4 stories from Jeanne DeWitt Grosser (COO at Vercel, ex-Stripe) on how AI is reshaping Go-To-Market

1) How Stripe tried to automate customer discovery before AI was a thing
In 2017, Stripe decided to launch an outbound customer discovery function.
Normally you’d staff this with ~30 SDRs, but Stripe ran lean, and Jeanne was given 4 people.

To make up for the lack of hands, the team created Project Rosalind: a massive database of potential companies with attributes like business model, industry, needs, and org structure.

The ambition was big: generate personalized outreach emails as a “template with variables” where the system would plug in the right arguments, case studies, and value props based on the company type.
But without AI, the accuracy just wasn’t there. Real personalization wasn’t achievable. The idea was right, but the tech wasn’t ready yet.

Years later at Vercel, they did the same thing again, but this time with AI agents that actually worked.

2) How Vercel cut a team of 10 SDRs down to 1 with AI
At Vercel, Jeanne set a goal: train an agent to perform like the best SDR on inbound leads.
A GTM engineer took a simple but brilliant route: he fully replicated the best SDR’s workflow:

• what they research

• how they find data

• how they decide a lead is “good”

• how they craft the first message

Six weeks later, the agent did it all end-to-end:

• analyzed the inbound lead

• gathered company context

• generated a personalized reply

• suggested the next step

A human just hit “send.”

Result: one SDR does the work of ten. The rest were moved to outbound prospecting, where the human touch matters more.

Agent cost: about $1,000/year.
Cost of a 10-person SDR team: over $1M.

3) How sales at Vercel started running in sprints
Vercel ships product updates so frequently that sales can’t always keep up.

To avoid drowning in chaos, they set up a system:

• An AI agent reviews all calls (via Gong transcripts) and flags where a rep explained a new feature incorrectly, missed a key point, handled an objection poorly, or confused the customer.

• Once a week, the team reviews these like bug reports.

• The next week becomes a “sales bug-fix sprint”: update demos, refine messaging, and practice better phrasing.

Sales starts to look like engineering: measurable, predictable, and continuously improvable.

4) Build vs Buy: why internal agents beat off-the-shelf tools
The market is flooded with AI sales tools right now, but they usually have two problems:

• They only solve slices of the workflow.

• They don’t know your internal context.

An internal agent, on the other hand:

• plugs directly into your CRM, email, Slack

• uses your data

• mirrors the logic of your best people

• costs basically nothing

For example, Vercel built a “lost-deals analysis agent” in two days with one engineer. Now it finds the real reasons deals are lost, often ones the team hadn’t even considered.

Source code for the mentioned agent.
Original interview with Jeanne: