Imagine a Treasury Manager looking at a forecast showing £12 million of expected receipts next month. It's a useful number.
But there's another question that may be just as important:
How much of that £12 million do we actually believe?
Perhaps £5 million relates to invoices already approved by customers with extremely reliable payment histories.
Another £3 million relates to invoices that have been raised but aren't yet approved.
£2 million relates to contractual milestones where delivery is still outstanding.
And the remaining £2 million comes from sales opportunities expected to close shortly.
Adding them together produces £12 million.
But those four components clearly don't mean the same thing. Yet many financial reports present them as if they do.
I think something is missing.
Confidence.
A Number Without Context
Financial systems are very good at representing amounts.
£250,000.
€1.4 million.
$8 million.
They're also very good at representing dates.
Invoice date.
Due date.
Payment date.
Accounting period.
What they rarely represent explicitly is:
How confident are we that this financial event will actually happen and when it says it will happen?
That's an important omission when we're trying to understand the future.
Historical accounting doesn't have this problem.
A bank transaction happened. An invoice was raised. A journal was posted.
These are facts.
But the moment we look forward, we're dealing with expectations and expectations contain uncertainty.
Not All Receivables Are Equal
Consider two £500,000 invoices, both due next Friday.
On a traditional aged receivables report they may look almost identical.
But suppose the first belongs to a customer who has paid every invoice for the past five years and has already approved this one for payment.
The second is being disputed by a customer whose account is already significantly overdue.
Same amount. Same due date. Same accounting status: outstanding.
Completely different information for Treasury, if Treasury is deciding whether it needs short-term funding next week, treating those two receipts identically could matter.
Financial Intelligence needs to preserve that distinction.
Confidence Isn't Probability
There's an important subtlety here.
When I talk about confidence, I don't necessarily mean producing a supposedly precise probability: "There is an 83.7% chance this invoice will be paid."
That degree of numerical precision can imply knowledge that simply doesn't exist.
Confidence might initially be much simpler:
High, Medium or Low.
What matters is that the classification is based on evidence and that the evidence can be explained.
For example:
High confidence
- invoice issued;
- customer approved it;
- no disputes;
- customer has a strong payment history.
Or:
Low confidence
- contractual milestone not yet achieved;
- delivery delayed;
- customer query unresolved.
The label is useful, but the reasons behind it are more useful.
Confidence and Timing Are Not the Same
This distinction is particularly important.
Suppose we have a £1 million customer receipt.
We're extremely confident the customer will pay, but we're uncertain whether it will arrive on the 12th, 15th or 18th.
That's high confidence in the event, but uncertainty about timing.
Now consider a potential £2 million contract.
If it is signed, the first payment will definitely be made 30 days later.
The timing rule is clear, but the contract itself hasn't been signed.
That's uncertainty about the event, even though its timing is relatively well understood.
Those two situations require different decisions.
Collapsing them into a single "forecast probability" loses useful information.
Confidence Changes
Confidence also isn't static. Think about the lifecycle of a customer receipt.
A sales opportunity is created. At that stage, there may be some evidence of future revenue, but confidence is relatively low.
As the opportunity progresses confidence increases.
The contract is signed and confidence increases again.
Delivery is completed, an invoice is raised and the customer approves the invoice.
At each stage, the organisation acquires additional evidence.
Then something unexpected happens.
The customer disputes part of the invoice and confidence may fall.
The dispute is resolved and confidence rises again.
Finally, the money reaches the bank.
At that point we no longer need confidence. We have fact.
This is why I think financial expectations should be thought of as living things rather than static forecast entries.
Their amount, timing and confidence can evolve as evidence changes.
Negative Evidence Matters Too
Financial Intelligence cannot simply accumulate positive evidence, it also needs to recognise contradictory evidence.
Suppose a customer historically pays reliably. That's useful.
But yesterday they requested an extension to their payment terms.
That's more recent evidence.
Or perhaps an invoice is contractually due next week, but Operations reports that the underlying delivery has slipped by a month.
The accounting record may remain unchanged for the moment.
Our financial expectation shouldn't.
A useful system needs to understand both evidence that strengthens an expectation and evidence that weakens it.
Why Treasury Cares
Confidence becomes particularly valuable when decisions have consequences.
Suppose Treasury expects a €10 million receipt next month and has a €9 million payment shortly afterwards.
On paper, the position looks comfortably covered, but what if the expected receipt is low confidence?
Treasury may choose to arrange liquidity earlier.
The same applies to foreign exchange.
An expected Dollar receipt could naturally offset an expected Dollar payment.
But should Treasury rely on that natural hedge if one side is highly uncertain?
The answer depends not just on amounts and dates, but on confidence.
That's why confidence isn't merely an interesting analytical measure. It can affect real financial decisions.
Confidence Should Be Explainable
There is an obvious danger. A system displaying:
Confidence: 87%
without explanation isn't necessarily Financial Intelligence. That’s simply a black box.
For financially significant decisions, users should be able to ask:
Why are you confident?
And receive an understandable answer with supporting evidence.
Perhaps:
- the invoice has been approved;
- this customer usually pays within three days of due date;
- there are no open disputes;
- delivery has been confirmed;
- the last eight comparable invoices settled successfully.
Equally, if confidence falls, Treasury should be able to understand why.
That explainability is essential for trust and also auditability and accountability if AI gets involved.
Where AI Fits
Artificial Intelligence clearly has a role here.
- It can help identify patterns in payment behaviour.
- It can interpret documents and communications.
- It can detect relationships between records.
- It can identify evidence that humans might otherwise overlook.
But AI shouldn't simply produce an unexplained confidence score.
Its more valuable role is helping gather, interpret and assess the evidence from which a financial conclusion is formed.
The decision-maker should still be able to see the reasoning.
Start With Evidence We Already Have
As with the wider Financial Intelligence concept, none of this requires connecting every corporate system immediately.
Start with Accounts Receivable.
- We know the invoice.
- We know the customer.
- We know the due date.
Then look at settlement history.
How late does this customer normally pay?
How variable is that behaviour?
As more information sources are connected, richer evidence becomes available.
Commercial activity. Contracts. Orders. Delivery status. Disputes. Invoice Approval
Confidence becomes progressively more informed.
That's consistent with a broader principle:
Financial Intelligence should become more valuable as more evidence becomes available, without requiring all the evidence before it becomes useful.
Adding a Missing Dimension
Financial reporting has traditionally organised information around two fundamental dimensions:
Value — how much?
Time — when?
For historical financial reporting, that works extremely well, but understanding the future requires another dimension:
Confidence — how much do we believe it?
Once we introduce confidence, a forecast stops being merely a collection of numbers assigned to dates.
It starts becoming a representation of what the organisation currently believes about its financial future — and the evidence supporting those beliefs.
That is a subtle change, but I think it's an important one because Treasury doesn't simply need to know:
How much money are we expecting next month?
It also needs to know:
How much of it can we confidently make decisions around?
And most financial systems today simply don't answer that question.
