All podcasts / Lenny's Podcast / Summary

What world-class GTM looks like in 2026 | Jeanne DeWitt Grosser (Vercel, Stripe, Google)

2025-11-30 - 86 min - source - Read full transcript
Lenny Rachitsky (host)Jeanne DeWitt Grosser

Key insights

A single GTM engineer, working part-time, replaced a 10-person outbound SDR team's inbound function with one agent in six weeks.
Vercel had 10 SDRs handling inbound leads. A go-to-market engineer spent roughly 25-30% of his time over six weeks building a lead-qualification-and-response agent, modeled on the team's best performer via shadowing. Lead-to-opportunity conversion rate held flat versus human performance while conversion cycle time shrank, since the agent responds instantly instead of leads sitting in a queue overnight. Nine of the ten SDRs were redeployed to outbound.
ai-go-to-market-engineering
The same agent that replaced 10 salaries costs about $1,000 a year to run.
The 10-person SDR function represented well over a million dollars in salary. The lead agent that replaced nine of those roles runs on Vercel's own infrastructure for about $1,000/year - over a 90% cost reduction relative to the labor it replaced, illustrating how cheap internally-built GTM agents can be once the workflow is well understood.
ai-go-to-market-engineering
Effective GTM engineers come from sales, not engineering, because encoding best practice requires actually knowing what good sales looks like.
Vercel's first three go-to-market engineers were technical sales engineers (former front-end developers) rather than pure software engineers. Jeanne found that agents modeled off a rep with only two years of experience made mistakes an experienced seller would catch; without 20 years of sales pattern-matching to draw on, you either need direct sales experience or have to deliberately study sales best practice to build a good agent.
ai-go-to-market-engineering
AI-run analysis of sales calls surfaced a completely different loss reason than the human account executive reported.
An AE attributed Vercel's biggest lost deal of the quarter to price. Running an agent over every Gong call, Slack message, and email showed the real cause was never reaching the economic buyer and failing to make the ROI/TCO case land - an inability to demonstrate value, not a pricing problem. This reframed how the team categorizes and acts on lost deals going forward.
ai-go-to-market-engineering
Sellers have historically spent only 30-40% of their time actually talking to customers; AI-driven automation of research and follow-up can push that toward 70%.
Annual industry benchmark reports have shown reps spending the minority of their time with humans for the 20 years Jeanne has been in sales. By offloading research, prospecting, and rote follow-up to agents, she expects sellers can spend the majority of their time on actual human interaction, deepening relationships rather than doing administrative work.
ai-go-to-market-engineering
About 80% of B2B buying decisions are driven by avoiding pain or reducing risk, not by chasing upside.
Jeanne cites this rough industry stat (heard secondhand, not independently verified) to argue that founders naturally pitch 'the art of the possible,' which resonates with other founders but not with most enterprise buyers, who are protecting their revenue targets, their career, and their team from the risk of a bad purchase decision. Effective enterprise GTM should foreground de-risking language over future-upside language.
sales-as-a-craft
Go-to-market should be designed and iterated on like a product, mapping the full customer journey and asking how to make every step feel human rather than transactional.
Jeanne's thesis, formed after seeing Gmail's technical edge over Yahoo/Hotmail erode as software commoditized: as products differentiate less on features, the experience of being sold to becomes a bigger driver of buying decisions. At Stripe, replacing the standard discovery-questionnaire call with a collaborative architecture whiteboarding session gave prospects a tangible artifact and built emotional buy-in, rather than feeling interrogated.
gtm-as-a-product
Segmentation should combine company size with a second axis specific to what actually predicts your buying behavior.
Beyond small/medium/large, Stripe layered in growth potential (a 200%-YoY grower was worth more attention than an 8%-YoY grower, given its consumption model) and business model (B2B vs. B2C vs. platform/marketplace, which determined whether a customer needed ACH/wires, consumer payments, or Stripe Connect). Vercel layers in site traffic (Google's CRUX score) and workload type (e-commerce vs. SaaS vs. crypto), since a small company with huge traffic can be more valuable than a larger low-traffic one. Jeanne's rule of thumb for founders: narrow segmentation to about three attributes, and treat it as a whole-company framework, not a sales-only exercise - Vercel teaches its 'Know Your Customer' segmentation to every new hire in their first week so product decisions account for it too.
segmentation-and-pricing
Founders should typically wait until roughly $1M ARR to hire their first dedicated salesperson, and must actually hand over the sales process, not just the title.
Before that point, customers usually look too dissimilar from one another for a repeatable process to exist yet - it's still an evangelist, founder-led sale. Once an ICP can be written down, founders should document their discovery questions, objection handling, and content so a hire can actually take over, while staying connected to customers to keep informing the roadmap.
founder-led-sales-transition
PLG has a ceiling and companies that delay building a sales motion too long hit a growth wall.
PLG works well at the outset for most non-enterprise-first products and remains a major growth driver at both Stripe and Vercel, but self-serve buyers generally won't spend seven figures through a self-serve flow. Companies that wait too long to build a predictable outbound sales engine (which itself takes time to mature) stall out; Jeanne can't name a $100B PLG-only company.
founder-led-sales-transition
Unbundling enterprise-only features into a self-serve tier can unlock significant PLG growth.
Vercel found that roughly half of customers buying its 'enterprise' SKU were actually startups, meaning enterprise-only packaging was blocking demand that wanted self-serve access. After a pricing restructure in August that split an enterprise SKU from a pro SKU and moved wanted features to self-serve, PLG funnel growth accelerated and reduced the need for human-intermediated sales.
segmentation-and-pricing
A go-to-market org earns credibility with product and engineering by having product depth deep enough to be mistaken for product managers.
Jeanne's litmus test: an account executive on her team should be indistinguishable from a PM within 10 minutes of talking to engineers. She believes the best GTM orgs are equal parts revenue-driving and R&D, translating the volume of customer conversations a large sales team has into signal that feeds the product roadmap, which requires skill in distinguishing a genuine market opportunity from a one-off objection.
gtm-as-a-product

