Dibein

Cash Flow Copilot

A 13-week forecast that updates itself.

Dibein reads your bank accounts and accounting ledger every day and rebuilds a rolling 13-week and 12-month cash forecast, broken down by customer, vendor, and category, with a confidence band instead of a single guess.

94.2%

median 4-week forecast accuracy across live customers

Most finance teams still forecast cash in a spreadsheet that someone updates on Monday mornings, by hand, from memory of which invoices are probably going to land. It's stale by Tuesday.

Cash Flow Copilot treats forecasting as a pipeline, not a chore. It ingests every transaction, categorizes it against your chart of accounts, learns the payment behavior of each customer and vendor, and recomputes the forecast automatically whenever new data lands — a payment clears, an invoice is issued, a payroll run is scheduled.

The output isn't a single line. It's a band: a most-likely path, plus the 10th and 90th percentile outcomes, so you can see how much your Friday actually depends on whether your slowest-paying customer comes through this week or next.

How it works
01

Connect read-only

Link bank accounts through Open Banking (Qonto, Revolut Business, N26 Business, and SEPA banks via Bridge/Powens rails) and your ledger (Pennylane, QuickBooks, Xero, Sage). Every connection is read-only — Dibein cannot move money.

02

Learn payment behavior

The model builds a payment-timing profile per counterparty from 12–24 months of history: median days-to-pay, variance, seasonality, and how that's trended over the last two quarters.

03

Forecast with a band, not a number

Confirmed items (payroll, rent, signed invoices) anchor the near term. Everything else is projected from behavior and your open pipeline, with an explicit confidence interval that narrows as the date approaches.

04

Refresh on every event

No nightly batch job to wait on. A cleared payment, a new bill, or an edited invoice triggers a recompute in the background, so the number on your screen is never more than a few minutes stale.

Try it

See Copilot in action on a sample workspace.

No signup required. This runs on fixed sample data — nothing here touches your own accounts.

Solen Studio · sample workspace

13-week cash forecast

€65,800
Week 13 balance
Forecast confidence band Baseline Your scenario

Scenario Studio

0 hires
On time
€0
€0

Every slider recomputes the full 13-week forecast against Solen Studio’s sample ledger, the same engine behind Cash Flow Copilot. Nothing here touches real data — this workspace is a fixed sample dataset.

Capabilities

Multi-entity roll-up

Forecast each legal entity separately, then consolidate into a group view with intercompany transfers netted out automatically.

Multi-currency

EUR, GBP, USD, and CHF accounts forecast natively and convert to a reporting currency using a rate you control, not a live spot rate that moves your numbers underneath you.

Confirmed vs. projected split

Every week of the forecast shows how much is contractually confirmed versus modeled, so you know exactly how much of your runway is a guess.

Covenant and threshold tracking

Set a minimum cash covenant or an internal floor and get the forecast week flagged the moment the median path — not just the worst case — crosses it.

Export without lock-in

Every forecast is one click from a clean spreadsheet export, formatted the way your board deck already expects.

FAQ
How far back does it need historical data to build a forecast?

Copilot produces a first forecast within minutes of connecting, using whatever history your bank and ledger expose (typically 12–24 months). Accuracy improves over the first 4–6 weeks as it observes your actual payment behavior directly rather than inferring it from history alone.

What happens to accuracy for a business with lumpy, seasonal revenue?

The model fits seasonality per counterparty and per category rather than one curve for the whole business, so a Q4-heavy retailer and a subscription business on the same instance get separately shaped forecasts.

Can I override a projected line item?

Yes. Any projected (unconfirmed) line can be manually adjusted or annotated, and the model logs the override so next week's forecast reconciles against what actually happened.