A balance is the one number a treasurer cannot change
Module · Liquidity and Funding
A balance is a receipt for a decision already taken. Every movement behind it, the receipt, the payment, the draw, the hedge, was still changeable a week ago.
Prophesee Treasury forecasts the movement rather than restating the balance, and alerts on the projected position rather than the actual one, so the treasurer is looking at the part that can still be moved. Operations owns the behaviour that feeds it.
Every treasurer has a cash forecast. Almost none has a cash prediction.
The thirteen week view is collected by email, assembled in a spreadsheet and reissued each Monday. It carries no confidence band and it is never scored against what happened, so by week six it is a narrative. Not because anyone is careless, but because no tool on the desk can do better.
Sources: The Hackett Group, European Working Capital Survey, Nov 2025 (1,000 largest European non-financial companies, drawn from financial statements) · The Hackett Group, US Working Capital Survey, Aug 2025 (top 1,000 US listed non-financial companies) · Sidetrade Data Lake analysis, Feb 2026 (approximately 285 million invoices captured in 2025) · Association for Financial Professionals, Payments Fraud and Control Survey, Apr 2025 (n=521 corporate treasury practitioners).
From reporting the balance to predicting the movement
ForesightThe forecast is collected from entities once a week and assembled by hand. It carries no band, and it is never scored against actuals, so nobody knows what week six is worth.
Time series models per entity, currency and cash flow category, each projection carrying a confidence interval and the drivers behind it. Time to event models predict when a receipt or a payment actually lands rather than when terms say it should. Monitored for drift and retrained on a schedule.
PulseHeadroom is calculated at signing and again at year end. Between them nobody tests it against the forecast, so a breach is discovered rather than avoided.
Write the covenant, the counterparty limit and the minimum balance in plain English. Rules run continuously against actuals and against the forward projection, so the alert fires when the forecast breaches, not when the balance does. Severity scored and routed to a named owner with the action attached.
HorizonDraw the facility, change the terms, repatriate the cash. Each option is argued in a paper, with no model of the no-action trajectory, so no improvement can be attributed to a decision.
Predict the no-action cash trajectory, then model the intervention: a terms change, a facility draw, a sweep structure, a hedge. Watch the projected curve update, then track actuals against the plan so the committee sees whether the action worked rather than that it happened.
NexusThe group position is assembled from bank portals, ledgers and entity submissions in different charts of accounts, so the answer arrives a day after the question.
Statements, ledgers, entity submissions and facility agreements resolved into one graph, with the same counterparty recognised across every system through fuzzy entity resolution and vector embeddings. Ask in plain language and get the statement line and the clause behind the answer.
Turning liquidity challenges into decisions
The forecast reads the payment date off the contract terms. Behaviour says something different, and the variance is absorbed silently every week.
ForesightEntities submit a number with no band and no driver, and nobody scores last quarter's submission against what happened.
ForesightThe same counterparty appears under three spellings across ledgers, statements and facility documents, so exposure cannot be totalled.
NexusHeadroom is a point in time calculation. Nothing tests it between the two dates it is reported on.
PulseOptions are compared in prose. There is no no-action baseline, so the effect of the choice can never be isolated.
Horizon12 AI applications that could be relevant
A sample of what becomes possible on the decision layer, not a fixed list: each application draws on the same data foundation and audit trail, and new ones are configured on the engines, not built from scratch.
A group cash projection by entity, currency and category, with a confidence band and drivers on every line.
The date each open item is actually paid, predicted per customer and entity with a confidence interval.
Forward exposure by currency and entity, with the band, before the hedge is placed.
Every past forecast scored against actuals by entity and horizon, so accuracy is measured, not assumed.
The cash effect of every scheduled outflow, modelled before the run is released.
Headroom tested against the forecast daily, with the projected breach date named before it arrives.
Counterparty, currency and minimum balance limits written in plain English and watched continuously.
Model the draw, the terms change or the sweep and watch the projected curve update.
Actuals tracked against the modelled plan so the improvement is attributed to the decision.
One cash position across every account, entity and currency, recomputed as statements land.
Every bank, customer and entity reconciled to a single identity across systems.
Where cash sits that cannot move, by entity and restriction, with the clause behind each one.
A day in a predictive treasury
Today: A group cash position assembled by hand from bank portals and entity emails, describing yesterday.
ForesightGroup cash by entity and currency, with a confidence band on every line and the drivers behind the movement.
Today: An entity that will breach its minimum balance, discovered on the day it breaches.
PulseThe forecast crosses covenant headroom in week seven. Severity scored, routed to her, with the two viable options attached.
Today: Draw the facility because the paper argued for it, then hope.
HorizonA terms change tested against a facility draw on the projected curve. She takes the one that holds headroom without the fee.
Today: No way to show which treasury action improved the cycle.
HorizonLast quarter's terms change tracked against the modelled plan. The cycle shortened as projected. That goes to the board.
Tomorrow's cash becomes today's decision.
Predict liquidity before it moves
We agree the metric and the baseline in week one, and measure the result on your data.