The Official SaaStr Podcast: SaaS | Founders | Investors

This week Canva Co-Founder and COO Cliff Obrecht joined Harry Stebbings, Rory O’Driscoll and Jason with honest, unfiltered insights on scaling to 240 million users, navigating AI transformation, and preparing for public markets.

Canva’s CoFounder Joins 20VC + SaaStr on The Coming IPO, Where AI Works Today, Employee Liquidity, Figma’s IPO, and Much More!

After 13 years building Canva into a $4 billion ARR juggernaut, Cliff Obrecht’s key insight is deceptively simple: “In the end, the thing that bails out our incompetence is your growth rate.” Whether facing 50x valuations in 2021 or 10x today, the fundamentals remain constant — compound growth covers a multitude of sins, while everything else is just noise.

The $4B ARR Reality Check: What Actually Drives Growth at Scale — 90% is Organic

Canva will close 2025 “very close if not at $4 billion” in revenue, growing nearly 40% and reaccelerating after a post-2021 adjustment period. For Obrecht, this trajectory validates a contrarian approach to scaling that most B2B companies get wrong.

“One thing you need to buck the trend of as you become a larger company is insular thinking and treating your user base like a wet tea towel that you need to ring out,” Obrecht explains. “90% of our user acquisition is organic.”

The reacceleration story breaks down to three core drivers, with AI playing a surprisingly modest role:

1. Core Flywheel Optimization (70%) “We just needed to reaccelerate all our core flywheels. We spoke about paying up for the team.”

2. International Expansion (10%) “Going really heavy on international enhanced that.”

3. AI Integration (20%) “I would probably say 20% [of reacceleration comes from AI].”

This distribution challenges the narrative that AI is the primary growth driver for established SaaS companies. Instead, Obrecht’s thesis is that AI amplifies existing strengths rather than creating new ones.

The AI Integration Playbook: Workhorses, Not Gimmicks

Canva runs billions of AI inferences monthly, but Obrecht’s approach differs markedly from AI-first companies. The philosophy: “We’re all about creating workhorses, not gimmicks. AI is just accelerating [our mission] massively, making it quicker, faster, and better for customers to achieve their goals.”

The 10% GPU Tax Is Real

Like Notion, Canva is paying the “GPU tax” — roughly 10% of revenue going to AI infrastructure and model providers. “100% yes, it already is [10% of revenue],” Obrecht confirms. “If you look at Lovable, their pass-through to Anthropic or model providers will be way more than 10%.”

But this isn’t sustainable at current levels. Canva’s optimization strategy reveals how smart SaaS companies should think about AI costs:

Near-term Reality:

* Heavy compute costs for new AI products

* Pass-through pricing to OpenAI, Anthropic, others

* Unified credit model for usage management

Long-term Optimization:

* “We’re betting on distilling these models down, understanding user queries and where I need the frontier model versus where I can deploy the model that’s on-device or self-hosted”

* “Companies will get better at picking the right model for the right job”

* “We view some of those upfront costs as more of a marketing cost than a long-term enduring cost of goods”

The Credit Model Solution

Facing 240 million users and October’s “whole slew of new AI products,” Canva solved the margin problem with usage-based pricing layered onto subscriptions:

* Free users: Limited AI credits

* Premium subscribers: Expanded credit allocation

* Heavy users: Usage-based pricing beyond thresholds

“We need to maintain that margin,” Obrecht notes. This hybrid approach lets Canva capture AI value without destroying unit economics.

The Billion-Dollar Balance Sheet Strategy

Perhaps Canva’s most contrarian move: maintaining a billion-dollar cash position while profitable. “We’ve had a billion sitting on our balance sheet for ages — that’s a flex, people,” Obrecht says with characteristic directness.

This wasn’t accident but strategy, informed by hard experience:

The 2021-2022 Lesson:

* 2021: $40 billion valuation at 50x revenue

* 2022: Crashed to $26 billion as markets corrected

* Lesson: “The markets will do what they’re going to do, but as a company, we can compound growth, compound margins, and deliver value to customers”

The Insurance Policy Approach: “The truth is it’s a rocky journey. There’s probably some bumps ahead. I think most founders will be happier with a bigger balance sheet. The really great ones are the guys who can take the capital and then have the discipline not to use it foolishly.”

For Canva, cash isn’t just defensive — it’s offensive optionality for when markets inevitably shake out competitors who “pissed it all away in performance marketing.”

IPO Strategy: Employee Liquidity Trumps Everything

After 13 years and eight years of profitability, Canva doesn’t need public market capital. But employees need liquidity, and secondary markets aren’t cutting it.

“We’re 13 years old as a company. Our employees should have liquidity. They’ve created all this value. How can we make it easy for them to access that wealth that’s built up?”

The secondary market reality check is brutal:

Problems with Private Liquidity:

* “While secondaries, annual secondaries are a mechanism for that, it’s pretty janky”

* “Particularly in some jurisdictions, it’s downright impossible”

* Employees need employer permission for their own wealth access

* Supply-demand imbalances when major companies (like Figma) go public

Public Market Arbitrage: “The public markets now are valuing companies a lot higher [than private markets]. Ultimately the volume of capital dictates that multiple and there’s such an immense amount of capital being deployed in public markets that just is driving up those valuations.”

Canva brought in Zoom’s former IPO CFO Kelly and is “gearing up to be an IPO ready company.” The timing question remains open, but the infrastructure is being built.

Distribution Moats in the AI Era

While AI companies fight for attention, Canva’s distribution advantages compound. The company is now the #5 most-referenced domain on ChatGPT and the #1 productivity app in the platform.

“Over 5% [of images uploaded to Canva] now come from ChatGPT,” up from 0.02% eighteen months ago. This represents “LLM SEO” — winning discovery in AI platforms the same way companies optimized for Google search.

But Obrecht’s insight goes deeper: “As soon as these LLMs started taking off, we had the conversation: is our SEO team working on LLM optimization? And there’s definitely a team at Canva working on that.”

The lesson: Distribution advantages require active maintenance and adaptation, not passive hope.

The Mercenary Problem vs. Mission Alignment

Discussing Meta’s AI talent acquisition spree, Obrecht revealed Canva’s cultural philosophy: “When I was a B2B founder trying to be driven but touchy-feely, I was sort of anti-mercenary. If you’re not on my journey, I don’t want you. This is a long path — Canva’s been doing this for 20 years.”

This mission-first approach creates different talent dynamics:

Advantages:

* Lower churn during difficult periods

* Aligned decision-making during market volatility

* Sustainable culture at scale

Trade-offs:

* Potentially slower talent acquisition

* Higher bar for cultural fit

* May miss short-term optimization opportunities

For Canva, the trade-off makes sense: “Sometimes you need mercenaries and sometimes there’s a tool for the job. But I think this is… Zuck knows this is a bunch of mercenaries. Some of them are going to fall in battle.”

Pricing Strategy: From Seats to Consumption

AI is forcing SaaS pricing model evolution. Canva’s hybrid approach previews the future:

“We need to rethink our seat-based pricing model because some of the tools, particularly around marketing tools we’re creating, enable a single marketer to deploy tens of thousands of pieces of content.”

The new reality:

* Traditional seats: Base functionality at fixed price

* AI consumption: Usage-based pricing above thresholds

* Value alignment: Pricing scales with customer outcomes

“One person can do inordinate amounts of work and that’s using a huge amount of compute. You can’t charge 20 bucks a seat for that level of breadth.”

Key Takeaways for B2B Leaders

1. Growth Rate Forgives All Sins Market valuations will fluctuate wildly, but sustained growth compounds through any cycle. Focus on fundamentals, not noise.

2. AI as Accelerant, Not Strategy AI should amplify existing strengths, not become the entire value proposition. Workhorses beat gimmicks at scale.

3. Cash Is Strategic Flexibility In uncertain markets, balance sheet strength enables opportunistic moves while competitors struggle.

4. Employee Liquidity Drives IPO Timing For profitable companies, public markets provide liquidity solutions more than capital access.

5. Distribution Moats Require Active Defense Winning new platforms (LLMs) requires the same systematic approach as traditional SEO.

6. Mission Beats Mercenaries at Scale Cultural alignment becomes more valuable as companies mature and face inevitable challenges.

The meta-lesson from Canva’s journey: building at scale requires different muscles than early-stage growth. The companies that understand this distinction — and build accordingly — separate themselves from the pack.

Cliff Obrecht co-founded Canva in 2012 and serves as Chief Operating Officer. Under his leadership, Canva has grown to 240 million monthly active users and approaching $4 billion in annual recurring revenue.

The full convo here:

Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

We had a great one this week — Marc Benioff joined Harry, Rory, and Jason on 20VC+SaaStr this week to deliver one of the most grounded and passionate takes on AI we’ve heard from any enterprise leader. In a market fueled by AGI promises and $10 billion funding rounds, Salesforce CEO’s cut through the hype while revealing his company is quietly building a billion-dollar AI business — by focusing on practical applications over futuristic fantasies.

And Marc shared for the first time how AI is letting them follow up on 1,000,000+ leads their human sales team … never followed up on.

Bottom Line Up Front

AI is working at enterprise scale, but not always in the way the hype machine suggests. Benioff’s Salesforce has achieved over $1 billion in AI and data cloud revenue—their fastest-growing product ever—by deploying agentic systems that actually solve customer problems today. Meanwhile, the venture ecosystem continues pouring unprecedented capital into AI infrastructure plays that may struggle to justify their valuations without dramatic changes in enterprise spending patterns.

The Bottom Line Up Front:

* Benioff’s Reality: “I don’t think that there will be a piece of software that we sell that will not be agentic.” Salesforce achieved $1B+ AI revenue faster than any product in their history by focusing on practical applications rather than AGI promises, while redeploying 4,000 support agents to higher-value roles.

* Harry’s Concern: “I don’t feel like we’ve ever had the concentration of value tied to AI in seven companies as we have today.” The MAG-7’s unprecedented market concentration around AI creates systemic risk, while traditional growth metrics become meaningless when 10% growth at $40B scale adds an entire Palantir annually.

* Rory’s Math: “You actually need these things to take vast chunks out of the labor budget and be worth 20, 30, $40,000 almost a head to the enterprise for the math to work.” Foundation model valuations require AI agents to capture massive enterprise labor budgets—a scale that current use cases haven’t yet reached.

* Jason’s Evolution: “The fundamental architecture of an enterprise software company in the future is not exactly as it was in the past.” Companies must redesign organizational structures around AI capabilities, with 80% of VCs now refusing meetings with non-AI founders regardless of fundamentals.

The Reality Check We Needed

Benioff opened with a direct challenge to the AGI narrative: “You’re talking to somebody who is extremely suspect of anybody who uses those initials, AGI. I think that we have all been sold a lot of hypnosis around what’s about to happen with AI.”

This isn’t technological pessimism—it’s operational realism. Benioff acknowledged AI’s power while stripping away the mysticism: “Large language models are two things. They are a finite set of algorithms… and a relatively finite set of data that has come off the internet. Those two things together really provide kind of the state of the art of large language models today.”

The warning about over-reliance resonated particularly strongly. Benioff cited articles about doctors becoming “intellectually lazy” due to over-dependence on inaccurate AI, calling it “a huge warning sign for all of us around AI.”

The $1 Billion Proof Point and the 100 Million Lead Revolution

While others chase AGI dreams, Salesforce is monetizing AI reality. Their data cloud and AI combination has exceeded $1 billion in revenue and represents their “fastest growing cloud product ever in 26 years.” This isn’t incremental feature revenue—it’s a fundamental platform shift.

The numbers tell the story of practical AI deployment:

* Reduced support agents from 9,000 to 5,000 through agentic systems

* 100+ million historical leads now being contacted through AI-powered sales agents

* AgentForce handling as many customer interactions as human support agents

“This is a product that a year ago we hadn’t even announced. This is a product that wasn’t even shipped until November of last year,” Benioff noted, highlighting the unprecedented speed of enterprise AI adoption when the use cases actually work.

The 100 Million Lead Follow-Up Challenge

The most striking example of AI’s practical impact came from Benioff’s revelation about Salesforce’s own massive lead management problem. “Over the last 26 years, Salesforce has had more than 100 million people contact us that we’ve not been able to call back. We just have not had the people. That’s just all there is to it.”

