Notes — Jeanne DeWitt Grosser on GTM Engineering

Four questions [Adler frame]

Q1 — What is it? A practitioner’s guide to modern B2B go-to-market, anchored in the emerging Go-to-Market Engineer role, a philosophy of building the buying journey as a product experience, and frameworks for segmentation and sales team design. Drawn from nine years at Stripe and current work as COO of Vercel.

Q2 — How is it argued? Through first-person operational accounts: Project Rosland (Stripe’s 2017 pre-LLM attempt at automated outbound), Vercel’s lead agent deployment (10 SDRs → 1 in six weeks), the deal-bott (real-time Gong-based deal intelligence), the whiteboarding discovery session, and the segmentation XY matrix applied at both Stripe and Vercel. Claims are specific: lead-to-opportunity conversion maintained at human parity; $1,000/year agent cost vs. $1M+ salary line.

Q3 — Is it true? Highly coherent. The 80% pain-avoidance stat is flagged as round numbers [?source]. The ROI claim on the lead agent is striking but the methodology is credible (human in loop throughout, KPIs tracked before/after). The segmentation framework is orthodox and well-illustrated.

Q4 — What of it? GTM engineering is structurally different from past sales automation — it encodes human expert workflows into agents rather than just sending more emails. The insight that “go-to-market is a product” reframes sales as an experience-design problem, which generalises to any business where technical differentiation is narrowing.

Glossary

GTM Engineer: Hybrid role combining sales domain knowledge with engineering skills to build automated agents and workflows for go-to-market functions. The earliest version is typically a former sales engineer who writes code.

SDR (Sales Development Representative): Pipeline generation specialist. Inbound: qualifies website enquiries before passing to an AE. Outbound: proactively targets prospects in the ICP.

AE (Account Executive): Closer. Runs the full sales process from qualified opportunity to signed contract.

Company universe: A database where every row is a company and every column is an attribute predictive of buying behaviour. Used to personalise outbound at scale. Grosser calls this Project Rosland (Stripe, 2017).

Deal-bott / Lost-bott: Vercel’s internal agent (running against Gong transcripts) that flags deal health in real time and retrospectively diagnoses true loss reasons after close.

Segmentation: Dividing the total addressable market by attributes (size, growth potential, workload type, business model) to surface different buying processes that need different sales motions.

Forward-deployed engineering: Embedding engineers directly inside customer environments to help deploy the product — an AI-era evolution of professional services.

AEO (Answer Engine Optimisation): Optimising content for appearance in LLM-based answer engines, analogous to SEO but for AI search. Grosser identifies this as an emerging driver of Vercel demand.

§The GTM engineer role

The role was conceptually prefigured by Project Rosland at Stripe in 2017: a “mad-libs” outbound email system built on a company-universe database, where 80% of the email was pre-filled by attributes (industry, business model, persona) and 20% was static template. It failed because LLMs didn’t exist yet and false-positive rates were too high.

At Vercel (2025), Grosser rebuilt the concept with a GTM engineer (former sales engineer, CS background) who spent 25–30% of his time over six weeks building a lead agent. Workflow: (1) shadow the highest-performing SDR; (2) enumerate every tab they open, every data source they hit; (3) encode that into a deterministic + agent-assisted workflow; (4) human review every output before send; (5) track lead-to-opportunity conversion and touches-to-convert as the benchmark; (6) when the agent matches human performance, redeploy the humans.

Outcome: 10 SDRs doing inbound → 1 SDR doing QA, with the other nine moved to outbound. Agent cost: ~$1,000/year.

Key insight: agents don’t replace sales reps when the task requires sophisticated multi-stakeholder enterprise navigation. They replace the repetitive top-of-funnel tasks that most salespeople found least interesting anyway, freeing them for the 70% customer-facing time that was previously impossible. [§Deal-bott below for post-qualification use.]

Ideal GTM engineer profile: prior sales engineering background (understands what good sales looks like) + ability to write code. Not a pure engineer dropped into GTM. Reverse also fails: a pure AE who can’t code can describe the workflow but can’t build the agent.

§Deal-bott and lost-bott

Loss analysis before the deal-bott: account executives self-report loss reasons in the CRM. These are systemically biased (loss reported as “price” when the actual diagnosis is “never engaged economic buyer” [§Segmentation]).

Lost-bott: feeds Gong transcripts, Slack interactions, and email threads for every closed-lost opportunity into an agent. The agent produces a root-cause analysis that frequently contradicts the AE’s self-reported reason. In the example given, the AE said “lost on price”; the agent identified “failure to demonstrate ROI to economic buyer” as the true cause.

