Write Pipeline Forecast Commentary from CRM Data

Workflow area: Sales · Tier: Advanced · Hands-on time: ~40 minutes

The dangerous part of forecasting is not arithmetic, it is optimism: the deal that has slipped twice but stays at 80%, the "verbal commit" with no paper. A model reading your pipeline coldly has no emotional stake in any deal — used correctly, that detachment is the feature.

When to use — and when not to

• Use it for the recurring forecast rhythmweekly or monthly commentary where the format repeats and the data is structured. That is where prompt reuse compounds.
• Use it as a pre-mortem before the forecast callthe AI's challenge questions are the ones your VP will ask; better to hear them an hour early.
• Skip it for the commit number itselfthe forecast call you make is a judgment you own. AI drafts commentary around your number; it does not pick the number.
• Skip it when the CRM data is fictionif close dates and stages are not maintained, fix hygiene first; commentary generated from stale data is confident nonsense.

The recipe

1. 1.

   Export and anonymize the pipeline. From your CRM, export open opportunities for the period: deal ID, stage, amount, close date, age in stage, times close date has moved, last activity date, and forecast category. Replace customer names with Deal-01, Deal-02 (keep a private mapping sheet). Add two context numbers the model cannot know: your quota for the period and quarter-to-date closed-won.

2. 2.

   Get the computed picture first — no prose yet. Upload to ChatGPT (data analysis mode) or Claude:

   Copy

   This CSV is my open sales pipeline. Quota for the quarter: [X].
   Closed-won so far: [Y]. Today's date: [DATE].
   
   Using real computation on the data:
   1. Total pipeline by forecast category and by stage.
   2. Coverage ratio: open pipeline ÷ remaining quota.
   3. Slippage list: every deal whose close date moved 2+ times, with
      amount and current stage.
   4. Stale list: deals with no activity in 21+ days, by amount.
   5. Concentration: what % of pipeline sits in the top 3 deals?
   
   Show the numbers as tables. Do not write commentary yet.
   

   Reconcile the totals against your CRM dashboard before going further — a misparsed column here poisons everything downstream.

3. 3.

   Run the skeptic pass on your call. Tell it your intended forecast and let it argue:

   Copy

   I intend to forecast [Z] for the quarter. Act as a skeptical sales
   director reviewing this pipeline.
   
   - Which deals am I most likely over-crediting, and why? Use the
     slippage, staleness and stage-age evidence in the data.
   - Which risks does my number depend on that I have not priced in?
   - What would have to be true for [Z] to land? List the 3 deals it
     hinges on.
   - Ask me the 5 hardest questions a VP would ask about this pipeline.
   

   Answer the questions honestly — in writing, to yourself. Adjust your call if the evidence says to. This step is the playbook's real value.

4. 4.

   Draft the commentary from verified numbers and your final call.

   Copy

   Write forecast commentary (max 350 words) using only the computed
   figures above and my final call of [Z].
   
   Structure:
   - The call: number, confidence, change vs last period.
   - Coverage & health: coverage ratio, concentration, stale pipeline.
   - Key deals: top 3 by amount with stage and honest risk note each.
   - Risks & slippage: what moved, what might move again.
   - Actions: 3 specific things the team does this week.
   
   Tone: factual, no hedging filler, no invented deal details. Refer
   to deals by their Deal-ID.
   

   Sample of the register you want:

   Copy

   Call: 4.2 Cr against 5.0 Cr quota (84%), down from 4.5 Cr last
   week after Deal-07 slipped a third time. Coverage on the remaining
   gap is 2.1x, but 46% of open pipeline sits in Deal-03 alone —
   this quarter lands or misses on that one deal.
   

5. 5.

   Re-identify and fact-check. Swap Deal-IDs back to real names in your document (not in the chat), then verify every number in the commentary against the step-2 tables and your CRM. Check the actions are ones you actually intend — models love inventing plausible next steps nobody agreed to.

Make it yours

• Week-over-week delta version. Keep last week's export; upload both and ask specifically what changed — new deals, stage moves, slips, closed-lost. Delta commentary is what forecast calls actually discuss.
• Rep-level rollups. For managers: group the computed tables by rep and generate one short paragraph per rep — coverage, biggest risk, one coaching question — before your 1:1s.
• Excel-first shops. If uploads are not allowed, compute steps 1–2 with pivot tables in Excel and paste only the aggregated tables into the chat for steps 3–4; the skeptic pass works fine on aggregates plus your top-10 deal list.

Quality check

• Every figure in the commentary reconciles to the computed tablesand the tables reconcile to the CRM.
• The forecast number is yoursthe AI challenged it; it did not choose it.
• Each top deal carries an honest risk noteno deal described only by its upside.
• Slipped deals are named as slippedmoved close dates are stated, not silently absorbed.
• Actions are real commitmentssomeone will actually do each one this week.