True Available Cash
The balance says you have it. It does not say you can spend it.
A concept for the finance lead who has to answer the one question a bank account cannot: how much can we safely spend today, and is this particular decision one of the safe ones.
- $84,500Shown by the bank
- $29,600Actually spendable
- 6 rulesBehind every verdict
Eighty-four thousand in the account.Twenty-nine of it is yours.
A business banking screen shows one figure and presents it as money. Most of it is already spoken for: owed on a date that has not arrived, reserved against a tax bill, or sitting inside an invoice that may never be paid. The balance is accurate, and it is not the answer to any question anyone is actually asking.
- Reserved$21,000Tax and buffersAlready belongs to someone else, on a date that has not arrived yet.
- Committed$24,700Payroll and suppliersOwed, scheduled, and not movable without breaking something.
- At risk$9,200Overdue and disputedA $9,200 invoice, twelve days late and formally disputed. $2,800 of it is expected to settle; the rest may not arrive at all.
- Available$29,600What survives all threeThe only figure in this list that can be spent. The overview leads with this same number under its product name, Truly Available.
The gap is $54,900: sixty-five per cent of the balance. That is not a rounding error, it is most of the money. And it is arithmetic a finance lead is expected to carry in their head, correctly, across four accounts at three institutions, every time somebody asks whether the company can afford something.
Can we spend $12,000 on equipment today?
Knowing that $29,600 is available does not settle it. $12,000 is comfortably less than $29,600, and the answer is still not a straightforward yes. Everything after this chapter is about why.
A better numberwas not the answer.
There is an obvious version of this product that stops at the previous chapter: surface Truly Available instead of Total Cash, colour the buckets, ship it. It would be a more honest dashboard, reporting a more accurate figure, to somebody who then has to do all of the actual thinking on their own.
But a finance lead is not employed to know the balance. They are employed to decide. Every accurate figure on that screen is an input to a question the screen refuses to answer, and the distance between seeing financial data and acting on it is precisely where the mistake gets made.
So the unit of this product is not the balance. It is the decision: proposed, evaluated, explained, and kept. Everything after this chapter is what that move cost.
Twelve thousand, judged againsta position, not a balance.
$12,000 is less than $29,600, so a balance check returns yes and stops. The concept does not check the balance. It runs the decision against every rule the company has written down, and reports which ones it clears and which one it depends on.
Configured in Rules & Settings as part of this scenario, not fixed by the product. A floor is a policy, and a tool that picks one on the company’s behalf is making the decision it was brought in to inform.
- $8,700On 29 August, if the receivable clearsThe $2,800 of Acme Retail INV-291 expected to settle on 28 August: one day before payroll. The remaining $6,400 stays disputed.
- $5,900On 29 August, if it does not$2,100 below the floor. Same decision, same day, opposite answer.
Five rules pass. One is a dependency, not a failure.
That distinction is the entire verdict. A system that returned “No” would be wrong, nothing here is unsafe today. One that returned “Yes” would be worse, because the answer stops being true on 29 August if the disputed Acme Retail receivable does not settle. So the product returns the condition itself, names the date it bites, and lists the three things that would remove it: the receivable settling, reducing the purchase to $9,900, or moving it to 1 September. $9,900 is the floor arithmetic and nothing else. $2,100 less, which is exactly the distance the position falls below $8,000. A bare no gets overridden by whoever wanted the equipment. A condition gets planned around.
A risk without a dateis a feeling.
Every finance tool can tell you that a decision carries risk. Almost none will tell you which morning you will find out. The projection runs the position forward twice, as it stands, and with the decision applied, and marks the first point at which the second line crosses the floor.
Here that is 29 August 2026, the day the $14,500 payroll leaves. Not "this could be tight"; a date, a figure, and the specific receivable it depends on. That is a thing a finance lead can act on three days early, which is the only window in which acting is cheap.
The chart is deliberately two lines and one threshold, and nothing else. Everything a projection could show that does not change the decision was left out: the version with confidence bands and scenario toggles was clearer to us and unreadable to anyone who had not built it.
Three decisions, andwhat each one cost.
The bet was written before any screen: the unit of this product is the decision, not the balance. Every screen was judged against it.
- 01
Truly Available is the headline, not Total Cash
The largest figure on the overview is $29,600. The bank’s own number is still present, but demoted to one of four inputs sitting beside it.
The trade-offIt contradicts the figure in the user’s banking app, on the first screen, before any trust has been earned. We accepted that the product has to explain itself immediately or be dismissed, which is why the breakdown sits directly beneath the number rather than one click away.
- 02
Every verdict carries its reasoning
The system never returns a bare Safe or Unsafe. Each verdict names the figures it was reached from: available cash, the floor, the commitments, the exposure, in the same view as the answer, not behind a details link.
