Stacked GTM

Cliff Simon, CEO & Founder of Polaris Ops, walks Andy and Noah through structuring GTM engineering across the full lead-to-cash cycle. From uncovering six months of AI Apex slop in a client's Salesforce org to crafting custom signal plays that outperform off-the-shelf data, Cliff shares what he's learned across 20+ engagements in 14 months.

Topics discussed:

• Build versus buy decisions for GTM engineering projects

• Total cost of ownership for AI-built GTM systems

• Structuring context layers with knowledge graphs and memory

• Positioning GTM engineers inside centralized RevOps teams

• Cleaning up AI-generated Apex slop in Salesforce instances

• Crafting custom signals that outperform commodity data vendors

• Running five go-to-market plays that reinforce each other

• Micro SaaS as a future path for GTM engineers

Amanda Kahlow founded 6sense. Her new company, 1mind, deploys AI "Superhumans" that orchestrate roughly 15 agents to solution, demo, and close across the full buyer lifecycle, replacing the traditional SDR-to-AE-to-SE handoff chain.

Amanda shares how her Superhuman "Nigel" joined a live meeting, held 31% of the talk time, and helped close a $262,000 deal in 11 days. She breaks down why solving for growth over efficiency changes how you build and staff your entire GTM org.

Topics discussed:

• Orchestrating 15 agents per Superhuman across buyer lifecycle
• Closing a $262,000 deal in 11 days with AI
• Shifting GTM from seller control to buyer control
• Using Superhumans for post-call sales coaching
• Building outbound without creating a bad buying experience
• Pricing Superhumans at the cost of a human
• Measuring deal cycle compression and ACV growth
• Deploying personalized Superhumans across ABM campaigns

Philip Cooper, Chief Customer Officer, Agentforce Sales at Salesforce, explains how the CRM is being rebuilt around agents, ambient data capture, and headless architecture. With the Momentum acquisition powering a self-driving CRM and MCP exposure decoupling sellers from the UI, Philip lays out what happens when the system of record starts updating itself.

Philip walks Noah and Andy through why the semantic layer still matters for consistent definitions, how Slack is becoming the bidirectional system of work where sellers forecast and invoke agents, and how revenue leaders can delegate business outcomes to orchestrated agent workflows.

Topics discussed:

• Positioning Agentforce as a horizontal capability layer across clouds

• Auto-updating CRM through Momentum's conversation capture

• Exposing Salesforce headlessly through MCP for agent integration

• Using memory fragments to surface undiscovered deal insights

• Layering a semantic layer on unstructured data for consistent KPIs

• Turning Slack into a bidirectional system of work with CRM

• Orchestrating revenue outcomes by delegating to multi-agent workflows

• Shifting the build vs. buy calculus with opinionated agent deployment

A mining safety company needed to find prospects, and no pre-built tool existed for that niche. Jai Toor at Deepline pointed a coding agent at the problem. It found a Reddit ML model, combined it with hiring signals, and delivered qualified contacts in under 30 minutes.

Jai walks through Deepline's developer-first approach to GTM data infrastructure, from waterfall enrichment to deploying deterministic code-based workflows. He shares how close-lost analysis surfaces niche buying signals specific to each company.

Topics discussed:

• Building code-native GTM infrastructure designed for coding agents
• Using waterfall enrichment across multiple data providers programmatically
• Discovering niche buying signals through close-lost regression analysis
• Deploying deterministic plays from iterative agent testing
• Adopting pay-as-you-go pricing inspired by OpenRouter's model
• Centralizing data while decentralizing agent-based rep interfaces
• Generating org charts from multiple data sources at $7 each
• Predicting cold call answer rates with ML-driven scoring models

Austin Hughes, Co-Founder and CEO of Unify, joined Noah and Andy to make the case that offer based outbound is quietly beating the classic ask for a meeting.

Austin walks through why Unify skipped Inbound entirely, how offer based outbound is beating classic meeting asks, and why he still believes reps need to own their own prospecting decisions even as agents take over the grunt work.

