Sales & CRM

How to forecast quarterly revenue with AI

Sales & CRM3 AI tools7 steps6 friction points

Want an agent to handle this on your live data? See the Starch version →

Quarterly revenue forecasting is the process of projecting what your business will bring in over the next three months — broken down by product line, customer segment, sales rep, or whatever cut matters most to how you operate. Most operators run this exercise at least once a quarter, often more frequently as plans slip or accelerate. The output drives hiring decisions, vendor commitments, and investor conversations, which is why getting it wrong has real consequences.

The workflow feels like an AI problem because it's mostly pattern recognition on structured data: historical revenue trends, pipeline coverage, seasonal adjustments, churn assumptions. You have the numbers somewhere — in your CRM, your accounting tool, your Stripe dashboard — and you need someone to help you synthesize them into a defensible forward projection. That synthesis step, the one where you connect actuals to assumptions to outputs, is exactly the kind of reasoning LLMs are genuinely good at.

ChatGPT, Claude, and Gemini can all contribute meaningfully here. They can walk you through a bottoms-up model structure, help you apply growth rate assumptions to historical ARR data, sanity-check your pipeline coverage ratios, and draft narrative commentary for whatever numbers you feed them. The catch is that 'feed them' part — you're always the one supplying the data, manually, before the session starts.

Skip the manual work → See the Starch version

Sales & CRM3 AI tools7 steps6 friction points

AI walkthrough

How to do it with AI today

A practical walkthrough using ChatGPT, Claude, and other off-the-shelf LLMs — what they're good at, what you'll have to do by hand.

Tools that work for this

ChatGPTClaudeGemini

Step-by-step

1 Pull your last three quarters of revenue data from Stripe, QuickBooks, or whatever system of record you use. Export it to a CSV or copy the key figures — MRR by month, new bookings, churn, expansion revenue — into a plain text block you can paste into a chat window.

2 Open ChatGPT (GPT-4o) or Claude and paste your historical revenue data. Ask it to identify the growth rate trend, average monthly net new ARR, and churn pattern across the period. Let it do the descriptive work before you ask it to project anything.

3 Export your current sales pipeline from your CRM — stage, deal size, close date, probability. Paste this into the same chat and ask the model to calculate weighted pipeline coverage for the quarter and flag any deals where the close date has slipped.

4 Prompt the model to build a bottoms-up quarterly forecast combining your historical growth rate, weighted pipeline, and any known variables you specify (a big renewal coming, a product launch, a rep departure). Ask it to output the model as a table with assumptions clearly labeled.

5 Run a second prompt asking it to stress-test the forecast: what does the quarter look like if your pipeline converts at 70% of weighted value instead of 100%? What if your largest renewal churns? Have it show the range, not just the point estimate.

6 Copy the model's output into a Google Sheet or Notion doc. Use ChatGPT or Claude to draft the narrative commentary — two to three paragraphs explaining the forecast, key assumptions, and risks — formatted for a board update or investor email.

7 Set a calendar reminder to repeat this entire process next month. Nothing you built in this session persists or updates automatically.

Prompts you can copy

Here is my monthly revenue data for Q1–Q3: [paste table]. Calculate the average monthly growth rate, net new ARR trend, and average churn rate. Show your work.

Here is my current pipeline by stage and probability: [paste CRM export]. What is my weighted pipeline coverage for Q4? Which deals by name are most at risk based on their close date versus stage?

Using the historical growth rate you calculated and the weighted pipeline above, build a bottoms-up Q4 revenue forecast. Show three scenarios: base case, upside (pipeline converts at 120% of weighted), and downside (pipeline converts at 60% of weighted). Format as a table.

Write a two-paragraph forecast commentary for our board update. Assume Q4 base case is $1.2M in new bookings, 4% churn, and net new ARR of $850K. Explain the key assumptions and flag the two biggest risks to the number.

Our largest customer ($180K ARR) is up for renewal in month two of the quarter. Walk me through how to model this as a scenario — both the retention case and the churn case — and show the impact on quarterly net revenue.

Reality check

Where this gets hard

The walkthrough above works — until your numbers change, the LLM hallucinates, or you have to re-paste everything next month.

No live connection to your Stripe, QuickBooks, or CRM data — every session starts with a manual export and copy-paste, which means your forecast is already stale the moment you run it.

Pipeline data from your CRM requires a separate export step, and there's no way to ask the model a follow-up like 'which of these deals slipped since last week?' without pulling a fresh export yourself.

The model has no memory between sessions — the forecast structure, assumptions, and prompt logic you built this quarter are gone next quarter, and you're starting from a blank chat window again.

Outputs aren't reproducible. The same prompt pasted two weeks later returns a differently formatted table or differently labeled scenarios, so comparing quarter-over-quarter requires manual reconciliation.

Scenario modeling requires re-prompting the full context each time you want to test a new assumption, which gets tedious and error-prone as you stack more variables.

There's no place for the output to live that stays connected to the underlying data — your forecast document is a snapshot, not a surface that updates as actuals come in.

Tired of the friction?

Starch runs the whole workflow on live data — no copy-paste, no hallucinated numbers, no re-prompting next month.

See the Starch version →

Starch alternative

The same workflow on Starch

Starch is an agentic operating system — agents build and run the software your work depends on, connected to your live business data. For quarterly revenue forecasting, that means an agent builds you a persistent app that pulls from your actual Stripe, QuickBooks, and CRM data on a schedule, so your forecast reflects what's true today, not what you exported last Tuesday.

Connect Stripe and Plaid once — Starch syncs your actual revenue, charges, and transactions on a schedule. Your forecast baseline is always built from live numbers, not a CSV you remembered to pull.

Use the Scenario Analysis starter app to model multiple forecast outcomes side-by-side. Connect your Stripe actuals as the baseline, then adjust growth rate, churn, or pipeline conversion assumptions to see how Q4 plays out under each set of conditions — without rebuilding a spreadsheet.

Use the Investor Reporting starter app to turn your quarterly forecast and actuals into a polished update automatically — MRR growth, runway, key assumptions, and narrative commentary — delivered to your investor list on the cadence you set.

Connect HubSpot, Apollo, or your CRM from Starch's integration catalog and ask the agent questions like 'what is my weighted pipeline coverage for next quarter?' against live deal data — no export required.

Describe the forecasting surface you actually want in plain English: 'Build me a dashboard that shows quarterly revenue forecast versus actuals, broken down by product line, pulling from Stripe and QuickBooks.' An agent builds it. It stays current.

The forecast app persists and runs continuously — next quarter you're not re-prompting from scratch. Assumptions update, actuals flow in from your connected sources, and the model reflects reality without manual intervention.

Get closed-beta access →

Toolkit