One decision, prepared before you open it · Illustration, fictional customer
Waiting for your decision
Asked by your sales lead this morning · prepared by the AI
A long-standing customer asks to pay their next large order on longer terms than usual. Do you agree?
What the decision depends on
They have paid every invoice of the past two years, a few of them late.
Your accounts
This order is larger than any they have placed before.
Your quotes
Your written terms allow longer payment only with your approval.
Your rules
The payment would land in the month your cash forecast shows as the tightest of the quarter.
Cash forecast
Your options, and what each one changes
A
Agree to the longer termsThe cash for this order arrives later, in your tightest month.
B
Agree, with a deposit before the work startsThey accepted a deposit on their last large order, according to the email history.Email history
C
Keep your usual termsTheir buyer wrote that they may then split the order into smaller ones.Email history
Not checked: a credit check on the customer. It is not in the data the AI can read.
Your decision
ABCAsk for more
Your choice and your reason are kept for the next brief
Our illustration: a fictional customer, and no real figures.
Some decisions can only be made by you: a customer asking for longer terms, a quote below the margin you usually accept, a purchase above the limit you set. Each one takes a few minutes to decide and much longer to prepare, because what it depends on sits in the accounts, the inbox and a spreadsheet, and someone has to gather it first. Until somebody does, the request waits, and so does the person who asked. This page explains what a decision prepared by AI contains, which decisions to prepare first, what is set up once and what stays with you.
Why decisions wait
Most of the decisions that reach an owner take time for a simple reason: the facts behind them are spread across systems that do not talk to each other. The payment history is in the accounts, the customer’s last messages are in someone’s inbox, the effect on cash is in a forecast spreadsheet, and the rules the business has agreed are often only in the owner’s head. Gathering them falls to the owner, or to the person who asked, and while it is not done the request waits on the owner’s desk.
This is also where AI most often disappoints an owner who already uses it every day. In a chat window, an assistant knows only what has been pasted into the conversation, so it answers a business question with general advice, or with figures nobody can trace. Set up properly, it reads the business’s own data, places each fact next to the system it came from, and hands over a short brief that the owner can decide from.
What a prepared decision contains
A prepared decision fits on one page and always follows the same order. It states the question in one sentence and who asked it. It sets out the facts the decision depends on, each one with the system it came from, so you can open the source when a fact surprises you. It lists the realistic options with what each one changes for the business, for the customer and for the people involved. And it says plainly what the AI could not check, so that a missing fact is a visible gap in the brief instead of a guess you cannot see.
At this level of independence, which we call copilot, the AI prepares the work and a person carries it out: the choice, the reply to the customer and any change in the accounts stay with you or with the person you name. We explain the other levels, and where a person’s approval sits in each, in handing a repetitive task to AI.
Which decisions to prepare first
The decisions worth preparing first are the ones that come back often and always depend on the same kinds of facts: a quote below the margin you usually accept, a refund outside your policy, a purchase above the spending limit, overtime or a change to the roster, longer payment terms for a large order, a late invoice to chase or to write off. For each of them, the facts sit in the same places every time, so the AI can be set up once to gather them, and every decision you make gives the next brief something to refer to.
The decisions that reach an owner, by how often they come back
The AI prepares a brief each time
Background only, you weigh it
Every week
A quote below your usual margin
A refund outside your policy
Overtime or a change to the roster
Every month
A purchase above the spending limit
Longer payment terms for a large order
A late invoice to chase or to write off
A few times a year
Hiring a manager
Opening another site
Changing your main supplier
Each mark is one time the decision comes back in a quarter, as an illustration
Rare decisions, such as hiring a manager, opening another site or changing your main supplier, are a different case. The AI can still gather the background: how the figures have moved over the past year, what was said about the question in meetings, the options already discussed. The weighing itself depends on judgement and on relationships the AI does not hold, and it stays with you and the people you trust.
What is set up once
Three things are set up before the first brief. The AI gets read access to the systems the decisions depend on, and only to those, with the same care as any other access in the business: it can read the accounts and the forecast, and it cannot pay, send or change anything, which we explain in adopting AI securely. The business’s rules are written down where the AI reads them at the start of every brief: who decides what, the limits, the standard terms and the margin you accept. And the format of the brief is fixed, so every decision reaches you in the same order and can be read on a phone between meetings.
Set up once, used by every brief
Your rules, written once
Who decides whatBelow the spending limit, your manager. Above it, you.
Payment termsYour standard terms. Longer terms need your approval.
QuotesThe margin you accept, and when a quote below it comes to you.
What the AI can read
Your accounts
Your email
Your forecast
It reads. It cannot pay, send or change anything.
Your past decisions, with your reasons
Longer terms, repeat customerAgreed with a deposit. Reason: their first order of this size.
Refund outside the policyAgreed once. Reason: our delivery was late.
Quote below your marginDeclined. Reason: no further work expected.
The next brief quotes the decision that matchesLast time, longer terms for a repeat customer were agreed with a deposit.
The fourth piece builds up over time: a record of each decision and the reason you gave for it. When a similar request comes back, the next brief quotes what you decided last time and why, so the AI does not treat a question you have already answered as a new one, and your team can see how you decide. This answers a problem owners meet as soon as they push AI on real business questions: it forgets yesterday’s corrections and repeats the same mistakes.
Where that data goes when the AI reads it matters to a business in the UK, Australia or New Zealand alike, and our guide on what you can safely share with AI lists the questions to ask about training, retention and location before anything is connected.
What stays with you
You keep the decision, the conversation with the customer and every exception. The AI never acts on a brief by itself: once you have chosen, the reply is written by you, or prepared for you to read and approve, and nothing is paid or changed in the accounts without the approval set for that action.
Once the rules are written and a kind of decision has been prepared the same way for a while, some of those decisions can move from you to a manager, who reads the same brief and applies the same written rules, with only the exceptions coming back to you. Your team then needs to know how to read and question what the AI prepares, which we cover in AI adoption training for your team.
How we start
We work on this with owners in two ways. We give you personalised guidance on applying AI to your responsibilities and making informed decisions about its use, so the briefs fit the way you decide. And when the figures a leadership team decides from come from several systems, we build them as a custom business tool, such as a dashboard whose definitions everyone has agreed, as we describe in how we help.
The AI AdoptionFree AI pre-audit 
Your accounts
Cash forecast
Email history