Companies

Techniques and frameworks

Summary

Jeanne DeWitt Grosser - COO of Vercel, former chief business officer at Stripe where she built its first sales team, and a Google Gmail alum - joins Lenny to lay out what a world-class go-to-market org looks like now that AI has entered the sales stack. Her central argument is that go-to-market should be treated as broadly as "any function that touches a customer or makes a dollar" (marketing, sales, customer success, support, partnerships), and that AI is collapsing what used to be roughly 17 specialized GTM roles back toward an integrated, product-like lifecycle.

The most concrete thread is the rise of the "go-to-market engineer," a technical hybrid role - Vercel's first three came from its sales-engineer ranks - whose job is to turn human sales workflows into AI agents. She walks through Vercel's own build: a single GTM engineer, working part-time over six weeks, replaced most of a 10-person inbound SDR function with an agent that held conversion rates flat while cutting response time, at a running cost of about $1,000/year against well over $1M in prior salary. The build process kept a human in the loop throughout - shadowing top performers to encode workflow, letting the agent draft and a human approve, then gradually removing the human as trust built. She's candid that this only works if the agent is built by someone who actually understands what good sales looks like, since an agent modeled on a two-year rep still makes mistakes a 20-year veteran would catch.

Beyond automation, Jeanne treats go-to-market itself as a product discipline. She traces her thesis back to working on Gmail at Google in 2004, watching technical differentiation fade as cloud computing commoditized software, which led her to bet that the experience of being sold to becomes the differentiator when products converge on features. At Stripe this showed up as replacing standard discovery calls with collaborative architecture-whiteboarding sessions; at Vercel it shows up as agents (Deal Bot, Lost Bot) that mine Gong calls, Slack, and email to surface real reasons deals are won or lost - in one case revealing that a deal an AE attributed to "lost on price" was actually lost because the team never reached the economic buyer.

On strategy and structure, she offers a repeatable segmentation framework (size plus one or two business-specific axes - growth rate, business model, site traffic, or workload type - narrowed to about three attributes), reinforces the well-known but underused idea that roughly 80% of B2B buying is driven by risk-avoidance rather than upside-chasing, and gives founders concrete guardrails: wait for roughly $1M ARR and a real ICP before hiring your first salesperson, and don't let PLG run past its natural ceiling without building a real outbound motion. She closes with pricing lessons from Stripe (killing an unnecessary freemium trial, unbundling an overloaded enterprise SKU at Vercel to unlock self-serve demand) and her approach to hiring a "diversified portfolio" of sales profiles - pairing sales-native reps with consulting/banking backgrounds so each side teaches the other.

Notable Quotes

"It should take them 10 minutes to figure out you aren't a product manager." - Jeanne DeWitt Grosser, on the litmus test for a sales org that has real product depth

"That's like a, you know, 90 plus percent reduction in total cost there." - Jeanne DeWitt Grosser, comparing the ~$1,000/year cost of Vercel's lead agent to the salary of the 10-person SDR team it replaced

"Those are just bugs. They're bugs in your go-to-market process. So you should not have them." - Jeanne DeWitt Grosser, on treating objection-handling gaps surfaced by AI call analysis the way an engineering team treats sprint bugs

"80% of customers buy to avoid pain or reduce risk as opposed to increase upside." - Jeanne DeWitt Grosser, on why enterprise pitches should lean on de-risking rather than "the art of the possible"

"Yeses are great, nos are great, maybe it's will kill you." - Jeanne DeWitt Grosser, quoting a sales trainer, on treating a clear "no" as useful data rather than a failure