Topics discussed:

- Building offer based outbound instead of meeting asks

- Why Unify skipped Inbound as a use case entirely

- Letting reps buy in individually before Rev Ops gets involved

- Owning a company wide context graph in the data warehouse

- Why AE demand surprised Unify more than SDR demand

- Why teams now want predictability and cost over frontier models

- Advice for hiring the first two outbound reps

- Why rep agency still matters as automation increases

Dust keeps coming up across the GTM engineering community, and this episode explains why. Gina Kabasakalis, Founding US GTM at Dust, has a clear-eyed view of where AI adoption actually breaks down: not in the build, but in the handoff from one technical person to an entire team.

She gets into the governance and permission infrastructure that makes multiplayer AI viable, with a concrete example of how RevOps encoding the right Salesforce schema field into a shared agent saves every sales manager from carrying that knowledge individually. She also covers what companies consistently get wrong about ROI measurement, and a new Dust feature called Pods that puts humans and agents in a shared workspace. If you're past the single-agent stage and trying to figure out how to scale this across an org, this one is worth your time.

Topics discussed:

• Why RevOps and IT/CISOs are the two buyer profiles gravitating toward Dust

• Multi-model routing: matching model complexity and cost to the task rather than defaulting to the highest-reasoning model for everything

• Encoding institutional knowledge at the system level so ICs stop carrying it individually

• Piloting vs. rolling out: validating long-term fit vs. the executive air cover and embedded AI ops function a real rollout requires

• Why time savings is the wrong AI ROI frame and what the CRM hygiene argument actually proves

• Pods: a shared workspace for humans and agents, with use cases from new hire onboarding to funding round comms

• Four-component agent framework: trigger event, data sources, recipient and destination, output transformation

• Agent sprawl: when to consolidate duplicate skills and when distributed creativity is the point

Everyone wants to run GenAI workflows. The problem is they're running them on commoditized data, team sizes, job titles, basic firmographics  and wondering why nothing converts. Joe Lehr, Director of GTM Engineering & Innovation at Primary Ventures, works across a portfolio of companies from pre-seed through Series D, backed by a $625M fund five, and he sees this mistake at nearly every stage.

Joe gets specific about how he builds out GTM infrastructure from scratch: which signals actually move deals, how Parallel.ai powers Clay's CLAGIN for enrichment at scale, and what it actually takes to move a sales team from single-player AI experiments to a shared multiplayer system built on Supabase and MCP. He also pushes back on a few widely held beliefs, including whether deep hyper-personalization still works, why your first GTM hire should not be a 23-year-old, and why CS may be the most under-automated function in the stack.

Topics Discussed:

• Non-commoditized signal mapping before any AI motion is built

• Why creativity, not tooling, is the actual constraint at early-stage companies

• The right first GTM hire: what to look for and what to avoid

• How Parallel.ai powers Clay's CLAGIN for signal enrichment at scale

• Mid-funnel data capture and why multi-opportunity edge cases break most tools

• Build vs. buy: taking vendor sales calls before deciding what to build

• Going multiplayer with Supabase, MCP, and structured account objects

• Why deep hyper-personalization is losing ground to smarter segmentation

• CS as the next frontier for agentic automation, expansion, and churn scoring

• Observability gaps when vibe-coded internal tools skip engineering involvement

James Underhill, Head of GTM Ops at Profound, runs the ops and systems function through one of the more extreme growth curves you'll hear about: five times ARR in nine months, headcount from sixty to two hundred, and an AE team scaling from thirty toward a hundred by year end. His operating principle is simple: buy infrastructure, build applications. That single bias shapes how his team is staffed, what they buy, what they build, and where they draw the line on both.

He built a deal desk bot in under twenty minutes on Dust without writing a line of code.  His team deflected 70% of support inquiries with an internal triage agent. And he replaced the traditional business partner role entirely by giving the field direct, semantic access to their own data.