This wasn’t a technology problem—it was a human capacity constraint that plagued even one of the world’s most successful software companies. Despite having “like 15,000 sales people,” Salesforce simply didn’t have enough SDRs to handle the volume of inbound interest. The math was brutal: 100 million leads over 26 years represents nearly 4 million leads annually that went completely uncontacted.

Think about the revenue implications. If even 1% of those 100 million leads could have converted to customers at Salesforce’s average deal size, that’s millions in lost revenue annually. This represents the classic enterprise software scaling problem: demand exceeding human capacity to respond.

The AI SDR Solution in Action

Benioff described how their AI SDR system now tackles this previously impossible challenge: “We have this agentic sales now. And not only are we doing support, but this agentic sales is calling everyone back and having conversations with them and then deeply integrating it through the omni-channel supervisor into our new agentic sales product.”

The system works through what Benioff called an “omni-channel supervisor”—an AI orchestration layer that coordinates between human agents and digital agents. This isn’t simple automation; it’s intelligent triage and routing that determines when human intervention is required and when AI can handle interactions independently.

The AI SDR process includes:

* Automated outreach to previously uncontacted leads from the 100 million backlog

* Conversation management through natural language interactions

* Qualification and scoring based on response patterns and engagement

* Seamless handoff to human SDRs when conversations reach complexity thresholds

* Integration with existing sales processes and CRM data

The “Customer Zero” Validation Strategy

Benioff emphasized that Salesforce serves as “customer zero” for their AI products, using their own 100 million lead challenge as the proving ground. This approach provides compelling advantages:

* Credible case studies: When selling AgentForce, Salesforce can point to their own massive lead follow-up success

* Real-world optimization: Internal usage reveals gaps and improvement opportunities before customer deployment

* Sales confidence: Representatives can speak from direct experience rather than theoretical benefits

* Risk validation: Salesforce absorbs the implementation risk before asking customers to do the same

The Talent War Reality

While Meta spends billions acquiring AI talent and Zuckerberg pays unprecedented premiums for teams, Benioff took a different approach: “No, and we’re not.” Instead of buying talent, Salesforce focused on redeploying existing headcount more effectively through AI augmentation.

This represents a fundamental shift in enterprise architecture. As both Marc and Jason noted: “The fundamental architecture of an enterprise software company in the future is not exactly as it was in the past.” Rather than replacing workers wholesale, successful AI implementation enables workforce optimization and value chain elevation. Thousands of headcount can be redeployed to new functions.

Palantir’s Enviable Pricing, and The Rise of the Forward Deployed Engineer

The Palantir discussion revealed perhaps the most telling moment of competitive envy from Benioff. Despite Salesforce’s $41 billion revenue dwarfing Palantir’s $4 billion, the market cap comparison stung: “That caught my attention. I’m like, how do I get that 100 times revenue multiple?”

The pricing revelation was even more striking. Benioff’s admission that Palantir’s publicly available price list made him question his own pricing strategy speaks to a fundamental shift in enterprise software economics. “I’m actually delivering, like, I’m automating the whole VA at this price. Like, what would they be charging? I mean, my prices are low compared to theirs.”

This isn’t just about individual deal sizes—it’s about market positioning. Palantir has successfully positioned itself as the premium AI solution for large enterprises, commanding prices that would be unthinkable for traditional SaaS products. They’ve broken through the conventional enterprise software pricing ceiling by selling transformation rather than tools.

The Forward Deployed Engineer Revolution

The forward-deployed engineer (FDE) concept represents a fundamental reimagining of the enterprise sales process. Traditional enterprise software follows a predictable pattern: demo, proof of concept, negotiation, implementation. Palantir flips this by deploying engineers before the deal is signed.

“We don’t have that kind of branding of, these are our forward deployed engineers where now we’re going to start building your product now before we’ve really signed a deal,” Benioff observed. “And I think that idea is very cool that all of a sudden you’re like in there kind of saying, yeah, we’re going to make a bet that we’re going to start doing business together. So we’re going to start building now.”

This approach solves the classic enterprise software chicken-and-egg problem. Customers can’t envision the solution until they see it working with their data, but vendors can’t build custom solutions without revenue certainty. FDEs bridge this gap by making the upfront investment in customer-specific development.

The model works particularly well for AI implementations because:

* Complexity Requires Customization: Unlike traditional SaaS where one-size-fits-most works, AI applications need deep integration with existing workflows and data structures.

* Proof Points Drive Adoption: Executive buyers need to see AI working with their actual data and use cases, not generic demos.

* Speed to Value: FDEs can achieve working prototypes in weeks rather than the months typically required for traditional enterprise software implementations.

* Premium Pricing Justification: The white-glove service model justifies significantly higher price points than self-service or standard implementation approaches.

Benioff’s interest in adopting the FDE model signals recognition that the enterprise software playbook is evolving. The companies winning the largest AI deals aren’t just selling software—they’re selling guaranteed outcomes through intensive human-AI collaboration during the sales process itself.

The broader implication is that enterprise software companies may need to fundamentally restructure their go-to-market organizations. Instead of separate sales, engineering, and professional services teams, the future may belong to integrated teams that can sell, build, and implement simultaneously. This requires significantly different hiring profiles, organizational structures, and economic models—but potentially unlocks the premium pricing that has made Palantir one of the most valuable enterprise software companies despite its relatively modest revenue scale.

The B2B Apps Survival Debate

Benioff delivered his most passionate response when addressing suggestions that SaaS applications would become “just CRUD databases.” He called this “one of the greatest disservices that has been done to our whole industry” and “crazy talk.”

His vision preserves application interfaces while adding agentic layers: “I need apps and I need agents and I need them to work together.” This represents the realistic middle ground between “AI replaces everything” and “nothing changes”—applications evolve with AI augmentation rather than disappearing entirely.

The Anthropic $10B Round Analysis

The discussion of Anthropic’s massive funding round revealed the disconnect between foundation model hype and enterprise adoption reality. While the round was “4x oversubscribed,” the math on market size remains challenging.

Rory’s analysis was particularly illuminating: For foundation models to justify these valuations, “you actually need these things to take vast chunks out of the labor budget and be worth 20, 30, $40,000 almost a head to the enterprise for the math to work.”

Even using Salesforce’s AI SDR as an example—potentially generating $3.6 billion in additional revenue with a 30% uplift—the LLM cost component at 20% of revenue would only represent $720 million flowing to foundation model providers. Scaling this across the enterprise software market suggests foundation model TAM may be smaller than current valuations imply.

The Consensus Investment Debate

Martin Casado’s tweet about consensus investing being “dangerous in early stage” sparked valuable discussion about AI investment strategies. The consensus view: most AI investments today are technically sound directional bets, even if individual companies fail.

As Jason observed: “When you’re on this mega trend of an architectural replatforming, a goodly amount of the correct investments to do are fairly consensus in terms of the broad macro themes.”

However, non-consensus bets face follow-on capital challenges. Jason’s advice to non-AI founders: “Don’t expect any money. 80% of the folks I can refer you to are not going to take your meeting.”

Market Timing Concerns

The concentration of value in AI-focused public companies raised reversion-to-mean concerns. As Harry noted: “I don’t feel like we’ve ever had the concentration of value tied to AI in seven companies as we have today.”

Yet recent earnings from MongoDB, Okta, Box and others showing AI-driven reacceleration suggest the enterprise software incumbents may finally be capturing AI value. This could provide more sustainable growth than pure-play AI companies facing eventual margin pressure.

Key Takeaways

* Enterprise AI is real and scaling fast: Salesforce achieved $1B+ in AI revenue faster than any product in their 26-year history

* Practical applications beat futuristic promises: Agentic support and sales systems deliver immediate ROI while AGI remains distant

* Workforce evolution is here and accelerating: Successful AI deployment rebalances and elevates human capabilities rather than eliminating jobs. But headcount may be radically repurposed in many segments (support, SMB sales, SDRs, etc).

* B2B apps will survive with AI augmentation: Applications evolve with agentic layers rather than being replaced by chat interfaces

* Foundation model valuations face math challenges: Current market caps require enterprise AI spending at unprecedented scales

* Consensus AI investing makes sense: Technical direction is clear even if individual companies face execution risk

* Follow-on capital flows to AI: Non-AI companies struggle to raise subsequent rounds regardless of fundamentals

Quotable Moments

Benioff on AGI hype: “You’re talking to somebody who is extremely suspect of anybody who uses those initials, AGI. I think that we have all been sold a lot of hypnosis around what’s about to happen with AI.”

Benioff on the future of software: “I don’t think that there will be a piece of software that we sell that will not be agentic.”

Harry on market concentration: “I don’t feel like we’ve ever had the concentration of value tied to AI in seven companies as we have today. And I am looking at it now going like, I really hope there’s not a blip here.”

Harry on growth at scale: “When you’re doing $40 billion, 10% growth is adding $4 billion, which is an entire Palantir every year.”

Rory on venture dynamics: “Venture is a game played by 6,000 people, and in the end, Sequoia wins.”

Rory on foundation model math: “You actually need these things to take vast chunks out of the labor budget and be worth 20, 30, $40,000 almost a head to the enterprise for the math to work.”

Jason on workforce evolution: “The fundamental architecture of an enterprise software company in the future is not exactly as it was in the past, that the fundamental architecture of the company will be different.”

Jason on follow-on capital reality: “80% of the folks I can refer you to are not going to take your meeting” (to non-AI founders).

*

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

When you think about Product-Led Growth (PLG) success stories, few companies exemplify the model better than Calendly. Founded in 2013 by Tope Awotona, Calendly has grown from a simple scheduling tool to a scheduling powerhouse that's touched "double digits" of the world's billion knowledge workers - meaning over 100 million people have used the platform at some point.

At SaaStr Annual + AI Summit, Awotona shared the hard-won lessons from building one of the most successful PLG companies of the last decade. What makes his insights particularly valuable is his transparency about the mistakes Calendly made along the way - and how they course-corrected.

Today, Calendly maintains a fascinating 90/10 revenue split between self-serve and sales-led motions, with customers ranging from individual contributors to million-dollar enterprise accounts. But getting that balance right took years of experimentation, data analysis, and some painful lessons about when PLG and enterprise motions can cannibalize each other.

Top 5 Key Learnings from Calendly's Journey

1. The Free Plan Is Your Marketing Engine - Protect It at All Costs

Perhaps the most counterintuitive insight from Awotona: Calendly spends "almost zero dollars on marketing campaigns." Their growth is entirely driven by user activity on the platform. Free users aren't just prospects - they're active marketing assets with measurable LTV.

"We're happy for people to use the product for free," Awotona explained. "Even free users have an LTV associated with them because we spend almost zero dollars on marketing campaigns."

The pressure to tighten free plans is constant. Sales teams consistently identify the free version as their #1 competitor, not any external product. But since putting up their first paywall in 2014, Calendly has never removed features from the free plan - they've only made it more generous.

The Takeaway: If your free plan drives viral growth, resist the short-term temptation to squeeze it. Instead, add more value to paid tiers rather than removing value from free ones.

2. Viral Loops Get Harder at Scale - But Scale Compensates

Calendly closely tracks two critical metrics: meetings-to-signups conversion rates and signups-to-activation rates (defined as five people scheduling with a new user). As expected, these conversion rates decline at scale - the viral coefficient naturally decreases as you reach market saturation.

But here's the key insight: "The good thing is the top line - the denominator - is getting bigger. That compensates a little bit for that degradation in conversion rate."

The Takeaway: Don't panic when viral coefficients decline at scale. Focus on optimizing the absolute numbers while understanding that percentage-based metrics will naturally compress as you approach market saturation.

3. Hybrid PLG + Enterprise Is Incredibly Hard to Get Right

This might be Awotona's most valuable insight for SaaS leaders. Calendly made both classic mistakes in balancing PLG and enterprise motions:

Mistake #1: Under-investing in enterprise and leaving seven-figure deals on the table Mistake #2: Over-investing in enterprise and having the sales team cannibalize PLG revenue

"What we found was the growth in our customer acquisition cost outpaced the incremental revenue growth. The enterprise business was really cannibalizing the PLG business," Awotona shared. By relaxing qualification rules to feed the sales team, they were simply converting self-serve prospects into sales-assisted deals - same revenue, higher cost, longer sales cycles.