The trade-offA reason is slower to read and much harder to build. A yes or no is one comparison; an explanation has to be assembled from every input and stay correct as any of them change. We took the cost because an unexplained no gets overridden, and a product that is routinely overridden is not being used.
- 03
Uncertainty is counted, not hidden
Stale feeds and unresolved disputes appear as At Risk alongside genuine receivables. A gap in the product’s own information is treated as a threat to the position rather than as a blank space.
The trade-offIt means telling the user that the product does not fully know the answer, which is the opposite of what a confident interface does. The alternative is worse: quietly treating unknown money as available is how a safe verdict turns into a missed payroll.
Every exposure, with the count of decisions resting on it The source data, classified, and one line the system admits it cannot read
Every verdict traces back toa rule the company configured.
The rules screen is not settings. It is the source of the verdicts, and it says so at the top: every Safe, Conditional and Not Safe answer is calculated from the six policies listed beneath it. No unexplained judgment, and nothing the finance team cannot inspect and change.
That closes the objection the product cannot survive without answering. A tool that tells a finance lead what is safe is either transparent or it is an oracle, and nobody delegates money to an oracle. Making the thresholds editable, and showing who last changed each one, and when. Is what turns a recommendation into a calculation the company already agreed to.
It is also why the operating floor is an input rather than a constant. In this scenario it is $8,000, entered by the company in Rules & Settings: a figure the concept reads and applies, never one it supplies. A product that picks the safety floor on the user’s behalf has quietly taken the one decision it was hired to inform.
What has to existfor the verdict to mean anything.
A verdict is only as good as the two things underneath it: a complete list of what the company already owes, and a record of what it has decided before.
What "committed" and "reserved" actually resolve to ObligationsEvery commitment the business has made, each with a confidence level and a decision relevance: protected, committed, or policy. This is what "committed" and "reserved" actually resolve to, and without it those two figures are assertions.
Kept, so they can be re-checked when the position moves Saved decisionsDecisions are kept rather than answered and forgotten, which is what makes the overview able to say that two of them depend on a receivable that has just gone into dispute. A verdict that cannot be revisited when its assumptions change is a verdict with a shelf life nobody is told about.
Nine screens in total, including the direction the rules screen exists to rule out.
What the productis not allowed to do.
Each of these removed an option a cleaner design would have taken.
- Read-only
The concept never moves money. It reads positions and returns verdicts; the payment still happens somewhere else. Advising and executing inside one tool needs a different class of permission and a different conversation about liability.
- Imperfect inputs
Bank feeds lag, reconciliations run late, and a clearing balance can be hours stale. The design has to stay correct while its own data is not, which is why staleness is a visible category rather than an error state.
- The floor is the company’s
Thresholds are policy. The product asks for them, applies them consistently and shows its working. It does not decide what a safe buffer is on the user’s behalf.
- Not advice
Every verdict is a calculation against the company’s own stated rules, and is worded as one. Nothing in the interface phrases itself as a recommendation about what the business ought to do.
Four things have to be true.None of them are proven yet.
This is a self-initiated concept, not a client engagement, and none of it has been put in front of a finance lead. Rather than dress that up, here is exactly what would have to be true for it to work, and what would have to be validated before any of it went near production.
- A1
That a finance lead will trust a figure which contradicts their banking app. This is the hardest assumption here, and the one the entire product depends on.
- A2
That commitments, reserves and exposure can be classified reliably enough from accounting and banking data to be worth automating. Every misclassification moves the headline number.
- A3
That an explained verdict actually changes behaviour, rather than being read once and then ignored the way a warning banner is.
- A4
That a business has, or can be persuaded to state, an operating floor. Without one there is no threshold and no verdict, only a prettier breakdown.
A1 is the one most likely to kill it. It is listed first for that reason.
Built as a concept,figures and all.
Nothing here is measured, and every amount on these screens is illustrative. This states what exists and what would be tested, not what was achieved.
- Problem framingConcept
- The user, and the decision they are actually making
- Four financial states, and why the balance is not one of them
- The design bet, written before any screen
- Four assumptions, including the one that would kill it
- Product designConcept
- Nine desktop screens at 1440
- Overview, cash position, obligations and transactions
- Decision evaluation with a rule-by-rule explained verdict
- Cash timeline marking the first point of financial pressure
- Risk register and editable rules with change history
- What we would testHypothesis
- Whether a finance lead accepts a figure lower than their bank’s, against a control shown Total Cash
- Decisions reversed after reading the reasoning, against decisions made on the balance alone
- Whether saved scenarios are ever reopened, or written once and forgotten
The claim is that a decision made against $29,600 is a better decision than one made against $84,500. That is the hypothesis. It has not been tested.