What Anthropic's sales team built in Q1 2026 — and what it tells us about the next wave of AI adoption

Eleanor Dorfman, Head of Industries at Anthropic, speaking publicly about how their sales team rebuilt itself around AI in early 2026.

There is a particular kind of organizational story that I find endlessly instructive: the one where a team is forced to innovate not because they planned to, but because reality left them no other choice.

Eleanor Dorfman, who leads commercial sales at Anthropic, told exactly that story at a recent conference. Worth watching in full (above) — and she has written about some of the specifics on LinkedIn, including the Cowork-based weekly workflow her team runs and Anthropic’s broader findings on enterprise agent adoption. The short version of the talk: Anthropic launched Opus 4.6 in December 2025. Demand went, in her words, “vertical.” The team came back from winter break to a pipeline they had not planned for, could not hire fast enough to serve, and had no playbook to address. What followed was a quarter of improvised but remarkably systematic reinvention.

I want to unpack what they actually built, because I think it illustrates something important about where AI adoption is heading — not in the abstract, but in the specific, unglamorous details of how a sales team actually operates day to day.

The instinct they had to fight first

The natural response to a demand spike is to hire. More reps, more coverage, more capacity. Dorfman is candid that this wasn’t really an option — not at the pace the market was moving, not without sacrificing the hiring bar they cared about. So they were left with a different question: how do you increase the effective output of the team you already have?

The answer they arrived at is deceptively simple: encode the habits of your best reps, and make those habits available to everyone else as AI-powered tools.

“We have to get reps in the door, they go through boot camp, we give them a territory, and we give them a sales plug-in.” — Eleanor Dorfman, Head of Industries, Anthropic

That plug-in — a combination of MCP connectors and what they call “skills” — is what I want to dwell on, because it is not magic. It is pattern recognition applied to institutional knowledge.

What they built, concretely

They identified five things their best reps were doing that everyone else was doing inconsistently or not at all, and turned each into a reusable skill inside Claude.

1. The morning brief

Pulls from Gmail, Gong, Slack, Salesforce, Google Docs, and calendar. Surfaces what needs attention today — unanswered emails, open action items, upcoming calls, centralized initiatives. Every AE’s day starts here. Dorfman gets hers at 7am via Slack.

2. Call prep

Before any customer call, one slash command generates a briefing: who’s on the call, their background, the deal history, discovery questions worth asking, competitive positioning, and what a good outcome looks like. Sent directly to the AE.

3. Customer follow-up

Extracts action items from Gong calls, emails, Salesforce notes, and Slack. Drafts the responses. Puts them in your inbox. Reminds you the next morning if you haven’t sent them. The goal: a 24-hour SLA on every customer commitment, with humans still in the loop on sending.

4. Competitive intel

Not a static battle card maintained by product marketing once a quarter. A dynamic, customer-tailored view of the competitive landscape generated fresh for each deal, because the landscape is changing daily.

5. Asset creation

Custom collateral — one-pagers, proposals, ROI calculators, prototypes — generated during or after a call, tailored to the specific stakeholder and moment. Previously, this required either being a top-five deal or having a friend on the design team.

What strikes me about this list is how un-exotic it is. None of these are capabilities that required Anthropic to invent something new. They are just the things your best salespeople have always done — research thoroughly before calls, follow up reliably, stay current on the competition, make customers feel like the collateral was made for them — now made available as baseline behavior for everyone.

The timeline is the interesting part

It is easy to read a conference talk like this and absorb it as a polished success story. But Dorfman is careful to give you the timeline, and the timeline tells a different story.

Q1 2026 — as it actually happened

• Dec ‘25 — Opus 4.6 launches. Demand goes vertical. Team returns from winter break to a pipeline no one planned for.
• Jan ‘26 — Enterprise self-serve MVP shipped. “How do you build an AI-native sales org from scratch?” becomes the actual question of the quarter, not a strategy exercise.
• Feb ‘26 — Self-serve funnel goes into production. Skills deployed. Supporting function workflows (deal desk, legal, RevOps) restructured around Slack as the front door.
• Q1 end — 54% of new enterprise logos closing through the self-serve funnel. Morning brief, call prep, and follow-up skills in daily use across the team.

That is a quarter of reactive construction — late nights reconciling billing, people physically walking over to RevOps to get approvals, reps on the East Coast and in Europe staying up to chase deal desk responses. The outcomes are real, but so is the cost of building under pressure.

What this tells us about AI adoption more broadly

I have been thinking for a while about why AI adoption inside enterprises has been slower than the headline numbers suggest. I think Dorfman’s story points at the real bottleneck, which is not model capability — it is the organizational and systems work required to turn model capability into reliable daily behavior.

The skills they built are not complicated prompts. What made them valuable was the connective tissue: MCP integrations into Salesforce, Gong, Slack, Gmail, Google Docs, Ironclad. The context routing. The governance to make sure proposals stay within policy. The muscle memory of getting every rep to actually use these tools every day. That is months of work even if you know exactly what you’re building.

A pattern worth naming: The most valuable AI deployments I have seen don’t introduce new workflows — they make the best version of existing workflows available to everyone. Anthropic’s skills are just the habits of their top reps, encoded. The leverage comes from closing the gap between median and excellent, not from reinventing the motion entirely. Dorfman’s read on how enterprises are building AI agents in 2026 is worth reading alongside this — the pattern holds across functions, not just sales.

There is also something worth noticing in how they handled the supporting functions — deal desk, legal, RevOps, customer support. Dorfman describes making Slack the “front door” for all of these, with Claude triaging tickets, pulling context from Salesforce and Gong and email, and either resolving or escalating with full background attached. The insight is that sales is not an island, and the bottleneck in a fast-growing team is often not the reps — it is the coordination overhead around the reps.

A practical note, if you’re thinking about building this yourself

This brings me to the reason I’m writing about this now, which I’ll be transparent about.

At AmpUp, we have spent the last year building exactly the infrastructure Dorfman describes — not for Anthropic’s sales team specifically, but as a platform that any revenue team can deploy without the quarter of construction. The morning brief, call prep, follow-up, competitive intel, and asset creation skills she describes are all things AmpUp ships out of the box, connected to the same tools her team spent Q1 wiring together.

I say this not to diminish what Anthropic’s team built — watching a team move that fast under genuine pressure is genuinely impressive. I say it because if you are a sales leader watching that talk and thinking “I need to build this,” the better question might be whether you need to build it at all.

Most companies don’t get a controlled runway to do this work. They get the same demand spike, the same hiring constraint, and then they spend six months rediscovering the same lessons. The patterns are not new. The plumbing can be.

If you’re curious what this looks like in practice for your team, AmpUp is worth a look. Happy to share more on how we think about the skills framework and what it takes to actually get a sales team using these tools daily — that last mile is where most of the work lives.

And if you watched the Dorfman talk and took away something different, I’d genuinely love to hear it. The thinking on this is still evolving. For more from her on deal strategy and account planning — the more durable parts of the sales craft underneath all of this — this Common Room conversation is worth your time.

All quotes are drawn from Eleanor Dorfman’s public conference talk, linked above. AmpUp is not affiliated with Anthropic.