The Takeaway: Be "incredibly analytical" with holdout groups and rigorous testing. The superficial metrics might look good while you're actually damaging your more efficient channel.

4. External-Facing Roles Are the Wedge Into Enterprise

Calendly's ideal customer profile focuses on external-facing roles: sales, customer success, recruiting. This represents about 25% of headcount in most organizations globally. But here's the strategic insight: these users become the wedge for enterprise expansion.

Their largest customer - a financial services company doing $1M+ in annual revenue - started with "a few people in the company using it. They loved it and then decided to expand." The expansion took 6-8 months from a sub-$20K footprint to seven figures.

The Takeaway: Identify the roles that get the most value from your product and use them as your enterprise wedge. Let the product prove itself with end users before enterprise sales gets involved.

5. Product Development Resource Allocation Must Match Business Model Evolution

With a 90/10 revenue split (PLG/Enterprise), how do you allocate product and engineering resources? Awotona's answer has evolved significantly:

* Initially: 90/10 allocation matching revenue

* 2020 Growth-at-all-costs era: Asymmetric allocation heavily favoring enterprise

* Today: "The pendulum has swung back" based on product maturity and diminishing returns analysis

"We look at the maturity of the product for different cohorts and analyze where we're hitting diminishing returns," he explained.

The Takeaway: Resource allocation should be dynamic, based on product maturity and ROI analysis, not just revenue percentages. Sometimes your PLG motion needs more investment despite generating most revenue efficiently.

Tope's Top 4 Mistakes (In His Own Words)

1. Over-Staffing Enterprise Sales During Growth-at-All-Costs Era

"We expanded that sales team greatly in 2020. Revenue grew, accounts over $100K grew, but customer acquisition cost outpaced incremental revenue growth. The enterprise business was cannibalizing the PLG business."

2. Under-Analyzing the Hybrid Motion

"I thought we were very analytical ourselves, but there's a lot more analysis we could have done, a lot more testing with holdout groups. When we ended up doing them, it was pretty eye-popping what we saw."

3. Relaxing Sales Qualification Rules Too Much

"We relaxed the qualification rules for letting people talk to a salesperson. What we would see is that revenue would convert, but if we just let them self-serve, a lot of them would have converted anyway - same revenue, but you add time and friction to the deal."

4. Not Being Strategic Enough About Product Allocation

"We've tried organizing teams by PLG vs Enterprise, by customer size, by capabilities. We've tried a number of different approaches. The most permanent design we've had is organizing by capabilities and having them own problems from individuals through enterprise."

The common thread in all these mistakes? Not being analytical enough. Awotona's biggest recommendation for founders building hybrid PLG-Enterprise motions: "Be incredibly analytical, do a lot of testing, use holdout groups."

What looks like success on the surface might be masking fundamental problems with your go-to-market efficiency. Sometimes the "best" sales reps are actually the worst when you analyze cohort retention. Sometimes growing enterprise revenue is actually just expensive PLG revenue in disguise.

The companies that crack the hybrid code - like Calendly eventually did - are the ones willing to dig deep into the data and make hard decisions based on what they find, not what they hope to see.

Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Gusto co-founders Josh Reeves (CEO) and Eddie Kim (CTO) came to SaaStr AI Summit to share their journey building “Gus,” an AI assistant now used by hundreds of thousands of small businesses.

Rather than chasing AI trends, they focused on solving real compliance pain points for small business owners navigating complex regulatory requirements across 50+ different state and local jurisdictions.

Their approach involved two key tracks: conversational interfaces that make software more intuitive, and automation that eliminates time-consuming tasks entirely. The result is an AI system that generates reports, executes actions like approving time-off requests, navigates complex compliance requirements, and creates optimal shift schedules. Gusto’s “startup within a startup” methodology, 90-day roadmap horizons, and hybrid interface philosophy offer practical lessons for any SaaS company serious about AI implementation.

Top 3 Takeaways from Gusto Co-Founders Josh Reeves & Eddie Kim:

* Create a “Startup Within a Startup” for AI Projects – Gusto built Gus by creating an independent team that operated outside normal engineering processes, shipped weekly, and didn’t even use project tracking tools initially. This allowed for the rapid experimentation needed in the fast-moving AI landscape.

* Focus on Problems Only You Can Solve – Be disciplined about which AI problems to tackle versus which ones the broader AI ecosystem will solve naturally. Gusto focused on Gusto-specific challenges while betting that general AI capabilities would improve on their own.

* The Future is Hybrid Interfaces, Not Just Conversational – While conversational AI is powerful for many tasks, the best user experience combines conversational and graphical interfaces at the right moments. Not everything is better done conversationally.

The Mission: Making Small Business Compliance Suck Less

Before diving into the technical details, Reeves made something crystal clear: “Companies don’t exist for the sake of it. We exist to go fix something, to go serve our customer and solve a pain point in their life.” For Gusto, that pain point is compliance hell.

Small business owners navigate a maze of local, state, and federal regulations. “The US in particular is more like 50 countries than one when it comes to all the different rules and requirements a business owner has to navigate,” Reeves explained. This isn’t about using AI because it’s trendy – it’s about using AI to solve a very real, very expensive problem.

Two Tracks: Conversational Interface + Automation

Gusto approached their AI strategy along two clear tracks that every SaaS company should consider:

Track 1: The Conversational Interface Revolution

“We are navigating a pretty massive paradigm shift in how software gets used,” Reeves noted. The conversational interface paradigm shift means customers can simply talk to software instead of figuring out which buttons to click or pages to navigate.

But here’s the key insight: this doesn’t replace traditional interfaces entirely. “We’re investing a lot of time in the Gusto web app, in our mobile app,” Reeves clarified. “We think they work in conjunction with each other.”

Track 2: Pure Automation

The second track focuses on literally automating tasks – both for internal team members (“Gusties”) and customers. “Think about something you spend 5, 10 minutes on. If it could just be done automatically for you, time savings is cost savings for a business owner.”

What Gus Actually Does (Beyond Just Answering Questions)

Kim broke down Gus’s capabilities, and they go far beyond typical chatbot functionality:

Advanced Reporting & Analysis: Gus generates reports and provides critical business insights, not just raw data dumps.

Action Execution: Tell Gus “I’d like to approve Sally’s time off request” and it will look up the request, show you details, and execute the approval upon confirmation.

Compliance Navigation: With tens of thousands of compliance rules affecting small businesses, Gus tells you exactly which ones apply to your specific business based on Gusto’s deep knowledge of your company.

Complex Scheduling: Gus can set up ideal shift schedules that automatically comply with break requirements, maximum daily hours, weekly limits, and other labor regulations.

The result? Gus is “generally available to every single one of our customers on Gusto’s platform” as of last week, actively being used by hundreds of thousands of small businesses.

The “Startup Within a Startup” Playbook

Here’s where Kim shared the tactical gold. Building AI products requires a fundamentally different approach than traditional SaaS development:

Start Fresh: “We started with a brand new codebase from scratch” and “politely excused ourselves from many of our engineering team’s rituals and processes.”

Move Fast, Track Later: “It wasn’t even until recently that we actually started to track our work in things like Jira and Asana.” Instead, they’d sit together, decide on priorities, and ship by week’s end.

Embrace Uncertainty: “We would just sort of sit all together in a room, talk about what are the most important things for us to work on. And by the end of the week, we would ship it.”

Three Hard-Won Lessons from Building Gus

Lesson 1: Vision Clear, Roadmap Fluid

“Although our vision for what problems we want to solve for the next few years is very, very clear… the roadmap of what we’re building cannot be more than 90 days out because we’re learning so much right now.”

Every three months, they learn something completely new that changes what they need to build. Traditional annual roadmapping doesn’t work in AI.

Lesson 2: Watch What Customers DON’T Do

While traditional SaaS metrics matter, Kim emphasized something more valuable: “It’s even more important to observe what customers are not doing on your product… what they’re doing right before they actually start using your product.”

He calls this “the work before the work” – the preparation customers do before using your product. “That is a gold mine of problems that you can be solving for your customers.”

Lesson 3: No Single Interface Rules Them All

“There is no single interface to rule them all. With the advent of AI, there’s so many things that are done better conversationally now, but not everything is done better conversationally.”

The future is hybrid – conversational at the right moment for the right task, graphical when that makes more sense.

The Internal Productivity Multiplier

Beyond customer-facing features, Gusto uses AI internally for massive productivity gains:

Customer Support Superpowers: Their support team uses internal AI tools to access information more quickly, reduce copy-pasting between documents, and pre-populate responses.

24/7 Self-Service: AI has made their help center “dramatically more efficient, more effective, better at getting you the information you need.”

Document Processing Automation: For tax filings, insurance enrollment, and other complex processes involving third-party documents, AI handles categorization and resolution. “Today, over half the tax notices that come in to the team get automatically flagged, organized, categorized, and then resolved without a human being involved whatsoever.”

The Business Impact: More Than Just Cool Technology

Reeves tied everything back to business fundamentals:

Increased Operating Margin: Better customer experience and internal productivity improvements create more operating margin to reinvest in new products and improvements.

Sustainable Growth: “Gusto’s been free cash flow positive for several years. We think if you’re at the scale we’re at, running a good business means you generate revenue, you generate free cash flow, you reinvest that, you build better product, you serve the customer better, and this technology basically adds to that flywheel.”

The Technology Tailwind Philosophy

Reeves closed with a broader perspective that every SaaS founder should internalize: “The building blocks constantly are changing.” When Gusto started 14 years ago, their technology tailwinds were cloud, paperless, and mobile. Their go-to-market tailwinds were social and search.

“AI is now a part of that and there’s going to be more in the future. And so the mindset here as a company builder is how do you not predict the future because no one has a crystal ball. But how do you stay on top of obsessively what the trends are, what the different technologies are.”

The key is matching technology capabilities with real customer problems: “It’s not using technology for the sake of it. It’s you’re looking at a customer problem, a customer pain point… And then if you think that technology can be a way for you to go make that pain go away to improve the quality of your experience, that’s a fantastic reason to dive in.”

Top 3 Mistakes to Avoid (Based on Gusto’s Experience):

* Don’t Try to Plan Too Far Ahead – Gusto learned that AI roadmaps can’t extend beyond 90 days because the technology and learnings evolve so rapidly. Traditional annual planning will leave you building the wrong things.

* Don’t Assume Conversational is Always Better – While conversational interfaces are powerful, not every task is better handled conversationally. Many things are still more effective with traditional graphical interfaces.

* Don’t Apply Traditional SaaS Development Processes to AI Projects – Heavy project tracking, rigid engineering processes, and traditional development rituals can kill the rapid experimentation needed for successful AI product development.

Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

The latest 20VC x SaaStr episode with Harry Stebbings, Jason Lemkin, and Rory O’Driscoll is here! The team is discussing Databricks’ $100B valuation, the coming IPO tsunami, CoreWeave’s massive debt raise, AI infrastructure spending that could hit trillions, the return of SPACs as a bubble signal, and why AI tool consolidation will happen faster than anyone expects

Bottom Line Up Front

Harry Stebbings: “We’re seeing the biggest wealth transfer in tech history unfolding before our eyes. When Databricks hits $100B and Andreessen could make $30B on a single deal, we’re not just in a bubble—we’re in a generational moment where the next wave of IPOs could make 2021 look like the appetizer.”

Jason Lemkin: “I think we’re all going to live in AI 24/7 and use 10 times the tokens and 10 times the compute in 24 months. The math is staggering—if we need 200x more infrastructure, how do we even finance that? But here’s the thing: we’ve already automated 5 humans with 10 AI agents at our small team. This isn’t theoretical anymore.”

Rory O’Driscoll: “All this depends on 3-5 more years of continued AI capex expansion. If that’s the case, everything works. If not, all bets are off. We’re way out on the risk curve, and the only thing between us and Armageddon is AI adoption continuing at this pace.”

Databricks Hits $100B: The New Normal for Private Valuations

When Databricks crossed the $100 billion valuation threshold this week, the most telling reaction wasn’t celebration—it was a collective shrug. As Jason noted, “If you’d said 5 years ago there’s going to be a hundred billion dollar market cap private company, you’d be like no way. Now you’ve got Anthropic at $170B, SpaceX at $360B, OpenAI at $500B. The correct response is yeah, whatever.”

