AI Cash Flow Forecasting: One Live Runway Figure Every Morning

AI cash flow forecasting works when an agent rebuilds your 13 week cash forecast every morning from the bank feeds, the open invoices and bills in your ledger and the payroll calendar, then sends the CFO one runway figure and the 3 lines that moved it since yesterday. The agent's job is timing, meaning when each customer will pay, while the arithmetic stays in a model it can't edit, and most finance teams get week one error under 5% within 8 weeks.

Why most AI cash forecasts go wrong

The usual first attempt is a tool that fits a curve to 24 months of bank balances and draws the line forward. It looks clever in the demo. Then the quarterly VAT payment lands, or payroll moves from the 28th to the 25th, and the curve has no idea either thing exists. Cash isn't a trend. It's a list of specific payments with dates on them, and most of that list is already sitting in your ledger.

The cost shows up in trust. If the runway figure jumps by 3 weeks on a Tuesday when nothing happened, the CFO stops opening it, and by month 2 the team is back to the spreadsheet someone rebuilds every Friday afternoon. A 150 person company can find a covenant breach 10 days out when the only cash view it has is last Friday's. A cash forecast earns its keep in the week things go wrong, so it has to be built from facts first and predictions second.

The five steps I use to build an AI cash forecast

1. Start from the bank, every account, every morning

Connect a daily feed for every bank account, every currency and every card. Most companies of 50 to 300 people have 4 to 12 of them, and there's always one savings account nobody remembers. Each morning the agent pulls the closing balances and checks that the sum matches the opening line of the forecast to the penny. If it doesn't, it stops and says which account is out. A forecast that starts from the wrong number is wrong in every one of its 13 weeks.

2. Build the 13 weeks from items you already know

Before any prediction, load the certain items. Open sales invoices and their due dates. Open supplier bills. Payroll and pension dates. VAT, PAYE and corporation tax on their calendar. Rent, loan repayments, software subscriptions and signed purchase orders. In most companies I work with, 70 to 85% of the next 13 weeks of payments are already known on day one. Write each one as a rule with a date, so "payroll, £410,000, the 25th, the Friday before if it falls on a weekend" sits in the model, not in someone's head.

3. Let the agent predict timing, customer by customer

This is where AI earns its place. Your ledger says an invoice is due in 30 days. Your customer's history says they pay on day 47, and later in December. The agent reads each customer's last 12 paid invoices, works out their own pattern and places every open invoice in the week it's likely to land. It does the same for supplier payment runs and for new sales from the CRM at the stage probability you set. It also flags anything odd: a customer who's paid on day 20 for a year and is now at day 35 goes to credit control that morning.

4. Send one figure and the 3 lines that moved it

The output is a short note in the CFO's inbox before 7am. Weeks of runway, cash at the lowest point in the 13 weeks and which week that is, the change since yesterday, and the 3 items that caused it. Nothing else. The model does the sums in a locked workbook or a saved query, and the agent writes about the result. It never types a number into a cell it wasn't given.

5. Score it every Friday and run it in shadow for 8 weeks

Each Friday, compare what the forecast said for the week with what the bank shows, line by line. I aim for under 5% error on week one and under 10% on week 4. Run the agent beside the old spreadsheet for 6 to 8 weeks. When it beats the spreadsheet 4 Fridays in a row, switch the spreadsheet off. If one line keeps missing, it's nearly always a rule nobody wrote down, a quarterly bonus or an annual insurance payment. Write the rule, and move on.

From my operating seat

I'm Hayat Amin, an AI-native CFO, AI operator and AI coach for CFOs who teaches finance leaders how to build AI into finance operations. I spent twenty years in the C-suite, with three exits and three FT100 listings. Now I build and run AI agent systems inside companies, across finance, sales pipeline, onboarding and compliance evidence, and I coach CFOs one to one on the same work.

I've run cash through a sale process, and a buyer reads your cash forecast the way a doctor reads a pulse. If the 13 week view you sent in week one of due diligence is close to what happened by week 13, every other number you give them gets easier to believe. Inside the companies I work with today, the live runway figure is the first thing the CFO reads each morning, built overnight from the bank feeds, and the same figure feeds the cash page of the board pack. I don't want a cleverer forecast. I want one number the CEO, the CFO and the board all read the same way.

Can AI do cash flow forecasting in Excel?

Yes, and for a first build I'd keep the forecast in Excel. The workbook stays the model: the formulas, the 13 columns, the runway line. The agent writes into one input tab each morning, the receipts it expects by week and the payments it knows about, and never touches a formula. Your team keeps the file they trust, and you can see every number the agent put there.

How accurate is AI cash flow forecasting?

Accurate enough to run the company on, if you measure it. I aim for under 5% error on the first week and under 10% on week 4, scored every Friday against what the bank shows. The gain over a manual forecast comes from receipts, where most teams assume customers pay on terms. An agent that knows a customer pays on day 47, not day 30, closes most of that gap.

What is a 13 week cash flow forecast and can AI build one?

It's a week by week view of cash in and cash out for the next quarter, built from real items rather than the profit and loss. Banks and lenders ask for one when headroom is tight. An agent can build and refresh it daily from the bank, the ledger and payroll, but a person still sets the rules for tax dates, debt and anything one off, and signs the version that goes to a lender.

Where I come in

This is the work I do with CFOs one to one and inside finance teams. In the first session we connect the bank feeds and list every known payment for the next 13 weeks, and within 2 weeks the agent is sending the morning note in shadow beside your spreadsheet. The profit and loss side of the same build is in AI financial forecasting, the cash page it feeds is in AI board reporting, and the tools I use day to day are in Claude for finance teams. The map of the whole topic is at AI for CFOs, the end state is at the AI-native CFO, and the sessions are at coaching for CFOs. How many weeks of runway do you have this morning, and how sure are you? Start at meethayat.com and tell me.