In this episode, Paid’s Manny Medina sits down with Mark Roberge, Co-founder of Stage 2 Capital, former CRO of HubSpot, Harvard Business School professor, investor, and author of The Science of Scaling, to break down what actually changes in go-to-market during an AI boom, and what absolutely does not.

Mark argues that while AI is accelerating workflows, it has not rewritten the laws of scaling. Human decision-making hasn’t changed. Retention still matters. Unit economics still matter. Hiring 17 reps overnight is still dangerous.

They go deep on product-market fit and why it’s not a feeling, the Stay / Go / Slow scaling model, founder vs. VC misalignment, systems of action vs. systems of record, and why the current AI cohort may see the highest failure rate in startup history.

“If you burned a billion dollars this year and you’re not OpenAI, that’s probably too much.”

First Principles Still Win

One of Mark’s central points is simple: AI does not change behavioral science.

“AI is not going to change the behavioral science of how humans make decisions.”

Much of AI 1.0 has been workflow streamlining, not workflow reinvention. The fundamentals of buyer psychology, sales process design, and value creation still apply.

What does change is leverage. Mark believes the first major unlock in go-to-market AI is increasing selling time.

“If you increase selling time from 25% to 75%, you triple X productivity right there.”

Product-Market Fit Is Not a Feeling

Most founders say they’re ready to scale when they ‘feel’ product-market fit. Mark rejects that entirely.

“Product-market fit is when you create customer value consistently.”

The metric? Retention. Specifically, net dollar retention is north of 100%. In early stages, you can’t wait a year to measure retention, so Mark pushes founders to define a leading indicator:

What usage behavior in month one predicts long-term retention?

• Slack used 2,000 team messages.
• HubSpot used three features adopted.
• Notion used weekly engagement.

If 80% of new customers hit that indicator, you have product market fit. If not, scaling is premature.

The Stay / Go / Slow Model

Instead of locking into rigid annual plans, Mark proposes a quarterly decision framework. After each quarter, you evaluate:

1. Demand generation for health
2. Conversion and quota attainment
3. Leading indicator of retention

If all are green, go faster.

If some are yellow, stay the course.

If any are red, slow down and fit it.

“It's so stupid that we build these annual plans when you're a $2,000,000 business and we abide by them like they're scripture.”

New Logos Should Not Be Slide One

In today’s AI cohort, Mark sees a dangerous pattern.

Boards ask for new logos, founders report ARR growth, then pilot revenue gets labeled as ARR.

But value creation lags.

“The first slide in your board deck should be your leading indicator of retention.”

Customer success should be a first-class citizen metric.

Founder vs. VC Incentives

VCs have 20 bets. Founders have one.

Some investors would rather see a company fail fast than ‘skimp along’ at 6-% growth.

But a founder who builds a durable $700M exit instead of chasing a trillion-dollar dream may protect life-changing outcomes.

This tension fuels overscaling.

Are We in an AI Bubble?

Mark’s answer: yes.

Signs of a classic bubble:

• Extreme valuation multiples
• Extraordinary burn ratios
• Experimental deployments counted as durable revenue
• First movers overcapitalized

“I think the last two-year cohort will see the highest failure rate in startup history.”

At the same time, the winners may be generational.

Today’s Value Prop Won’t Win Tomorrow

The most strategic insight of the episode: The product printing money today is unlikely to be the long-term moat.

Mark references Amazon’s early book focus as a wedge.

You design big, start small, and build the infrastructure for what the market will want in five years.

If you build the future too early, the market isn’t ready. If you only optimize for what sells today, you lose the long game.

Companies Mentioned

• HubSpot
• Salesforce
• Workday
• ZoomInfo
• OpenAI
• Microsoft
• Amazon
• Slack
• Notion
• Cursor
• Day AI
• Harvard Business School
• Boston Consulting Group

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.