But beneath this seemingly blasé attitude lies a fundamental shift in how we value AI infrastructure companies. Databricks is now worth ~50% more than Snowflake while growing 2x as fast—crossing $4 billion ARR at 50% growth versus Snowflake’s $4 billion at 26% growth. At 25x revenue, it actually feels undervalued given the growth trajectory and AI positioning.

The real story here isn’t just another unicorn—it’s about generational wealth creation concentrated in private markets. If Andreessen Horowitz led Databricks’ seed round in 2013 with follow-on investments across multiple funds, they could own 15% of a company heading toward a $200B IPO. That’s potentially $30 billion in returns from a single deal.

“If A16Z invested out of a $650 million fund at the time and they’ve turned a billion and a half into $30 billion, you’re going to feel good in the morning,” Rory observed. “The one thing we don’t take enough into account is just how many people make money when these go out. There are so many LPs in SPVs, SPVs on SPVs. There are dentists who are going to make 5x.”

The Coming IPO Tsunami: Why This Is Just the Beginning

What makes this moment particularly significant is that we haven’t seen the epic IPOs yet. CoreWeave, Circle, Hinge Health, even Figma—these are still “niche plays” compared to what’s coming. As Harry pointed out, “The mega ones are still to come. We might get the next wave of IPOs that are even better very quickly, and the amount of froth that could create in the ecosystem could create a bubble on top of a bubble.”

Consider the pipeline: If Canva IPOs next year at $4 billion ARR growing 40% (versus Figma at $1 billion growing 48%), and if Databricks goes public at an implied $200B valuation, we’re looking at a wealth creation event that dwarfs anything we’ve seen before.

“There’s a significant absolute sum locked up in private markets that at some point is going to seek a public market where a number of them will be worth $50 billion, which is unprecedented. A few will be worth $100 billion and maybe one will be worth a trillion dollars in the private markets,” Rory noted. That trillion-dollar private company prediction isn’t hyperbole—it’s math.

CoreWeave’s $11B Debt Bet: Canary in the Coal Mine or Smart Capital Allocation?

While everyone celebrates Databricks’ equity valuation, CoreWeave’s $11.2 billion debt raise tells a different story about AI infrastructure financing. The company spent $22 billion in capex this year while generating $1 billion in quarterly revenue—a capital intensity that would make even utilities blush.

“Once you decide to spend $22 billion in capex, you’re going to have $11 billion in debt,” Rory explained. “The way you think about CoreWeave is they’re the guys doing the pointy end of the bet that Meta, OpenAI, Anthropic are talking about—we want to deploy a ton of GPUs and data centers. CoreWeave is stepping up as the financing vehicle.”

But this creates a fascinating risk dynamic. CoreWeave becomes the canary in the coal mine for AI demand. “You’re not going to see it in Microsoft. OpenAI isn’t public. You’re not going to see it in Google—they’re too big. But you could see that canary in the coal mine in this public one due to exposure,” Jason observed.

The company’s debt structure depends entirely on matching long-term customer commitments with long-term financing. If Microsoft and OpenAI have seven-year take-or-pay contracts and CoreWeave has seven-year debt, the math works. But if customers start paying rather than taking, or if new data center construction slows, CoreWeave stock could signal broader AI infrastructure headwinds before they show up in hyperscaler earnings.

The Return of SPACs: Peak Bubble Signal?

Nothing says “bubble” quite like Chamath Palihapitiya announcing a new SPAC with the disclaimer that if investors lose everything, they should “embody the adage from President Trump that there can be no crying in the casino.”

“It does seem to remarkably correlate—the return of Chamath to the SPAC market appears to correlate with the most recent bubble in 2021 where it worked in the bubble and didn’t work after,” Rory noted. But the bigger issue isn’t timing—it’s structure. SPACs create adverse selection because promoters get paid when deals get done, regardless of quality.

“If you pay people to do deals, deals get done regardless of quality,” Rory explained. “I just think it’s an imperfect vehicle for fundraising, and the fact that it’s back is indicative of bubbly times.”

The casino metaphor is particularly galling because capital markets aren’t supposed to be zero-sum games. “The point of Wall Street is to funnel money from savers into profitable long-term investments. It’s not a freaking casino, which is a zero-sum game where only the house wins.”

AI Labor Replacement: The Trillion-Dollar Question

Sam Altman’s prediction that AI infrastructure will require “trillions” in spending sounds hyperbolic until you consider the scale of potential labor replacement. Jason’s team has already replaced 5 humans with 10 AI agents, and their AI tool costs are approaching $500,000 annually—heading toward $1 million.

“I think we’re all going to live in AI 24/7 and use 10 times the tokens and 10 times the compute in 24 months,” Jason predicted. “If we’re using AI 20 times more a day and 10 times more tokens, how much more do we need to spend on infrastructure? What’s 200 times today?”

But Rory pushes back on the trillion-dollar timeline: “The $365 billion being invested by the four most profitable companies on the planet is already straining their balance sheets. We’re reaching the limits of what people want to finance.” Even if the direction is right, the pace matters enormously when you’re talking about 4-5% cost of capital on hundreds of billions in investment.

The key insight is that AI adoption needs to transition from technology budgets to human labor budgets to justify trillion-dollar infrastructure spending. “If you see that transition, you’ve got $10 trillion of value unlocked, and then spending trillions becomes a lot more feasible,” Harry noted.

The Consolidation Wave Is Coming Faster Than Expected

One of the most practical insights from the discussion is that AI tool consolidation will happen much faster than the SaaS consolidation cycle. When Jason’s small team is already spending $500,000 annually on AI agents, with every B2B AI startup trying to charge $60-100K per year, the math becomes unsustainable quickly.

“We may see a wave of consolidation in AI even next year because it’s just too many AIs that are six figures and up,” Jason predicted. “Everyone thinks their thing is going to save labor, but you can’t all get credit for the same labor.”

This creates a different dynamic than traditional SaaS, where companies could afford to have dozens of specialized tools. In AI, you might want everything from one vendor in 12 months, which favors companies like Rippling that can bundle multiple AI agents with an orchestration layer.

The implication for venture investing is profound: “We’re ignoring so many risks in AI B2B. Platform risk is everywhere, and we’ve given up worrying about it because the growth is so attractive.”

Fintech’s Global Winners: The Nubank Story

While everyone focuses on AI, Nubank’s $2.5 billion quarterly profit (up 42% year-over-year) with 123 million customers shows what winning looks like in fintech. The Brazilian neobank is now worth $60 billion and has essentially become the primary bank for an entire generation.

“New Bank has 35% market share of Gen Z in Brazil, and Revolut has almost 20% of everyone in the UK with a Revolut account,” Jason noted. This isn’t just fintech disruption—it’s complete market capture.

The success pattern is consistent across geographies: find the weakest, most inefficient part of incumbent banking infrastructure and attack it relentlessly. Nubank went full-stack banking in Latin America, Revolut started with FX in Europe, and Chime focused on deposit arbitrage in the US. The bigger the opportunity (weaker the incumbents), the bigger the outcome.

Revolut’s product expansion strategy is particularly instructive: 26 new products in development, run like internal venture teams with separate funding goals and weekly testing metrics. As Harry observed, “Out of all the founders I’ve interviewed—hundreds and thousands—Nick [Storonsky] is the best I’ve ever met at this playbook.”

Key Takeaways

* Private market valuations aren’t crazy when you adjust for growth rates and AI positioning—Databricks at 25x revenue growing 50% actually looks reasonable compared to public comps

* The next 12-18 months could see the biggest IPO wave in history, with companies like Canva, Databricks, and Stripe potentially going public at unprecedented valuations

* CoreWeave’s debt levels make it the leading indicator for AI infrastructure demand—watch their quarterly results for early signals of market shifts

* AI tool consolidation will happen much faster than SaaS consolidation due to budget constraints and the overlap in labor-saving claims

* Trillion-dollar AI infrastructure spending is directionally correct but timeline matters—the pace of adoption versus capital deployment costs will determine winners and losers

* Successful fintech companies capture entire demographic cohorts, not just market share—35% of Gen Z isn’t just usage, it’s generational platform lock-in

* We’re all living in an AI bubble whether we realize it or not—QQQ and VTI exposure means everyone is betting on continued AI infrastructure expansion

Quotable Moments

On the IPO pipeline: “The mega ones are still to come. We might get the next wave of IPOs that are even better very quickly, and the amount of froth that could create in the ecosystem could create a bubble on top of a bubble.”—Harry Stebbings

On bubble dynamics: “If you’d seen a $100 billion private company valuation a year ago, I would have fallen out of my chair. Now it’s funny how I don’t blink an eye in the age of AI.” —Jason Lemkin

On wealth creation: “When Andreessen could make $30 billion on the Databricks IPO, how does that further reinforce the cycle we’re in? 30 billion is a lot.” —Harry Stebbings

On AI infrastructure: “CoreWeave are the guys doing the pointy end of the bet that Meta, OpenAI, Anthropic are talking about. They’re stepping up as the financing vehicle to make that happen.” —Rory O’Driscoll

On consolidation timing: “We may see a lot of these AI folks get consolidated out sooner than happened in the SaaS wave. Everyone wants an agent for their department. They want five or six agents.” —Jason Lemkin

On market risk: “We’re way out on the risk curve, every one of us, right now. The only thing between us and Armageddon is AI adoption. If AI adoption keeps happening, this wagon train keeps rolling. If it slows down, it’s going to get ugly real fast.” —Rory O’Driscoll

On fintech disruption: “New Bank is old bank. If the CEO of Bank of America went to Brazil today, he’d be like, ‘I know exactly what these guys are doing’—banking the middle class, top to bottom.” —Jason Lemkin

On the trillion-dollar question: “Sam Altman said they’re going to spend trillions on infrastructure. Do you think that’s a real number or a metaphor? No company on the planet has spent a trillion dollars in capex ever.” —Rory O’Driscoll

Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

At SaaStr's packed AI Summit 2025, Box CEO Aaron Levie and IBM VP of AI Raj Datta did a deep dive together with SaaStr's Jason Lemkin on how B2B companies should think about AI agents. With 10,000 attendees—a massive jump from last year—the energy around AI agents was electric.

Here are the top 10 learnings from their convo:

1. AI Agents Represent a Fundamental Shift from Software to Digital Labor

The Key Insight: We've moved beyond chat interfaces to AI that actually performs work autonomously.

Levie explained the evolution: "We've had AI models for five-plus years. Then we had assistants like ChatGPT. But now we're seeing agents that fundamentally go do work for you—and that work could take a minute, an hour, or 100 hours for the agent."

For B2B companies, this changes everything. Instead of selling software to 10 lawyers in a company, you're now selling "infinite legal capacity." IBM proved this works at scale, saving $3.5 billion internally through AI agents handling HR and procurement functions.

Takeaway: Start thinking about your software as digital labor, not just tools. What work can your AI agents do autonomously for customers?

2. Your Customer Data Is Your Biggest Competitive Advantage

The Key Insight: The companies with the richest, most proprietary datasets will win in the AI agent era.

"What data are you sitting on that is proprietary to you?" Levie asked. "Very quickly you realize more companies are actually in the data business than they initially thought."

Box exemplifies this perfectly. Customers who've stored documents for years can now ask complex queries like "Tell me everything where I have the wrong indemnity provision" or "What contracts I shouldn't have signed." This transforms static data into dynamic business intelligence.

Takeaway: Audit your proprietary data assets. What unique insights could AI agents extract that your competitors can't replicate?

3. Enterprise AI Adoption Is Happening 1000x Faster Than Cloud

The Key Insight: Unlike cloud adoption, which took over a decade, AI is being embraced immediately by enterprises.

"It's going 1000x faster than cloud adoption because everyone's using it," Datta noted. While pitching cloud to banks in 2008-2009 was a "non-starter," today there isn't an enterprise that doesn't already have an AI strategy in development.

The speed difference is staggering: ChatGPT reached 500 million users in roughly two years—faster than any technology in history.

Takeaway: Your enterprise customers are ready to buy AI solutions now. The question isn't if they'll adopt AI, but which vendor they'll choose.