Topics Discussed

• Why calibrating to the first derivative of growth matters more than current headcount

• Using Dust as a low-code agent platform across GTM, CS, and internal ops

• Deflecting 70% of support inquiries with a confidence-gated triage agent

• Buy infrastructure, build application as a disciplined team operating principle

• Why GTM engineering at this level requires actual software engineers, not just technical curiosity

• GitHub fluency as the new hiring proxy for this function

• Snowflake plus a semantic layer as the foundation for real-time, conversational data querying

• Centralizing Claude Code skills in Notion for rep-accessible, field-ready workflows

• The hidden maintenance cost of vibe-coded tools and why it compounds fast

• Replacing business partner roles with self-serve data access

• Where Gong's call intelligence falls short and why that becomes a build opportunity

• How to decide what to build versus buy when your team could technically do both

Seth Marrs spent six years at Forrester as their lead sales AI and technology analyst, which means he evaluated every major wave of GTM tech and has the receipts to back his calls. Now at Sandler as Chief Strategy Officer, he's seeing the practitioner side up close. His read on the agentic SDR market is blunt and sits well outside the vendor narrative, and he supports it with something most guests can't: actual data on what happens to sellers when you train them, test them, then watch them on live calls the next day.

He also gets specific on the parts most people gloss over. Why removing the seller from data capture is the first move that makes everything downstream possible. Why "next best action" is a flawed idea that quietly argues against needing sellers at all. And why the agentic opportunity in sales may sit with the SE and the human selling skill, not the SDR everyone is trying to automate away.

Topics discussed:

• Why "fire your BDRs and automate with AI" is the vendor red flag to watch for

• The recording-permission gap that breaks the data flywheel on outbound

• Keeping BDR research depth while moving from 10 calls to 50

• Sales motion as the deciding factor: high-velocity vs. long-cycle BDR economics

• Removing the seller from data capture entirely to build a reliable dataset

• Compete score: revenue per rep normalized across net-new, farmer, and CSM roles

• The 40 to 80 to 40 adoption collapse measured on live calls, not surveys

• Certifying sellers on what they actually say to customers, not test completion

• Why top-three suggested actions beat a single prescribed next best action

• The agentic SE: training a bot on the technical layer while humans own the conversation

• Extracting intelligence from the bottom of the funnel where capture doesn't exist

• Buy the infrastructure, build the AI layer on top: the maintenance trap with vibe-coded tools

Ryan Milligan started at Quotapath as Director of RevOps, spent four and a half years building the data and systems foundation, and is now CRO. His team has run at 100% blended quota attainment in 8 of the past 10 quarters, never below 90%, and has grown closed ARR per rep 1.7x in 18 months. The GTM engineering team doing most of the building is two people.

In this episode, Ryan gets specific about how the whole system works: the data architecture decision he made on day one that still underpins everything, how he splits Dust and Claude into distinct roles across the sales cycle, and why he thinks the current wave of everyone building their own tools is a bubble with a painful correction coming. He also makes a sharp case that comp plan design is one of the highest-leverage tools a CRO has for changing the mix of revenue being closed, not just paying people.

Topics discussed:

• Processing data in the warehouse nightly and reverse ETL-ing into CRM so both always speak the same language

• The build vs. buy litmus test: uniquely bespoke and relatively fixed vs. everything else

• Rep-built V1 prototypes handed to RevOps for productionizing and org-wide rollout

• Dust as system of record, Claude as system of action, and how they split across the sales cycle

• Thursday multi-thread standup: every rep required to arrive with all multi-threads queued for active opps

• The "what would it take to close twice as many deals" framework for identifying which agents to build next

• Warm outbound architecture using Clay, Unify, and product interaction data as intent signals

• Comp plan design as a lever for changing the shape of revenue reps close, not just incentivizing volume

• Why data architecture is the only real defense against confident AI hallucination

• GTM engineer defined as the owner of the full prospect-to-renewal lifecycle

Listen to more episodes: 

Apple 

Spotify 

YouTube