Deal-bott (real-time): runs in production. Pushes insights into per-customer Slack channels: “You are mid-process and have not spoken with an economic buyer,” “That call with the economic buyer went poorly — here’s a suggested follow-up.”

Additional use: rapid enablement. With a Vercel shipping velocity described as “something every other day,” the agent runs against calls post-launch, identifies objection-handling gaps, and surfaces them weekly in a sprint-style review. Bugs in the go-to-market process are triaged and fixed with the same cadence as product bugs. [?] Grosser calls these “bugs in your go-to-market process.”

§GTM as product — the buying journey

The thesis emerged from Grosser’s time on Gmail (2004–): when Gmail launched, it had a full year’s lead over Yahoo Mail on storage and performance. That level of technical differentiation no longer exists in most SaaS categories. As software commoditises, the experience of being sold to becomes a differentiator.

Practical form at Stripe: the first post-qualification call was traditionally a discovery interrogation. Stripe replaced it with a whiteboarding session — the customer draws their payments architecture. Grosser’s observation: most customers had never mapped their own payments architecture. They left with a new asset and a sense of a collaborative partner rather than a vendor running a script. The session was also more informative for Stripe than structured discovery, because it surfaced the full competitive landscape and buying complexity organically.

Principle: add value at every touchpoint regardless of whether the customer buys. Grosser’s nine-year perspective at Stripe: “Half a decade later, here they are and they bought.” Investing in the relationship even through a lost deal produces win-back rates that compound over time.

At Vercel: outreach includes core-web-vitals performance benchmarking (with peer comparison) and AEO content even when there is no immediate conversion path. Builds authority that converts later.

The product parallel: map the buying journey the way a PM maps a user flow. Step 1 through Step N. At each step: is this an experience or a transaction? Transactions are forgettable; experiences create customers for life.

§Segmentation — the XY matrix

Standard segmentation (size: SMB / mid-market / enterprise) is necessary but insufficient. Grosser’s framework: add at least two more axes that reflect how your specific product is bought.

At Stripe:

• X-axis: size (SMB, mid-market, enterprise)
• Y-axis: growth potential (consumption-based model = high-growth customers are disproportionately more valuable)
• Third cut: business model (B2B → billing + ACH; B2C → consumer payments; platform / marketplace → Stripe Connect)

At Vercel:

• X-axis: size + promote (Crux rank from Chrome data = traffic volume; a 3,000-person company with top-25 internet traffic is treated as enterprise)
• Y-axis: growth potential (same as Stripe)
• Third cut: workload type (e-commerce → different language, different tooling; SaaS → migration barrier; crypto → cloud-native stack)

Key insight: segmentation is a company-wide concept, not just a GTM concept. Every PM building a feature should know which segment they are targeting and why. Grosser delivers the segmentation framework to every new hire in their first week at Vercel. The segmentation question is: “Do I have a point of view on where I’m trying to win and why?”

Practical approach on joining a new company: in the first 30 days, partner with data science to run a regression on what attributes of a customer predict high revenue. That is the empirical foundation of the segmentation framework.

§Building sales teams engineers trust

Grosser’s litmus test: if you put an account executive in front of ten engineers from the company, it should take them ten minutes to realise the AE is not a product manager.

What this requires:

1. Deep product knowledge — credibility with the engineering org
2. Discipline about signal vs. noise — not every customer request is a product gap; the best AEs can distinguish objections to overcome from genuine market opportunities
3. GM mindset — thinking about building a business, not closing deals; able to say no

Implication: the go-to-market org is an R&D extension. A 20-person sales team talks to more customers per week than most UXR programmes. If that insight is translated well, it becomes a product intelligence function, not just a revenue function.

Sales hiring mix: classic AE profiles (sales skills, pipeline discipline) paired with consultant or banking profiles (quantitative, P&L-oriented, able to run a TCO analysis). The two learn from each other: the consultant learns sales process; the AE learns how to speak to a CFO about total cost of ownership.

Sales comp hot take: annual comp plans lock in incentives 12 months in advance. In a fast-moving product company (Vercel shipped an entire AI cloud mid-year after plans were written), this rigidity creates misalignment. Grosser advocates for more flexibility in comp design, though acknowledges she doesn’t yet have a clean solution.

PLG note: PLG is not a permanent state. Every company eventually needs a sales motion for deal sizes that can’t close through self-serve. The mistake is waiting too long before building the outbound muscle, which takes time to become predictable.