Finance & FP&A

How to forecast runway and months of cash with AI

Finance & FP&A3 AI tools7 steps6 friction points

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

Forecasting runway and months of cash means taking your current bank balance, subtracting your net burn rate, and projecting how far that takes you — ideally broken out by expense category, compared against incoming revenue, and updated often enough to actually influence decisions. For most early-stage operators, this lives in a spreadsheet that gets touched once a quarter, is never quite current, and requires thirty minutes of copy-pasting every time someone asks 'how long do we have?'

The workflow feels like an AI problem because the math itself isn't the hard part — the hard part is collecting the inputs, doing the arithmetic consistently, and then re-running it every time a number changes. If you could just describe what you want and have something pull the numbers together, the actual forecast would take seconds. That gap between 'I know how to think about this' and 'I have current numbers in front of me' is exactly where people reach for ChatGPT or Claude.

General-purpose AI tools like ChatGPT, Claude, and Gemini can genuinely help with the analytical layer — writing the formulas, checking your burn rate math, building a projection model from sample data, or stress-testing your assumptions. Where they fall short is everything around that layer: getting the actual numbers in, keeping them current, and making the output somewhere you can return to next month without rebuilding it from scratch.

Skip the manual work → See the Starch version

Finance & FP&A3 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 Export your last 3-6 months of bank transactions from your bank or accounting tool as a CSV. If you use QuickBooks or Plaid, export from there — you want categorized data, not raw debits.

2 Open ChatGPT or Claude and paste in the CSV rows (or a summarized version if it's large). Tell the model what each category represents and ask it to calculate average monthly burn, separated by fixed and variable costs.

3 Ask the model to pull your average net burn and, given your current cash balance, project how many months of runway you have. Specify whether you want it to include expected revenue or calculate gross burn only.

4 Paste in your Stripe revenue data or MRR number separately, then ask the model to recalculate net burn and update the runway projection. You'll likely do this as a second prompt because the context gets crowded.

5 Ask the model to build a simple 24-month forward projection table — current balance minus net burn per month — formatted as a table you can paste into a doc or spreadsheet.

6 Run a scenario prompt: ask what happens to runway if you reduce headcount by one person, or if revenue grows 15% per month. The model will recalculate given the inputs you've already provided.

7 Copy the output into a Google Sheet or Notion doc. This becomes your 'current' forecast until the next time you need to update it, which means repeating this entire process.

Prompts you can copy

Here are my categorized bank transactions for the last 6 months as a CSV. Calculate my average monthly gross burn and break it down by expense category. Identify which categories are growing month over month.

My current cash balance is $340,000. My average net burn is $28,500/month after $14,000 in monthly Stripe revenue. How many months of runway do I have, and at what date does cash hit zero?

Build a 24-month cash projection table starting from $340,000 with $28,500 net monthly burn. Show ending cash balance each month. Flag the month where balance drops below $100,000.

If I hire one engineer at $12,000/month all-in starting in month 3, and my revenue grows 8% per month from $14,000, what does my runway look like? Show month-by-month cash balance for 18 months.

My burn last month was $31,200 and the month before was $26,800. Is this variance significant enough to change my runway estimate? How should I think about which number to use as my baseline?

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 bank or Stripe — every run starts with a manual export, a copy-paste, and hoping nothing got miscategorized since last time.

Context windows cap out around a few hundred rows of transaction data, so if you're past the seed stage with real transaction volume, the model is working with a truncated picture.

Revenue and expense data live in separate systems; you're stitching them together manually in the prompt, which means the net burn calculation is only as accurate as what you remembered to include.

The output is a block of text or a table in a chat window — nothing persists. Next month you run the same prompt chain from scratch and hope the structure comes out the same way.

Scenario modeling requires manually re-entering all your base assumptions every time you want to test a new variable, because the model has no memory of last week's conversation.

The model can't tell you when your numbers have changed — there's no alert when burn spikes, no daily update, no way to glance at runway without running the whole workflow again.

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 — it builds and runs the software that handles this workflow continuously against your live business data, instead of a one-off prompt you re-run manually every month when someone asks about runway.

The Runway Analysis starter app connects Stripe and Plaid directly — Starch syncs your real transaction data and revenue on a schedule, so your burn rate and months-of-cash figure reflect what actually happened this week, not last month's export.

Net burn is calculated from real numbers: Stripe revenue minus Plaid-tracked expenses, categorized automatically. No manual CSV wrangling, no mismatched time periods — the inputs are always current.

The Scenario Analysis app lets you test assumptions against your actual baseline. Describe the scenario you want to model — 'what if I delay hiring two months and revenue grows 10% slower' — and it runs against live Stripe and Plaid data, not a static snapshot you built by hand.

You get 6-month historical trends and 24-month forward projections in a persistent dashboard you can open any morning. The numbers update on a schedule; you don't have to rebuild the model to get a current answer.

If neither pre-built app fits your exact setup, describe what you want in plain English — 'show me runway assuming 90-day collections lag on invoices, with QuickBooks bills as the expense source' — and Starch builds that dashboard for your specific model.

Set up an automation to Slack you a runway summary every Monday morning. One description to Starch; it runs on a schedule without you touching it again.

Get closed-beta access →

Toolkit