How to Use Claude for Sales Forecasting: The Hannah Skill Guide

Every sales team runs the same quarterly ritual: a number gets committed, leadership believes it, and then reality lands twenty per cent lower. The gap is rarely effort — it is method. When people try to use Claude for sales forecasting with an open prompt, they get a tidy-looking projection built on the most dangerous assumption in sales: that the stage a deal sits in tells you whether it will close. AI sales forecasting only helps if it forecasts the way a disciplined analyst does.

Hannah is that discipline. She is a forecasting persona you load once into Claude, ChatGPT, or any AI chat, and her first principle is blunt: qualification over stage. A deal parked in “Proposal” with no economic buyer identified is not really in Proposal, whatever the CRM says. From that one rule, a forecast you can actually defend to the board starts to take shape.

Why generic AI forecasts miss

A bare chatbot trusts the labels. It takes CRM stages at face value, applies textbook stage probabilities that have nothing to do with your business, and quietly absorbs the optimism that reps and managers build into every pipeline. It never inspects a deal — never asks whether budget is confirmed, whether there is a timeline, whether anyone with signing authority is even involved. So it produces a number that looks rigorous and behaves like a guess.

What changes with the Hannah skill

Hannah judges deals on what is known, not where they sit. She enforces clear forecast categories — commit (you would stake your reputation on it), best case (likely, missing one element), and pipeline (possible, not fully qualified) — and refuses to let them blur together. She anchors projections on your real historical close rates rather than assumed probabilities: what actually closed from best case last quarter, not what should have. And she inspects at-risk deals, surfacing the optimism that masquerades as qualification.

What it actually produces

You get a forecast methodology with category definitions, a commit / best case / pipeline split weighted by your own close rates, a pipeline coverage analysis, an at-risk deal list with the qualification gap named for each, and a weekly forecast report you can take into a leadership meeting. Ask it to pressure-test the number — “which commit deals would not survive an inspection call?” — and it will tell you.

How to get the most out of it

Give it the real pipeline and, crucially, your historical close rates by category — that history is what turns a guess into a projection. Be honest about qualification rather than hopeful, and let it challenge you; the value is in the pushback, not the agreement. Treat its commit number as the floor you defend, not the ceiling you hope for.

Who this is for

Founders carrying the number themselves, sales managers running a forecast call, RevOps teams building the process, and anyone who has to stand in front of the board and be believed. It works with Claude, ChatGPT, or any AI chat that accepts a system prompt. For the rest of the revenue motion, the sales skills collection covers outreach, deal qualification, and pipeline reviews — each a focused assistant rather than a general chatbot.