Go-to-market as a product: lessons from Stripe and Vercel

Executive overview

Most go-to-market orgs operate as disconnected silos — marketing, sales, and success each using slightly different segmentation, targeting, and messaging. The fix is to treat the entire customer lifecycle as a single integrated product, designed with the same intentionality as a software system.

Jeanne DeWitt Grosser (COO at Vercel, former Chief Business Officer at Stripe) lays out a framework covering segmentation, the rise of the GTM engineer, sales as R&D, and how to build a sales org that earns the trust of engineers.

Selling to reduce risk almost always outperforms selling to increase upside — 80% of buyers are motivated by avoiding pain, not chasing gain.

What go-to-market actually means

• GTM is every function that touches a customer or makes a dollar: marketing, sales, technical sales, customer success, support, partnerships.
• The industry has over-specialised into 17+ roles; many will collapse as AI re-architects workflows.
• The goal is to map the full lifecycle — from awareness to high-LTV retention — and orchestrate it like a product.
• Most companies start with sales (or marketing + PLG), then layer in customer success and technical sales as they scale.

The GTM engineer: what changed and why it matters

• GTM engineers bring technical and AI capability to bear on sales and marketing workflows — rebuilding them as agents rather than human-executed processes.
• At Stripe in 2017, the team tried to build a personalised outbound system ("Project Rosalind") — a company universe database with Mad Libs-style email templates. It barely worked without AI.
• At Vercel in 2024, the same concept was rebuilt with a GTM engineer and AI. It now works.
• The GTM engineer shadows top performers in each function, maps their workflow, then encodes it as an agent — deterministic steps plus AI judgment for less structured tasks.
• Start with the most legible, replicable workflows (e.g. inbound lead qualification) before moving to complex ones (e.g. deep enterprise outbound).

The lead agent: a live example

• Vercel ran 10 SDRs on inbound qualification. One GTM engineer built a lead agent in six weeks, spending ~25–30% of his time on it.
• The agent does deep research, pulls database attributes, and drafts a response; a human reviews and sends.
• After the build, Vercel moved from 10 SDRs to 1 managing the agent — the other 9 shifted to outbound.
• Lead-to-opportunity conversion held flat; time-to-first-response dropped because the agent handles after-hours leads instantly.
• The agent runs on Vercel's own infrastructure; total cost is ~$1,000/year versus seven-figure SDR payroll.

The deal bot: AI as sales coach

• Vercel feeds all Gong call transcripts, emails, and Slack interactions into a deal bot agent.
• The bot first ran as a lost-opportunity review tool: in one case, the AE attributed the loss to price; the bot identified the real cause as failure to reach the economic buyer and weak ROI framing.
• Now the deal bot runs in real time — surfacing insights per customer in dedicated Slack channels (e.g. "You're mid-process and haven't spoken to a CFO-level buyer").
• After new product releases, the bot scans calls to flag objections the team handled poorly, generating a sprint backlog of "bugs" in the sales process.

GTM engineer profile and when to start

• The best first GTM engineers come from the sales side, not engineering — ideally former sales engineers or technically minded AEs who understand what good looks like.
• Sales domain knowledge matters: agents modelled on a 2-year rep will encode 2-year-rep habits; a 20-year veteran spots the gaps.
• A rough trigger for bringing in a GTM engineer: ~10 people in go-to-market, with a documented, replicable process.
• Build vs. buy calculus is shifting: internal agents are cheap and fast to build; off-the-shelf tools often solve narrow problems and create a 20-tool stack.

Segmentation: how to carve up the market

• Standard size-based segmentation (SMB / mid-market / enterprise) is necessary but rarely sufficient.
• At Stripe, a second axis was growth rate (consumption-based model → fast growers worth more) and a third was business model (B2B, B2C, marketplace, platform → different products and personas).
• At Vercel, the axes are size, Crux rank (Google's public traffic data — a proxy for spend potential), and workload type (e-commerce, SaaS, crypto, etc. — each requires different language and content).
• Segmentation is a company decision, not just a GTM decision. Grosser delivers the segmentation framework to every new hire at Vercel in their first week.
• Aim for no more than three attributes defining your ICP — beyond that, you can't staff a team around it.

Thinking about go-to-market as a product

• The thesis: as software commoditises, the buying experience itself becomes a differentiator — "we buy a lot of things because of how we feel about them."
• At Stripe, the standard first discovery call was replaced with a whiteboarding session where prospects drew their own payments architecture. Customers left with a useful asset; Stripe learned everything it needed without an interrogation.
• Every touchpoint should add value regardless of whether the prospect buys. Lost deals often return 3–5 years later if the experience was genuinely helpful.
• At Vercel, outreach leads with concrete performance benchmarks (Core Web Vitals, peer comparisons) before any product pitch.

Sales tactics that actually work

• Lead with unique insight: show prospects exactly where they're underperforming, using data they couldn't easily find themselves.
• Go beyond docs: create well-architected guides and implementation blueprints specific to a customer's business model (e.g. Stripe's marketplace setup guide).
• Talk less than 50% of the time. Ask a follow-up question before answering. Help the customer arrive at the conclusion themselves.
• Frame around risk avoidance, not upside: 80% of buyers act to avoid pain or reduce risk; only 20% buy for growth. Enterprise buyers fear career damage from a bad vendor choice.

Building a sales org product and engineering will respect

• The litmus test: put an AE in front of 10 internal engineers — it should take 10 minutes for them to realise the AE isn't a product manager.
• Deep product knowledge earns internal credibility and makes the sales org a genuine R&D function — translating customer signal into roadmap input.
• A 20-person sales team speaks to more customers per week than any product manager; that data pipeline is wasted without disciplined signal extraction.
• Hire a mix: experienced salespeople paired with consulting or finance backgrounds. Each teaches the other — sales rigour meets analytical depth.

PLG, pricing, and comp

• PLG makes sense at the outset for most companies (exceptions: products built explicitly for enterprise from day one). It typically has a ceiling; every large company eventually adds sales.
• Waiting too long to build a sales motion is the most common PLG mistake — outbound takes time to become a predictable engine.
• Pricing should be designed like a product: understand where customers drive value, where you incur costs, and align the model accordingly. Challenge freemium by default — Stripe killed a Billing free trial and saw zero downside.
• Sales comp locks in priorities 12 months ahead; in a fast-moving market this creates rigidity. Grosser doesn't have a clean answer yet but sees this as a structural problem to solve.
• On hiring, value sales experience as a skill but pair it with non-traditional profiles (consulting, finance) who bring analytical and consultative depth.

Timing your first sales hire

• Wait until roughly $1M ARR, but only if you have a repeatable pattern — similar customers, a documentable ICP, and enough consistency that someone else can learn the motion.
• Founders must enable the first salesperson with discovery questions, objection handling, and content — then stay connected to customers to continue informing the roadmap.