Anthropic published a behind-the-scenes look at how its own sales leader runs the week with Claude, and the takeaway is simple: AI works best when it’s wired into a repeating rhythm, not summoned ad hoc. According to Anthropic, the value compounds when Claude shows up at the same checkpoints every week, with the same prompts, against the same data. What stands out here is the discipline. Treating Claude like a calendar habit beats treating it like a magic box.
Here’s a practical weekly framework you can lift from that approach. Use it as a starting structure, then tune it to your stack.
Quick Start
You’ll learn how to build a five-day Claude routine for sales work: pipeline review, prep, follow-ups, forecasting, and reflection. You’ll need a Claude account (Pro or Team), access to your CRM exports (CSV is fine), call recordings or transcripts, and a notes app for prompts you want to reuse.
Monday: Pipeline Triage
Start the week by giving Claude a CSV of your open opportunities. Ask it to flag stalled deals, missing next steps, and accounts with no activity in 14+ days. Why it matters: most reps lose deals to silence, not competitors. Claude surfaces the silence fast.
Reusable prompt: “Review this pipeline. Group deals by stage. Flag every deal with no logged activity in 14 days and propose a specific next action for each.”
Tuesday: Call Prep
Feed Claude the prospect’s website, recent news, LinkedIn bios of attendees, and your previous call notes. Ask for a one-page brief: their priorities, likely objections, three questions to open with, and one insight you can bring.
Tip from the Anthropic playbook: keep briefs short. A page you actually read beats a deck you don’t.
Wednesday: Follow-Up Drafting
Paste in call transcripts or notes. Have Claude draft follow-up emails, recap notes for the buyer, and an internal Slack update for your manager. Three artifacts from one input.
Warning: always edit before sending. Claude drafts the scaffolding, you add the judgment. If the message sounds like a template, your buyer will smell it.
Thursday: Forecasting and Account Plans
Give Claude your weighted pipeline and ask it to stress-test your commit number. Have it list which deals are at risk of slipping and why. For your top three accounts, ask for an account plan: stakeholders, gaps, expansion paths.
This is where the compounding kicks in. After a few weeks, Claude can compare this week’s forecast to last week’s and tell you what shifted.
Friday: Reflection and System Updates
End the week with two passes. First, ask Claude to summarize the week’s wins, losses, and patterns across calls. Second, refine your prompt library. If a prompt produced sloppy output, rewrite it. If one produced gold, save it.
Anthropic’s point throughout the piece: the prompts themselves are an asset. Treat them like reusable code.
Why This Routine Beats One-Off Use
Most people try Claude once, get a so-so answer, and walk away. Sales leaders who get real lift run it on a schedule against structured inputs. The model gets better outputs when it sees the same context shape every week, and you get faster because you stop reinventing prompts.
This is significant because it reframes the AI conversation in sales from “will it replace reps” to “which reps will out-prepare everyone else.” The ones with a Claude routine will.
Next Steps
- Pick one day this week and run the matching block above.
- Save every prompt that works in a single doc. That doc is your moat.
- After two weeks, audit: which Claude outputs actually changed a deal? Double down on those, kill the rest.
- Share the routine with your team so the prompt library grows beyond one person.
For the full account of how Anthropic’s sales leader structures the week, check the original post on Anthropic’s site.