Know your runway before it becomes a surprise

Businesses rarely fail because they were unprofitable on paper. They fail because cash ran out on a Tuesday while a large invoice was still thirty days from being paid. This employee exists to make that Tuesday visible six weeks early, and it is read-only from top to bottom.

Steps

Hire it with no write capability whatsoever

In Controls, block every write. This role never moves money, never touches the books, and never issues an invoice. Removing the capability is better than gating it, because there is no legitimate case where you would approve one.

Teach it your collection reality

The whole value is in the weighting. An invoice is not cash, and a forecast built on invoice dates is a fiction you will make decisions against.

Forecast on actual collection behaviour, not invoice terms: - Customers on 30-day terms historically pay at 45 - Enterprise accounts pay at 60 regardless of terms - Anything already 30+ days overdue: weight at 50% - Anything 60+ days overdue: exclude from the forecast - Two accounts always pay on time: [names] State the weighting you applied for each large item.

Pull the figures deterministically

Position, aging, and committed outgoings come from the ledger through a command node. The model interprets; it never transcribes. Every number in the briefing should be traceable to a query rather than to a sentence.

Demand plain language and a single verdict

Write for a business owner, not an accountant. No jargon. Lead with the answer: are we fine, tight, or in trouble in the next 30 days, and what is driving it. Then: current position, expected in (weighted), committed out, and the week where it gets tightest. Name the two things that would most change the picture.

Monday morning, before decisions get made

"trigger": { "type": "schedule", "cron": "0 8 * * 1" }

Verify

Reconcile the opening position against your bank balance to the penny
Check the weighting against what those customers actually did last quarter
Run it against a historical month and compare the forecast to what happened
Confirm it never proposes an action that moves money

When it goes wrong

The forecast is consistently optimistic

It is trusting invoice terms. Your weighting rules are either absent or being ignored; require it to state the weighting applied to each large item so you can see the reasoning.

It misses a large committed outgoing

Annual payments, tax, and anything paid outside your main account are the usual suspects. List recurring large outgoings explicitly in the persona.

The briefing is full of accountancy language

Give it a sample briefing written the way you would say it out loud. Imitation beats instruction.