01 the business
Joinery and fit-out business
55 staff, project-based work
The owner knows today's bank balance but not next month's. The forecast spreadsheet was built once and is rarely updated. Big payroll and tax weeks arrive as surprises.
02 before
The problem
Most owners know their bank balance today but not where it will be in six weeks. The forecast spreadsheet was built once, nobody has time to update it, and big payroll or tax weeks arrive as a surprise.
- No current view of the next six weeks.
- Customers' real payment habits are ignored.
- Tight weeks are found too late to act.
03 the build
What the AI does
- Step 1: Pulls the bank balance, open invoices and bills, and recurring payments from Xero or QuickBooks every Monday.
- Step 2: Estimates when each customer will actually pay based on how they have paid before, not just the due date.
- Step 3: Adds known outflows such as payroll, rent, loan repayments and tax dates, plus likely deals from HubSpot or Pipedrive.
- Step 4: Rebuilds a week-by-week forecast and highlights any week where cash drops below the floor you set.
- Step 5: Writes a short summary of what moved since last week and sends it to the owner by email or Slack.
paused: waiting for a person to approve
How it was put together
- Pulled bank balance, open invoices, bills and recurring payments every Monday.
- Estimated when each customer will actually pay, based on past behaviour.
- Added payroll, rent, loans, tax dates and likely deals from the CRM.
- Rebuilt a week-by-week forecast and flagged weeks below the cash floor.
04 human in the loop
Where a person stays in charge
The owner or finance lead reviews the forecast and adjusts any assumptions, such as a big deal slipping or a planned purchase, before it is shared with the bank, investors or the wider team.
- The owner reviews assumptions before the forecast is shared with anyone.
- Each line shows where it came from.
- The forecast never moves money or changes the books.
05 rollout
From kickoff to live in 3 weeks
- Week 1Connect the books and CRM; agree the cash floor and known outflows.
- Week 2Back-test against the last three months of actual cash.
- Week 3Start the Monday summary.
06 after
What changes
You see tight weeks coming early enough to act: chase a payment, move a purchase or talk to the bank. The forecast stays current because nobody has to rebuild it by hand.
Assume a monthly payroll of $250,000 and a customer who usually pays 20 days late on a $120,000 invoice due the same week.
Expecting the $120,000 on its due date would leave you $120,000 short in payroll week. A forecast using real payment habits shows the gap weeks ahead.
What to measure
- Forecast versus actual cash, by week.
- Weeks below the cash floor spotted in advance.
- Time spent updating the forecast.
- Use of overdraft or short-term borrowing.
07 evidence
What the research says
Gartner research put the average cost of poor data quality at $12.9 million a year per organisation.
Gartner, Data quality: why it matters and how to achieve it (opens in a new tab) · 2020
About 75% of generative AI's potential value sits in four areas: customer operations, marketing and sales, software engineering, and R&D.
08 lessons
Mistakes to avoid
- Using due dates instead of real payment habits.
- Forgetting irregular outflows such as tax and insurance.
- Sharing without review. Assumptions need the owner's eye.
09 questions
Common questions
- How far ahead does it forecast?
- Usually 13 weeks, with the first six in the most detail.
- Can it model scenarios?
- Yes. You can test what happens if a big deal slips or a purchase moves.