4. IBM's Agent Catalog Could Revolutionize B2B Distribution

The Key Insight: Even small B2B startups can now access enterprise sales channels through IBM's new agent marketplace.

"Even if you're a five-person shop, you submit your AI agent to our catalog, and IBM sellers are able to sell it for you," Datta explained. This democratizes enterprise sales in unprecedented ways.

IBM's massive sales force essentially becomes the distribution channel for thousands of specialized AI agents, potentially creating the "App Store moment" for enterprise software.

Takeaway: Explore partnerships with larger platforms that could distribute your AI agents. Distribution partnerships may be more valuable than ever.

5. The Next Generation Will Completely Reshape B2B Work

The Key Insight: AI-native workers entering the workforce will challenge every existing business process.

A Stanford student approached the moderator saying "20% of my class has already dropped out to do AI" after building an AI sales agent that reached $2 million ARR in months.

Levie predicted: "They're going to come into our enterprises and say, 'You take 2 weeks to come up with a marketing plan? That was just generated by Claude in 5 seconds. Why would you meet about that?'"

Takeaway: Prepare for a workforce that expects AI-first processes. Your internal tools and customer solutions need to match this expectation.

6. Data Assets May Soon Appear on Corporate Balance Sheets

The Key Insight: We're heading toward a fundamental revaluation of what makes B2B companies valuable.

"Right now there's nothing on a company's balance sheet that approximates the value of their data," Levie observed. "In 10 years, will we see 'how good is your data?' emerge as a measurable business asset?"

This suggests enterprise valuation models will need to account for data quality, uniqueness, and AI-readiness.

Takeaway: Start thinking about your data as a balance sheet asset. Clean, organized, proprietary data will become increasingly valuable.

7. You Have a 2-Year Window to Win or Lose Everything

The Key Insight: The AI transition window is much shorter than cloud, creating extreme risk and opportunity.

"You could lose your entire position probably in a 2-year period," Levie warned, "but it's also where you could cement your leadership position in a very unique way."

Unlike cloud, where companies had years to decide, AI advantages will compound quickly. The winners will get locked in fast.

Takeaway: Move aggressively now. Waiting 6-12 months could mean losing your market position permanently.

8. Model Context Protocol (MCP) Strengthens Rather Than Threatens Moats

The Key Insight: Making your data more accessible actually increases customer stickiness.

When questioned about whether MCP would weaken competitive advantages, Levie argued the opposite: "The more places where customers can use their data is only a net positive. We want to connect to every MCP environment possible."

This mirrors how APIs strengthened platforms by creating more integration points, not weakening them.

Takeaway: Embrace interoperability standards. They'll make your platform more valuable, not less defensible.

9. AI Agents Work Best When Customers Don't Know They Exist

The Key Insight: The best B2B AI implementations are invisible to end users.

Datta emphasized: "There'll be thousands of agents working in the background, but typically one interface for the client. You don't know if you're talking to SuccessFactors or other products—you're just talking to the Watsonx bot."

Customers want outcomes, not technology. They don't care about having "a thousand agents"—they want their problems solved seamlessly.

Takeaway: Hide the complexity. Focus on delivering better outcomes, not showcasing your AI architecture.

10. Traditional Business Fundamentals Now Matter More Than Ever

The Key Insight: As AI capabilities become commoditized, execution and customer focus become the primary differentiators.

"The things that transcend AI will become even more important," Levie concluded. "Building a company, hiring a sales team, iterating, building better products, serving customers—the lowering of barriers to build doesn't make it easier to build sustainable businesses."

While AI democratizes software creation, it doesn't make building great companies any easier. Long-term thinking, customer obsession, and execution excellence remain paramount.

Takeaway: Don't let AI distract from business fundamentals. The companies that win will combine great AI with great execution.

The Bottom Line for B2B Leaders: AI agents aren't just the next feature to add—they represent a fundamental shift to selling digital labor instead of software tools. The companies that recognize this shift and act on it in the next 24 months will cement dominant positions in their markets. Those that don't risk becoming irrelevant.

The race is on. Your customers are ready to buy AI solutions now. The question is whether you'll be the vendor they choose, or whether a competitor—potentially one that didn't exist six months ago—will capture that opportunity instead.

Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

How We Built ChatGPT Enterprise's Sales Team from Absolute Zero: The Complete Playbook with Maggie Hott, GTM Leadership at OpenAI

About Maggie: Maggie Hott has spent 15 years building go-to-market teams at four unicorns that collectively represent over $50B in market value. She started as the 2nd SDR at Eventbrite, became the first sales hire at Slack (helping scale from $50M to $1B ARR and a $27B Salesforce acquisition), served as Director of Sales at Webflow (scaling from $40M to $140M ARR), and now leads go-to-market at OpenAI where she built ChatGPT Enterprise from scratch. She also runs a venture fund with seven other women investors, backing 30+ founders. These are her personal views, not those of OpenAI.

It was early 2023. OpenAI had just launched ChatGPT, the fastest-growing consumer app the world had ever seen. We were riding an incredible wave, but we had a critical hypothesis: ChatGPT Enterprise would require a fundamentally different go-to-market motion than our API business.

When OpenAI hired me to build ChatGPT Enterprise from scratch, I walked into what can only be described as a beautiful blank slate—and a terrifying challenge.

Our entire sales and go-to-market organization was less than 10 people. No SDRs. No solution consultants. No customer success managers. No sales operations. No RevOps. No marketing enablement. We didn't even have a working Salesforce instance.

What we did have: six incredibly talented account directors and one technical success partner, all laser-focused on selling our API to developers and technical teams.

Here's the complete playbook for how we built what we believe became the fastest-growing enterprise application in history.

The Build vs. Adapt Strategic Decision

Most companies in our position would have taken the "efficient" approach: enable the existing team to sell both products, maybe hire a few specialists, and gradually expand capabilities.

We chose the opposite path: build a dedicated ChatGPT Enterprise team from absolute zero.

This wasn't just about headcount. It was about creating an entirely separate organizational DNA optimized for enterprise selling.

Why We Built Separate vs. Adapted Existing

Product Complexity Was Fundamentally Different

* API sales required deep technical conversations about integrations, rate limits, and model parameters

* ChatGPT Enterprise needed business impact discussions about productivity, compliance, and organizational change management

* The buyer personas didn't overlap—CTOs vs. CHROs, CFOs, and business unit leaders

Sales Cycles Had Different Rhythms

* API deals often moved quickly with technical evaluation periods

* Enterprise required lengthy security reviews, compliance discussions, and change management planning

* Different stakeholders, different timelines, different objection patterns

Go-to-Market Motions Required Different Muscles

* API was largely product-led with sales-assist for larger accounts

* Enterprise needed traditional enterprise selling: demos, pilots, RFP responses, and executive alignment

Speed Trumped Efficiency in This Moment The market window was massive but narrow. Every week mattered. Having a dedicated team meant:

* No competing priorities or split focus

* Ability to move at startup speed even within a scaling company

* Clear ownership and accountability for outcomes

Phase 1: Foundation Building (Months 1-3)

The First Three Critical Hires

Enterprise Account Executive #1: Financial Services Specialist

* 8+ years selling enterprise software to banks and insurance companies

* Deep understanding of compliance requirements (SOX, PCI, etc.)

* Existing relationships with CISOs and risk management teams

* Experience with 12+ month sales cycles and complex procurement processes

Enterprise Account Executive #2: Technology Sector Specialist

* Background selling to high-growth tech companies

* Understanding of developer tools and technical infrastructure

* Experience with both startup buyers and enterprise technology teams

* Ability to bridge technical and business conversations

Enterprise Account Executive #3: Healthcare/Life Sciences Specialist

* Healthcare technology sales background

* HIPAA compliance expertise

* Relationships with healthcare CIOs and innovation teams

* Understanding of clinical workflow integration challenges

Why Vertical Specialists First? Enterprise buyers expect deep industry knowledge. They want to know you understand their specific compliance requirements, regulatory challenges, and business context. Hiring generalists would have slowed our credibility-building process by months.

Building the Foundational Systems

Customer Qualification Framework We couldn't use our API qualification criteria. Enterprise buyers had different needs:

* Company Size: 1,000+ employees (later expanded down to 500+)

* Budget Authority: Direct access to decision-makers with budget

* Use Case Clarity: Specific productivity or efficiency goals

* Security Readiness: Existing enterprise software deployment experience

* Timeline: Willingness to run pilots and structured evaluation processes

Initial Pricing and Packaging Structure Started with three tiers based on our early customer research:

* Starter: Small teams, basic enterprise features

* Business: Department-wide deployment, advanced security

* Enterprise: Organization-wide, full compliance and customization

Basic Sales Process Framework

* Discovery Call: Understand use case, stakeholders, and decision process

* Technical Demo: Customized demo showing specific business scenarios

* Security Review: Deep dive on compliance, data handling, and enterprise requirements

* Pilot Program: Structured 30-60 day pilot with clear success metrics

* Business Case Development: ROI modeling and executive presentation

* Contract Negotiation: Enterprise terms, security requirements, implementation planning

The Customer Success Foundation

Why Customer Success from Day One Enterprise deals aren't won when the contract is signed—they're won when the customer is successfully deployed and seeing value. We hired our first Customer Success Manager in month two.

First CSM Profile

* Enterprise software implementation experience

* Change management background

* Technical enough to understand AI/ML concepts

* Business-focused enough to measure productivity impact

Phase 2: Systems and Process Scaling (Months 4-9)

Adding the SDR Engine

Why SDRs for Enterprise? Enterprise deals require extensive relationship building and multi-threading. Our AEs needed to focus on advancing qualified opportunities, not prospecting.

First SDR Hires (3 people)

* Industry Focus: Each SDR aligned with our AE verticals (Financial Services, Technology, Healthcare)

* Enterprise Experience: All had experience prospecting into large organizations

* Multi-Threading Skills: Ability to identify and engage multiple stakeholders

SDR Success Metrics

* Qualified meetings booked (not just meetings)

* Multi-stakeholder engagement (average 2.3 contacts per account)

* Account penetration depth (director level and above)

Implementing Proper Sales Operations

The Salesforce Build-Out This was bigger than just CRM implementation. We needed:

* Custom Objects: Enterprise-specific fields for compliance requirements, use cases, stakeholder mapping

* Workflow Automation: Lead routing, opportunity progression, approval processes

* Integration Stack: Marketing automation, customer success platforms, billing systems

* Reporting Infrastructure: Pipeline forecasting, activity tracking, conversion metrics

Sales Operations Hire

* Background: Enterprise SaaS sales operations experience

* Technical Skills: Salesforce administration, data analysis, process automation

* Strategic Thinking: Ability to design scalable processes for rapid growth

Building Enablement and Competitive Intelligence

Sales Enablement Program

* Industry Training: Deep dives on financial services, healthcare, and technology sector challenges

* Competitive Intelligence: Detailed battlecards for Microsoft, Google, and emerging AI competitors

* Demo Certification: Standardized demo flows for different use cases and industries

* Objection Handling: Frameworks for common enterprise concerns (security, compliance, change management)

Content Creation

* Industry-Specific Case Studies: Early customer success stories by vertical

* ROI Calculators: Productivity impact modeling tools

* Security and Compliance Documentation: Detailed technical specifications for enterprise buyers

* Executive Briefing Materials: Board-ready presentations on AI strategy and implementation

Phase 3: Full Enterprise Motion (Months 10-18)

Scaling the Core Team

Additional AE Hires

* Expanded to 12 enterprise AEs across verticals

* Added government/public sector specialist

* Hired retail/e-commerce focused rep

* Added manufacturing/industrial specialist

Solution Engineering Team

* Technical Specialists: Deep AI/ML expertise for technical evaluations

* Industry Solutions Engineers: Vertical-specific implementation expertise

* Demo Engineers: Specialized in creating compelling enterprise demonstrations

Expanded Customer Success

* Implementation Specialists: Focused on enterprise deployment and change management

* Strategic CSMs: Relationship management for largest accounts

* Technical Success Engineers: Post-deployment optimization and advanced use case development

Channel Partnerships and Ecosystem

System Integrator Partnerships

* Deloitte: Enterprise AI strategy and implementation services

* Accenture: Large-scale transformation projects

* IBM: Hybrid cloud and enterprise integration

* PwC: Risk management and compliance implementation

Technology Partnerships

* Microsoft: Azure integration and enterprise infrastructure

* Salesforce: CRM integration and workflow automation

* ServiceNow: IT service management and workflow optimization

* Slack/Teams: Productivity platform integrations

Marketing Engine Development

Account-Based Marketing Program

* Target Account Lists: 500 highest-value prospects per vertical

* Personalized Campaigns: Industry-specific content and outreach

* Executive Events: CIO roundtables and AI strategy workshops

* Webinar Series: Industry-specific use case demonstrations

Content Marketing Strategy

* Industry Reports: AI adoption trends by vertical

* Executive Whitepapers: Strategic guides for AI implementation

* Customer Success Stories: Detailed case studies with ROI metrics

* Thought Leadership: Speaking engagements at industry conferences

Phase 4: The Great Integration (Months 19-24)

The Bold Organizational Decision

By early 2024, we faced a new challenge: many customers wanted both API and ChatGPT Enterprise capabilities. Having separate teams was creating customer confusion and internal inefficiencies.

