Know it weeks early.
Operations tells you what happened. The predictive layer tells you what's about to — models that forecast the numbers that matter, so a problem surfaces while there's still time to act on it.
Every number on the platform has a history — every payment, return, settlement and balance, posted to one ledger. That history is exactly what a forecast needs. The predictive layer reads it and projects forward: the runway, the churn, the returns and the funding gaps that would otherwise only become visible once they'd already happened.
Because the models sit on the same ledger as operations, a forecast is never a spreadsheet guessing from stale exports — it's computed from the live book, tenant by tenant.
Forecasts on the numbers that decide the quarter.
Runway & break-even
Projected cash position and the date the money runs out — or the month it turns — from actual inflows and outflows on the ledger, not a static model.
Churn & retention
Which accounts are drifting toward the exit, scored from payment cadence and behaviour — so retention is a call you make before the cancellation, not after.
ACH-return prediction
The debits most likely to come back, flagged before submission — so origination quality stays inside NACHA thresholds instead of tripping them.
Funding-gap forecasting
For platforms that front money — payroll, advances, draws — the projected gap between what's paid out and what's collected, days ahead, per counterparty.
Capital underwriting
Advance and credit decisions scored against collateral the platform already holds and values — a BOL, an invoice, retainage — on the same ledger the money moves on.
Cash-flow forecasting
Inflows and outflows projected forward across rails and tenants, so treasury sees the shape of next month rather than reconciling last month.
Fed by the ledger. Read by the operator.
One source of history
Every payment, hold, settlement and return is already on the double-entry ledger. The models train and forecast on that book — no separate data pipeline to drift out of sync.
Per tenant, per vertical
Each vertical's risk profile and payment patterns differ, so forecasts are scoped to the tenant — a factoring book and a rent roll are never averaged into one number.
Surfaced where you work
A forecast that stays in a notebook changes nothing. Predictions surface in the operator view next to the live numbers they're about — and feed the agents in Layer 3.
See → predict → act.
The predictive layer sits between live operations and the agents that act on them. Explore the layer above and the layer below.
Request access → Autonomous agents →