I just watched 2 Anthropic GTM leaders break down how their sales team runs on Claude.

13 AI use cases. Here's the full list.

Plus the part most teams will copy wrong.

The setup:

Claude connects to their whole stack through MCP. Salesforce stays the system of record. Claude becomes the system of intelligence.

Brand and product context get set once, in shared org-level skills. No rep re-explains the company on every prompt.

Read that twice. That's the part to copy.

The 13 below are downstream of it.

The use cases:

1. Market mapping. One prompt sorts the full market into clean, non-overlapping segments. Their cybersecurity map went from weeks to an afternoon.

2. Dynamic account planning. Calls summarize into a rolling doc. The account plan updates itself in a sheet. A color code flags any account untouched for 30 days.

3. Morning brief. Every overnight account update, pulled before the day starts.

4. Call prep. One command builds full prep from the CRM, past calls, and email history.

5. Call coaching. Every call scored against a real methodology. Then Claude reads across the whole team to find where reps keep stalling. One scored call tells a rep what to fix. Reading every call tells you what the team keeps getting wrong. That second part is the point.

6. Customer follow-up. Recap and next steps drafted after every call, against a 24-hour SLA.

7. Pipeline reviews. Claude flags missing fields, stale steps, and ARR that doesn't add up.

8. Forecast scoring. Every deal scored across 14 factors. Not gut feel.

9. Competitive battle cards. Built and refreshed as the market moves.

10. Asset creation. The one-off decks and one-pagers reps used to wait days for.

11. Support front door. Deal desk, legal, and revops requests start in Slack, triaged and routed by Claude.

12. Brainstorming. Reps pressure-test deal strategy with Claude before they act. Ash called this the most overlooked use case in sales. I agree. It's the one that keeps a rep human while everything else gets automated.

13. Enterprise self-serve. With the system above carrying the load, 54% of Anthropic's new enterprise logos now start self-serve.

My take:

The list is the easy part. Every one of these 13 is a weekend project now.

The foundation is the hard part: the MCP wiring and the shared org skills. The unsexy work nobody screenshots.

It's also the only reason any of this compounds.

Most teams skip the foundation and bolt the use cases on anyway. Each one works in isolation. None of them reinforce each other.

Build the foundation first and every use case compounds on the last. That's the whole difference.

So if you run an outbound team.

Start with the layer underneath.

The use cases follow.