We made a dramatic decision: unify both organizations into a single go-to-market team.

The Integration Process

The Scope: 500 people across both teams The Challenge: Most people had been at OpenAI less than 6 months The Change: Everyone got new roles, new managers, new workflows, and had to learn the opposite product

Integration Timeline

* Week 1: Announced the change and new organizational structure

* Week 2-4: Individual role assignments and team restructuring

* Month 2: Cross-training program launch (API sellers learning ChatGPT Enterprise, ChatGPT Enterprise sellers learning API)

* Month 3: New process implementation and system integration

* Month 4-6: Performance optimization and culture integration

The Results

* Faster Execution: Eliminated duplicate processes and conflicting priorities

* Better Customer Experience: One team, one relationship, unified solution selling

* Reduced Costs: Eliminated redundant systems and overlapping functions

* Enhanced Capabilities: Every seller could now handle both simple API needs and complex enterprise requirements

Integration Lessons Learned

What Worked

* Clear Communication: Transparent about the why behind the change

* Fast Timeline: Ripped the band-aid off quickly rather than prolonged transition

* Investment in Training: Intensive cross-product education program

* Leadership Alignment: United leadership team modeling the integration

What Was Challenging

* Learning Curve: Everyone had to master new products and processes simultaneously

* Cultural Integration: Merging two different team cultures and ways of working

* System Complexity: Integrating different CRM instances and operational tools

* Customer Communication: Managing customer relationships during the transition

The Key Success Factors

Hire for Chaos, Not Comfort

Every early hire was a "chaos translator"—someone who thrived in ambiguous, rapidly changing environments. We specifically avoided people who needed clear processes and defined roles. Instead, we found builders who could create structure from nothing.

Maintain Extremely High Hiring Standards

Despite the pressure to scale quickly, we never compromised on quality. One exceptional enterprise seller was worth three average ones, especially in the early days when they were helping define our entire go-to-market approach.

Invest in Industry Expertise Early

Generic enterprise selling skills weren't enough. The investment in vertical specialists from day one paid massive dividends in credibility, deal velocity, and win rates.

Build for Scale from the Beginning

Even with a small team, we designed processes and systems that could handle 10x growth. This meant more upfront investment but prevented costly rebuilds later.

Embrace Bold Organizational Changes

The integration decision was risky and painful, but it was the right move for customers and the business. In AI, the pace of change requires constant organizational evolution.

The Numbers Behind the Success

While I can't share specific revenue figures, here are some metrics that demonstrate the impact:

* Team Growth: 10 to 500 people in 24 months

* Win Rate: Maintained >60% win rate even during rapid scaling

* Sales Cycle: Average enterprise deal closed in 4-6 months (fast for enterprise AI)

* Deal Size: Average contract value increased 5x through pilot program optimization

* Customer Success: >95% of enterprise customers expanded usage within first year

* Integration Success: Post-integration, unified win rates increased 15% over separate teams

What I'd Do Differently

Start with Sales Operations Earlier

We should have hired sales operations in month one, not month four. The operational debt we accumulated in early months took significant effort to clean up.

Invest More in Change Management

Enterprise AI adoption requires significant organizational change management. We should have built change management expertise into our customer success function earlier.

Build Competitive Intelligence Faster

The enterprise AI landscape evolved incredibly quickly. We should have invested in systematic competitive intelligence gathering and analysis from day one.

Create More Structured Onboarding

With rapid hiring, our onboarding process became inconsistent. A more structured program would have reduced time-to-productivity for new hires.

The Broader Lessons for Building Enterprise Teams

Product-Market Fit Looks Different in Enterprise

Consumer product-market fit is about engagement and retention. Enterprise product-market fit is about business impact and organizational change. The metrics and success criteria are fundamentally different.

Enterprise Buyers Want to Buy from Experts

Generic enterprise selling doesn't work in specialized markets like AI. Industry expertise and technical depth are table stakes, not differentiators.

Speed Matters More Than Perfection

In rapidly evolving markets, the team that moves fastest often wins, even if their processes aren't perfect. Iteration speed trumps initial optimization.

Culture Scales Through People, Not Processes

Our best cultural decisions were hiring decisions. The right people created the right culture, which then influenced all our processes and systems.

Customer Success is Revenue, Not Cost

In enterprise software, customer success directly drives expansion revenue, renewal rates, and reference-ability. It's not a support function—it's a growth engine.

Building ChatGPT Enterprise from zero taught me that enterprise go-to-market isn't just scaled-up SMB selling. It requires different people, different processes, different systems, and different organizational structures.

The most important lesson: in fast-moving markets like AI, organizational agility matters more than organizational perfection. The teams that can build, scale, and adapt quickly will win.

Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

The latest 20VC + SaaStr: The Figma IPO Breakdown: $3B Left on the Table, Broken CEO Comp, and Why VCs Are Worse Than Public Markets

This week, Brian Halligan (Co-Founder & Executive Chairman of HubSpot) joins the regular crew of Harry Stebbings, Rory O’Driscoll, and Jason Calacanis for an inside look at IPO dynamics, CEO compensation, and the current state of public markets.

The Bottom Line Up Front

On IPO Pricing Reality: “The people who said, ‘Oh, Figma left $3 billion on the table.’ The $98 price only happened because the IPO happened at $38. Had someone walked in and said, I know this IPO is going to price at $100 bucks a share to open tomorrow morning. Let’s raise at $80. They wouldn’t have had a book becayse no one has bid at that thing. So that money wasn’t accessible.” — Rory O’Driscoll

On Going Public vs. Staying Private: “I think VCs are a much bigger pain in the ass than public investors. VCs are a much bigger pain in the ass than the typical public investor and slightly less of a pain in the ass than the public activist investor.” — Brian Halligan, Co-Founder & Executive Chairman, HubSpot

On CEO Compensation: “CEO comp is pretty broken at the moment. Everyone really relies heavily on RSUs. When I grew up in the industry, it was mostly ISOs, it was options until 2006 and regulations changed. It just creates sort of a risk averse behavior in the CEO.” — Brian Halligan

The Figma IPO might be the most misunderstood public offering in tech history. With a 250% first-day pop — pricing at $38 and opening at $85 and trading up to a day 1 high of $124— everyone from X pundits to Bill Gurley called it catastrophic mispricing. But the real story, told by those who’ve been in the room where it happens, reveals a much more nuanced truth about how IPOs actually work.

The Night Before: When Exhaustion Meets High-Stakes Poker

Brian Halligan, who took HubSpot public in 2014, pulls back the curtain on what actually happens in those final 24 hours. “You’ve never been as tired in your entire life as you are when you’re making this decision,” he explains. “You’ve been on the road for the last two weeks. You hit 12 countries. You had six pitches a day. Your battery is on red.”

Then comes the moment that defines everything: the pricing committee meeting the night before trading begins. “The investment bankers sit you down and say you got two big decisions to make. One is who are the investors going to be… and then what’s the price.”

The founders and bankers, perfectly aligned throughout the entire road show, suddenly find themselves “across the table from each other” in that final hour.

The Fidelity Gambit: A $1 Decision Worth Billions

The most revealing part of Halligan’s account centers on what he calls “the Fidelity discussion.” At HubSpot, Morgan Stanley pushed for a $24 pricing because “Fidelity has told us that they’re in at 24, they’re out at 25.”

“You really want Fidelity because Fidelity has trillions of dollars and they could own, you know, they could be a massive banker,” Halligan explains. But HubSpot pushed back, pricing at $25. The result? A 32% first-day pop to around $33.

Rory O’Driscoll, who’s been through this multiple times, confirms this is standard operating procedure: “They clearly give the same speech every time… It’s always Fidelity or one of the other two. It’s like ‘you want Fidelity, it’s only another dollar,’ right?”

The Real Story Behind Figma’s “Mispricing”

So what actually happened with Figma? O’Driscoll delivers the most important insight: the astronomical pricing wasn’t accessible at IPO time.

“The $98 price only happened because the IPO happened at $38,” he explains. “The discussion they were having on the day was… do I go two bucks more and exclude Fidelity or two bucks less and take Fidelity… Had someone walked in and said, I know this IPO is going to price at $100 bucks a share to open tomorrow morning. Let’s raise at 80. They wouldn’t have had a book cuz no one has bid at that thing.”

This is crucial for understanding IPO dynamics. The massive first-day pop wasn’t money “left on the table” — it was euphoria-driven demand that materialized only after the stock started trading.

The Unintended Consequences of Mega-Pops

But here’s where it gets problematic. O’Driscoll reveals a counterintuitive truth: “All those big buyers who came in at Figma $38 before they buy they have an internal process with a price target to exit. None of those price targets are going to be more than $110 bucks a share.”

The result? “Most of them are selling those shares right now… If you’re running money as one of these institutions and you bought at 35 and you built a business case that says, ‘We think this will be worth 50 bucks a share in two years and suddenly it’s worth $100 a share in two days.’ At least half those mutual funds have to say should we lose our position.”

This creates the exact opposite of what founders want — instead of long-term institutional holders, you get a trading-oriented shareholder base.

The CEO Compensation Crisis Nobody’s Talking About

Perhaps the most underreported aspect of the Figma story is Dylan Field’s $2 billion compensation package — and what it reveals about broken CEO incentive structures across the industry.

Halligan is blunt: “CEO comp is pretty broken at the moment.” The core problem? The shift from stock options to RSUs (Restricted Stock Units) after 2006 regulatory changes.

“When I grew up in the industry… it was mostly ISOs, it was options until 2006 and regulations changed and the expensing of that changed. So the world kind of moved to RSUs. It just creates sort of a risk averse behavior in the CEO.”

The Performance Stock Unit Solution (And Its Problems)

Figma tried to solve this with Performance Stock Units (PSUs) — essentially recreating options by tying stock grants to performance triggers. The problem? “They put this in place before the IPO. And if you read the triggers, they’ve already achieved the triggers.”

Because of the IPO pop, Field hit all his performance milestones immediately. “He’s made all the triggers already… the performance element vanished very quickly.”

O’Driscoll advocates for revenue and operational income targets instead of stock price triggers, but acknowledges the disclosure problem: “If we put that on the table and then circumstance change, we have to disclose it. So then all the analysts will start saying, ‘Oh my god, they think they can make 35%.'”

HubSpot solved this by tying CEO compensation to “net new ARR” and “earnings number” — much more controllable metrics than stock price.

The Case for Going Public: Why VCs Are Worse Than Public Markets

The most surprising revelation from Halligan: public markets are actually easier to deal with than VCs.

“When we were a private company, we had a bunch of quirky, slightly misaligned venture capital investors who were definitely in our shorts. Then we flipped to public and then we got rid of those VCs. Now we had a bunch of quirky slightly misaligned public investors who are less in our shorts.”

He continues: “I think VCs are a much bigger pain in the ass than the typical public investor and slightly less of a pain in the ass than the public activist investor.”

Why Timing Matters More Than You Think

Both Halligan and O’Driscoll emphasize the seasonal nature of IPO success. HubSpot went public three months after Zendesk, which “didn’t get the big long onlies and they never really got them. They always had a lot of hedge funds in there.”

