In partnership with
Over the weekend, I listened to Lenny Rachitsky's podcast with Jeanne DeWitt Grosser, Vercel's COO. The numbers from one story are worth paying attention to: six weeks, one engineer working maybe 25-30% of his time, built an AI agent that now does the work of nine full-time SDRs handling inbound lead qualification.
Annual cost to run the agent: roughly $1,000. Previous cost for those nine SDRs: well over $1 million in salaries.
The agent maintains the same lead-to-opportunity conversion rate the humans did, and actually does it faster because leads don't sit in queues overnight. By the way, those nine SDRs weren't fired. They got moved to outbound prospecting, work that actually uses their full capacity instead of the repetitive inbound grind.
But the economics are just the opening act.
Vercel built a second agent—the "deal-bot"—that runs against every Gong transcript, every Slack message, every email in their sales process. When their biggest Q2 loss came in, the account executive coded it as "lost on price." Standard story. The deal-bot analyzed the entire deal history and returned a different verdict: you never actually reached an economic buyer, and when you discussed ROI with the people you did talk to, your message clearly didn't land. Real loss reason: failure to demonstrate value. That insight changed what Vercel's marketing team needed to build—not price sheets, but value quantification frameworks.
The deal-bot now runs in real-time during active deals, feeding insights into Slack channels. "You're this far into the sales process and haven't talked to an economic buyer—you should think about that." It's not just automating tasks but seeing patterns humans miss or won't admit.
Who's building this stuff?
The “GTM Engineers” at Vercel aren't traditional software engineers or data scientists. They're former front-end developers who moved into sales. They understand both the code and the workflow. They can shadow a top-performing SDR, watch them work across seven open tabs (LinkedIn, company research, ChatGPT, various databases), and translate that human process into an agent workflow. Then they let the agent make the call but keep a human in the loop to review and hit send. The human reviews become training data. Eventually the human says "yes" enough times that you're confident letting the agent run. The profile that's working: sales experience first, technical chops second.
The build-versus-buy calculus is shifting underneath all of this. These agents aren't that hard to build anymore, and they're not expensive to run. Your specific context—your sales process, your customer insights, your workflows—is actually your competitive advantage. Generic software can't encode that.
Vercel is already finding that building their own agents is easier and more effective than waiting for perfect vendor solutions in a space where everything is still figuring itself out.
The skills that matter for marketing teams are changing. It's not about learning to prompt better, though that helps.
It's about being able to articulate your workflow clearly enough that someone can encode it. It's about understanding where AI genuinely adds leverage versus where you're automating something that didn't need to exist in the first place.
If you're leading a team, the person who deeply understands your marketing process AND can translate it into an agent workflow is going to be extraordinarily valuable.
That might be someone from your team who learns to code. It might be an engineer who embeds with your team long enough to understand what you actually do. Either way, the combination is what unlocks this.
The companies that figure out this build motion early—and develop the internal capability to do it—are going to move at a different speed than everyone else. Worth thinking about who on your team has both the domain knowledge and the technical curiosity to become your first "GTM Engineer."
— Torsten & Peter
Our sponsor today:
Startups who switch to Intercom can save up to $12,000/year
Startups who read beehiiv can receive a 90% discount on Intercom's AI-first customer service platform, plus Fin—the #1 AI agent for customer service—free for a full year.
That's like having a full-time human support agent at no cost.
What’s included?
6 Advanced Seats
Fin Copilot for free
300 Fin Resolutions per month
Who’s eligible?
Intercom’s program is for high-growth, high-potential companies that are:
Up to series A (including A)
Currently not an Intercom customer
Up to 15 employees
Our newsletter partner today:
Stop letting weekly publishing drain your time or your team's time.
LetterOperators delivers a ready-to-send Beehiiv newsletter every week for $499/month — no hiring, no freelancers, no distractions — just consistent, high-quality publishing done for you.
We build a brand-trained AI workflow paired with human editing — refined through years of multi-LLM editorial work — so you stay focused on strategy instead of content cycles.
Subscribe to AI Operators for weekly insights on using AI to scale newsletters — and see the system CMOs rely on for consistent output without extra headcount.
Become a Pro member
Upgrade to get access to subscriber-only content.
Upgrade for $9/monthThe subscription gets you:
- Workshop library with prompts and transcripts
- AI CMO self-paced course
- Resource library with guides, frameworks, tools
- 1:1 AI marketing advisory call