The result years later? “When they hit an issue and they had activist pressure… you had a stronger investor base the whole way through.”

For companies considering going public now, O’Driscoll’s advice is clear: “Run forest run. The market’s wide open. The valuations are good. There’s a lot of demand. It’s very seasonal… If I were Canva, it’s amazing company. I’d be lining everything up to go public.”

The AI Revolution’s Impact on SMB Software

The conversation also touched on a critical question for B2B software: can AI work for SMBs? Lemkin and Halligan, who built a $30B company serving small businesses, is skeptical of many current approaches.

“80% of folks have an AI team to do this. Even with what they call SMB and VC, 80% have a team. SMBs don’t have a team, right? The owner of one restaurant… doesn’t have an AI team.”

The key insight: successful SMB AI products will need to be “pre-baked” and self-training, not requiring company-by-company customization that works for enterprise.

Meta’s AI Spending: Sustainable Strategy or Bubble?

The group also analyzed Meta’s monster Q3 results: 38% year-over-year EPS growth, 22% revenue growth, but a 22% drop in free cash flow due to AI infrastructure spending.

O’Driscoll’s take: “This isn’t an AI-enabled success. This is I have an awesome existing business, and it kicks off so much money that I’m allowed spend that money on building this great AI vision.”

The key insight: hyperscalers can sustain massive AI capex because their existing businesses generate enormous cash flows. “Real men with $70 billion in free cash flow get to spend $40 billion of that on servers.”

Key Takeaways for B2B Leaders

* IPO Timing Is Everything: Market conditions and investor appetite matter more than perfect fundamentals. The difference between getting marquee institutional investors versus hedge funds can define your public company experience for years.

* Rethink CEO Compensation: Move away from RSUs toward performance-based structures tied to operational metrics (revenue, profitability) rather than stock price. Stock price triggers often get hit accidentally or become impossible to achieve.

* Public Markets Aren’t the Enemy: Properly managed, public investors are often less intrusive than growth-stage VCs. The key is conservative guidance and clear long-term vision.

* SMB AI Requires Different Architecture: Traditional enterprise AI deployment models won’t work for small businesses. Success requires pre-trained, self-configuring solutions.

* The Best Time to Go Public Is When You Don’t Need To: Companies with strong cash generation and optionality should consider public markets during favorable windows, not when forced by cash needs.

* Long-term Shareholder Base Matters More Than Valuation: A 20% pop with the right institutional investors beats a 200% pop with day traders and hedge funds.

The Most Quotable Moments

On Founder Exhaustion: “You’ve never been as tired in your entire life as you are when you’re making this decision. You’ve been on the road for the last two weeks… your battery is on red tired and then the investment bankers sit you down.” — Brian Halligan

On Going Public Now: “Run forest run. The market’s wide open. The valuations are good. There’s a lot of demand. It’s very seasonal. Timing really matters.” — Rory O’Driscoll

On CEO Compensation: “If you paid Elon Musk like Mary [Barra] makes $29 million a year. Do you think Elon cares about $29 million? So you have to kind of comp it to the CEO’s net worth as opposed to just the peers.” — Brian Halligan

On AI and SMBs: “How do you solve the fact that it can take six months to roll out a Palantir grade deployment? How do you do that in 60 seconds? And any founders that crack that code, I want to invest this hour, this second.” — Jason Lemkin

*

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

Ross Biestman joined ServiceTitan in 2018 as employee #354 when the company was doing less than $30M ARR. As Head of Sales, he helped scale the vertical SaaS platform from Series B to IPO, growing revenue to over $860M ARR and achieving an $11B market cap. Before ServiceTitan, Ross spent nearly a decade in enterprise software sales, serving companies like Bloomberg, United Airlines, and Sprint. A Cal Bears rugby player turned sales leader, Ross has become one of the most respected voices in vertical SaaS go-to-market strategy.

Christina Shen is Managing Partner and Co-founder of Chemistry, a $350M early-stage fund investing in seed and Series A companies. Previously a Partner at Bessemer Venture Partners, Christina led ServiceTitan's Series A investment and has been tracking the company's remarkable journey for nearly a decade. She and Ross were undergraduate classmates at UC Berkeley - she in political science (destined for sales, as she jokes), he in business school (destined for venture capital).

This SaaStr Annual deep-dive represents a full-circle moment between two Cal Bears who took different paths but remained connected through one of the most successful vertical SaaS stories ever told.

Ross' Top 5 Learnings to Transform Your Go-to-Market Strategy

* ICP Discipline Creates Exponential Returns: ServiceTitan's laser focus on ideal customer profile (ICP) drove their ability to scale from $30M to $770M ARR in 7 years. Ross's key insight: "Everything we do is intentional" - they put blinders on and refused to chase dollars outside their ICP, even when it meant turning down revenue.

* Merit-Based Lead Distribution Beats Traditional Territory Models: ServiceTitan revolutionized sales by using AI to route leads to the AE with the highest propensity to close that specific deal type, not round-robin or named accounts. This merit-based system includes monthly scoring and relegation/promotion like English Premier League soccer.

* Industry Immersion Unlocks Multi-Vertical Expansion: To expand beyond plumbing into all trades, ServiceTitan hired industry professionals directly into R&D, spent countless hours in the field with customers, and methodically tested new verticals with their top performers before opening floodgates.

* Customer Site Visits Are Your Secret Weapon: In an age of Zoom and AI efficiency tools, Ross mandates that every new hire - from SDRs to C-suite - must learn the product, industry, and customer by spending time on-site. The founders still spend 95% of their time with customers.

* Vertical SaaS Creates Unprecedented Value When Done Right: By serving an ignored $1T+ industry (trades) with true dedication, ServiceTitan achieved what seemed impossible - 11x revenue growth in 7 years and an $11B market cap, proving vertical software's massive potential.

The Making of a Vertical SaaS Giant: Inside ServiceTitan's Journey

When Ross Beastman joined ServiceTitan in 2018 as employee #354, venture capitalists were telling him he was "crazy" to sell software to plumbers. Today, that crazy decision has paid off to the tune of an $11 billion market cap and over $770 million in ARR. But this wasn't luck - it was the result of methodical execution, unwavering discipline, and a deep understanding of what makes vertical SaaS work.

The Unlikely Beginning: Why Plumbers?

Ross's initial reaction to ServiceTitan was skepticism. Coming from enterprise software, serving companies like Bloomberg and United Airlines, the idea of "software for plumbers" in Los Angeles seemed like a step backward. But the founders, Ara and Vahe Kuroghlian, had a different vision.

"Their parents worked really hard as trades people, plumbers and HVAC contractors in Southern California," Ross explains. "Instead of going into their parents' garage to create Apple, they went into their dad's plumbing shop in Los Angeles and said this is an industry that has been ignored and trapped in the past."

The breakthrough came during customer due diligence. Walking into HVAC shops in places like Fresno, California - where it's 110 degrees outside - Ross witnessed something remarkable: "Every single person in this business was deeply embedded and committed to using this technology as their operating system, whether the person answering calls at the front desk, the CFO, the owner, the person in the warehouse checking inventory, or the technician in the field."

When he sized the industry at over a trillion dollars in domestic gross transaction volume, the opportunity became clear. The phone call to his wife was simple: "I'm going to Los Angeles to sell software to plumbers."

The $30M to $860M Playbook: Execution Excellence

Year One Reality Check

ServiceTitan in 2018 was "organized chaos in the best possible way." As a Series B company doing less than $30M ARR, they were laser-focused on new logo acquisition with a single product serving just plumbing, heating, and electrical businesses. The challenge was clear: identify the path to $50M, then $100M, then $500M.

The ICP Discipline That Changed Everything

Ross's biggest learning from enterprise software didn't translate - but his biggest insight did. "The focus on ideal customer profile pays real long-term dividends," he emphasizes. ServiceTitan's approach was ruthlessly disciplined:

* Marketing dollars, R&D investments, and go-to-market efforts were laser-focused on their defined ICP

* They refused to chase revenue outside their target, even when facing pressure to hit numbers

* This prevented the "thrash" that comes from serving customers who aren't ideal fits

"I find companies at earlier stages trying to chase dollars - maybe because they want to hit a specific target to be deemed more valuable in the venture community, or literally because they need dollars to make payroll. That approach allows for too much thrash."

The result? Customers born outside the ICP often become "born red" with heightened churn propensity, destroying unit economics.

The AI-Powered Sales Revolution

ServiceTitan's approach to AI isn't just product-focused - it's revolutionizing how they run their internal operations. Their sales organization uses AI for lead distribution in a way that's completely upended traditional models:

Merit-Based Lead Routing Instead of round-robin or named accounts, incoming leads are distributed to the account executive with the highest propensity to close that specific type of business. This system includes:

* Monthly scoring based on quota attainment

* Efficiency metrics defined by close rates

* Quality performance scoring through pitch and demo analysis

* A relegation/promotion system like English Premier League soccer

"If they perform really well, they get the best type of pipeline in that next month. If they perform poorly, they get relegated. If you get good, you got to stay good to be good."

Customer-Facing AI Solutions For customers, ServiceTitan built AI into core workflows like automated dispatching - sending the right technician to the right job to minimize drive time while maximizing skill set match and close propensity. This allows customers to scale technicians without proportionally scaling dispatchers, improving both top-line and EBITDA.

Multi-Vertical Expansion: The Methodical Approach

ServiceTitan's expansion from plumbing to all trades wasn't accidental - it followed a precise three-step methodology:

1. Hire Industry Professionals Into R&D. They recruited former technicians, call center agents, CFOs, and business owners directly into their R&D organization. "We want you to come work for us, report into our R&D function, and teach us how to build product for verticals we're not in today."

2. Intensive Field Research. The team spent countless hours on ride-alongs with technicians, visiting commercial properties and residences, and studying office operations to identify every value point for a completely tailored solution.

3. Methodical Testing with Top Performers. Rather than opening floodgates, they pulled their best salespeople from established markets to test new verticals. "That was scary for me as the person responsible for the P&L - I don't want to take my 10 best salespeople helping us keep the lights on and have them focus on something that could be a failure. But it was that discipline around establishing success criteria that earned us the right to go into multiple trades."

The Power of Customer Immersion

In an era of Zoom meetings and AI efficiency tools, ServiceTitan maintains an "old school mentality" that Ross believes creates massive competitive advantage:

Mandatory Customer Immersion. Every new hire, from entry-level SDRs to C-suite executives, must spend time with customers in the field before they can be effective in their role. This isn't optional - it's core to the ServiceTitan way.

"Good things happen when you go on site and spend time with your customers. Whether you're in marketing, sales, R&D, or HR - you need to spend time with customers way more than you spend time in the boardroom."

The founders exemplify this approach, spending the majority of their time with customers either on calls or in the field, despite running an $11B company.

Surviving the Curveballs: Lessons in Resilience

ServiceTitan's journey hasn't been smooth. The pandemic, economic uncertainty, and a pulled IPO in 2022 tested the organization's resilience. Ross shares a powerful story about overcoming the psychological impact of the failed IPO attempt:

His rugby coach's story about Army Rangers training resonated deeply. In Ranger school, candidates face a grueling physical test at 3 AM with sirens, fireworks, and loud music. The twist? When they think they've reached the finish line, they discover it's actually several more miles. "People didn't fail physically, they failed psychologically. They couldn't bring themselves to accept that this wasn't the destination."

This experience taught Ross that "the IPO is truly just a milestone, not a destination" - a reframe that helped ServiceTitan focus on long-term business durability rather than short-term milestones.

The Customer-First Philosophy That Drives Everything

ServiceTitan operates on a fundamental principle that Ross calls "order of operations": Every decision, dollar raised, product built, and person hired focuses solely on creating better outcomes for contractors. Only after delivering that value do they expect business returns.

"It's almost thinking about this from the value we're bringing to the customer and their ability to drive successful outcomes as the prerequisite to anything we become the beneficiary of."

This philosophy shows up in practical ways:

* Turning down revenue that doesn't serve the ICP

* Investing R&D dollars in customer-requested features vs. market expansion

* Hiring industry professionals who understand customer pain points

* Mandating field time for all employees

The $70,000 Credit Card Moment

Ross's most memorable customer win illustrates the power of providing disproportionate value. His first sales call required a cross-country journey to Rogers, Arkansas, including a missed flight, a middle-of-the-night drive through country roads, and a 7 AM arrival at a Hampton Inn.

The customer - a business owner in a small brick house with a tin roof - possessed incredible business acumen despite appearances. Forty minutes into the presentation, he stopped the demo, slid his American Express across the table, and said: "Stop talking. I know if you guys do anymore, the product's just going to get more expensive. Here's $70,000."

The full-circle moment came six years later when that same customer became a key reference for ServiceTitan's largest deal ever - proving that every customer interaction matters and you never know which will become most valuable.

The Vertical SaaS Opportunity

When Ross joined ServiceTitan in 2018, vertical software wasn't even a recognized category. Today, it's one of the hottest investment areas, but Ross's experience offers crucial insights for success:

Market Size Validation. The trades industry does over $1 trillion domestically in gross transaction volume, making it larger than many horizontal markets. But size alone isn't enough - the industry had to be underserved and ready for disruption.

Mission-Driven Approach. The best vertical SaaS companies don't just serve a market - they're passionate about transforming it. ServiceTitan's founders weren't just building software; they were honoring their parents' legacy and improving lives of hardworking people.

Customer as Secret Weapon. Every ServiceTitan customer call ended with the same request: "Don't tell my competition - this is my secret weapon." That level of customer advocacy is the gold standard for vertical SaaS product-market fit.

Building a Learning Organization

ServiceTitan's scale required constant learning and adaptation. Ross emphasizes challenging conventional wisdom: "Those preconceived notions and industry best practices aren't perfect translations. Those things need to be challenged every single day."

Key principles include:

* Testing everything rather than assuming best practices apply

* Building capacity models specific to your market and sales motion

* Iterating based on customer feedback rather than competitor benchmarks

* Maintaining beginner's mind even at scale

The Future of AI in Vertical SaaS

Ross predicts AI will become table stakes, just like mobile and cloud: "No one calls a company a mobile company anymore. Everyone will have to have an AI offering. In the end, we're just going to be talking about companies."

For ServiceTitan's customers, AI adoption has accelerated as private equity consolidation creates pressure for operational leverage. Business owners are "looking for opportunities to drive incremental leverage that is not dependent on human-based labor" across the entire customer funnel.

The key is focusing AI on real customer problems rather than technology for technology's sake - true to ServiceTitan's customer-first philosophy.

Ross's Top 4 Mistakes: Hard-Won Lessons from the Journey

Based on Ross's candid sharing throughout the interview, here are the critical mistakes he learned from building ServiceTitan:

1. Applying Enterprise Best Practices Without Question

"Those preconceived notions and industry best practices aren't perfect translations"

Ross initially tried to apply traditional enterprise software playbooks to vertical SaaS, including standard capacity models and go-to-market approaches. The mistake was assuming what worked at Bloomberg or United Airlines would work for plumbers in Fresno. The lesson: Challenge every assumption and build practices specific to your market.

2. Prioritizing Growth Over Profitability Too Long

"We were focused on hyper-growth... the idea of having to be profitable was something we can worry about later"

Like many B2B companies in 2021-2022, ServiceTitan optimized for growth over unit economics, which made the market correction and pulled IPO more painful. Ross learned that sustainable businesses need profitability discipline from earlier stages, not just growth-at-all-costs mentality.

3. Treating the IPO as a Destination Instead of a Milestone

"The IPO is truly just a milestone, not a destination"

The psychological impact of the pulled IPO in 2022 was devastating because the entire organization had oriented around it as the ultimate goal. Ross learned that focusing on long-term business durability and customer outcomes should always trump short-term milestones, even major ones like public offerings.

4. Not Investing Enough in Customer Immersion Early

"Good things happen when you go on site and spend time with your customers"

While ServiceTitan eventually built a culture of mandatory customer field time, Ross wishes he had institutionalized this practice even earlier. The biggest breakthroughs - from product insights to sales wins - came from face-to-face customer interactions, not boardroom strategy sessions. The lesson: make customer immersion non-negotiable from day one, for every role.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com

How three high-growth companies are actually implementing AI across their revenue operations — and what it means for your AI-informed GTM strategy today.

The AI transformation in go-to-market isn't coming — it's here. But unlike the broad promises we've been hearing, the real story is in the specifics. How are actual revenue leaders at scale implementing AI today? What's working, what isn't, and where should you focus your limited budget?

SaaStr Annual's AI Summit brought together three top GTM leaders who are pushing the envelope on AI implementation across their organizations, together with one a great VC to guide the convo:

* Sahana Sarma, Global VP GTM Strategy and Operations at Databricks, brings deep expertise in scaling revenue operations at one of the fastest-growing data and AI companies. She's been instrumental in building Databricks' "Ask Mo" internal agent and driving AI adoption across their field teams.

* Rob Schmeltzer, Head of Strategic Customer Success at Monday.com, oversees customer success operations for the work management platform serving over 180,000 customers. He's leading Monday.com's implementation of AI-powered deal desk assistance and automated engagement mapping.

* Uri Ouziel, Global Head of Deal Management at Benchling, manages complex deal processes for the life sciences R&D platform. His team has been pioneering automated CSM engagement tracking and customer intelligence aggregation at scale.

* Itamar Novick, Founder & General Partner at Recursive Ventures, provides the investor perspective, having backed multiple AI-enabled GTM tools and witnessed the evolution from early promise to practical implementation across his portfolio companies.

What emerged was a practical blueprint for how leaders are really using AI to drive efficiency gains.

The Foundation: Why Point Solutions Aren't Enough

The first major insight that emerged was the limitation of point AI solutions. As Databricks’ Global VP GTM Strategy noted: "I feel like a lot of the point solutions you see in AI are great, but that's where they are. There are point solutions."

The companies seeing the biggest wins are those building on platforms they already have — leveraging Gemini's integration with G Suite or ChatGPT's API capabilities to create customizable, wide-ranging tools rather than adding another vendor to their stack.

The Monday.com Approach: Building Your Deal Desk Co-Pilot

Monday.com's revenue operations team spent two quarters training Gemini to become what they call a "deal desk assistant." Here's how it works:

* Deal Intelligence: They feed Gemini their deals, playbooks, and connect it with their CPQ system

* Optimization Engine: The AI suggests optimal deal structures and pricing

* Efficiency Gains: Significant reduction in clicks and time-to-close for complex deals

The key insight? They didn't buy a new tool. They used an existing, already-adopted platform to solve a specific workflow problem. "When you can do it yourself using an existing tool that is already adopted and already integrated, it's a great win," the Monday.com leader explained.

Customer Intelligence: The New Competitive Advantage

All three companies have made customer research and intelligence a core AI use case, but their approaches vary significantly.

Databricks: The "Ask Mo" Internal Agent

Databricks built their own tool called "Ask Mo" that serves as an entry point into Salesforce, allowing their field teams to:

* Research customer organizational changes and triggers

* Identify next-best-product opportunities by analyzing similar customer patterns

* Generate executive briefing documents by pulling data across their entire stack

* Access consistent information quickly as they scale their team

The tool aggregates information across their entire tech stack and serves it to field teams in a conversational interface. This isn't just about efficiency — it's about consistency and quality of customer interactions at scale.

The Research Stack Evolution

Beyond internal tools, field teams are using a combination of:

* Gemini and ChatGPT for content creation and analysis

* Perplexity for deep customer research

* Gong for deal intelligence, pipeline forecasting, and coaching insights

But here's what's interesting: they're not just using these tools individually. The real value comes from synthesizing insights across multiple platforms.

Operations Efficiency: Where AI Shows Immediate ROI

The operations use cases are where AI is delivering immediate, measurable value.

Benchling's Automated Engagement Mapping

Benchling implemented Salesforce's Agent Force to automatically categorize and log CSM engagements. Previously, customer success managers had to manually log every interaction. Now:

* Automatic Detection: The system identifies calls, emails, and activities across all systems

* Smart Categorization: Engagements are automatically categorized by type

* Time Savings: CSMs focus on high-value activities instead of data entry

"Really just an example of trying to take manual repetitive things off of our team's plates," their CS leader explained.

Process Documentation with Scribe

Multiple panelists highlighted Scribe as a game-changer for operations teams. The ability to automatically document processes and turn them into sales enablement courses is "really impactful and saves a bunch of time."

Data Enrichment and Prospecting: The Clay Revolution

Clay emerged as a standout tool across multiple companies, but not for the reasons you might expect. It's not just about data enrichment — it's about building custom workflows.

Clay works because "it is not a point solution but integrates with a bunch of point solutions and enriches data, but you can build it into your own thing." This flexibility allows revenue teams to create custom prospecting and account research workflows without heavy engineering resources.

Touchless Prospecting: The Holy Grail

Multiple companies are running POCs on touchless prospecting — using AI to generate and send prospecting emails without human intervention. But there's a crucial caveat: "You still need to have some kind of human in the loop. We're not yet at the point where things can go out customer-facing without some level of risk and quality control."

Customer Intelligence at Scale: The Interpret Case Study

One of the most sophisticated implementations came from a company using Interpret to analyze customer communications at scale.

The Challenge: They were excellent at understanding individual customer needs but struggled to identify trends across their customer base.

The Solution: Interpret analyzes all customer communications to surface trends around product usage and feature requests, segmented by customer type.

The Impact: This aggregated intelligence has "really helped our conversation that I have and my team has with our product organization." Instead of anecdotal feedback, they now provide data-driven insights to product teams.

The Budget Reality: AI Isn't Free

Here's where the conversation got real. As one panelist put it: "We don't necessarily have more budget. We're already paying for all those tools, and the AI tools are incremental. They want more money, and I don't have a budget."

This budget constraint is forcing companies to make strategic choices:

Build vs. Buy Decision Framework

* Leverage Existing Platforms First: Can you solve this with ChatGPT/Gemini integration rather than a new vendor?

* Focus on Workflows, Not Features: What specific process are you trying to improve?

* Measure Integration Costs: How much effort will it take to get this working with your existing stack?

Enablement and Assessment: The Underrated AI Use Case

Companies are using AI-powered platforms like Udely and Uplimit for sales enablement, but the real value isn't just in content delivery — it's in assessment and impact measurement.

"We are using [these tools] to actually drive enablement and allow us to better assess... how does enablement really impact success?" This data-driven approach to enablement is helping companies optimize their onboarding and ongoing training programs.

The Glean Factor: AI-Powered Knowledge Management

Glean appeared in multiple companies' stacks as a solution for fast-scaling teams. As one leader explained: "We're growing so fast, bringing so many people in. Glean is a pretty effective way for us to get information out to our field that's consistent, especially from a content management perspective."

The key insight: AI-powered search and knowledge management becomes critical as companies scale beyond the point where tribal knowledge works.

What This Means for Your 2025 GTM Strategy

Based on these real implementations, here are the key takeaways for revenue leaders:

Start with Integration, Not Innovation The companies seeing the biggest wins are building on platforms they already have rather than adding new vendors. Before buying a new AI tool, ask: "Can I solve this with our existing ChatGPT or Gemini integration?"

Focus on Process Efficiency First The clearest ROI is coming from automating manual, repetitive tasks. Look for processes where your team is doing data entry, research, or documentation that could be automated.

Customer Intelligence is the New Moat Every company is implementing some form of AI-powered customer research and intelligence. This isn't optional — it's becoming table stakes for competitive sales teams.

Budget for Integration, Not Just Licenses The real cost isn't the AI tool — it's the integration work, training, and ongoing optimization. Budget accordingly.

Human-in-the-Loop is Still Required Despite the hype, no company is comfortable with fully automated customer-facing communications. Plan for human oversight and quality control.

The Road Ahead

The AI transformation in GTM is real, but it's more tactical than transformational right now. Leaders from Databricks to Monday are seeing significant efficiency gains, better customer intelligence, and improved process automation. But they're also being strategic about where they invest and how they implement.

The winners won't be the companies with the most AI tools — they'll be the ones who thoughtfully integrate AI into their existing workflows to solve specific problems. As these implementations mature over the next 12-18 months, the competitive advantages will become even more pronounced.

Thanks for reading The Secrets To Scaling in The Age of AI! Subscribe for free to receive new posts and support my work.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit cloud.substack.com