Write the decision brief

Most strategic work starts with an ambition, a statement of direction such as "we should expand services" or "we need to grow into the North." Ambitions point the way, but nobody can approve or reject them, so they tend to drift.

A decision question names three things:

• What will be chosen, from a set of named options
• Who makes the call
• By when it will be made

This lesson uses a sample company: a 60-person commercial cleaning and facilities firm that is thinking about adding a maintenance service.

• Ambition: "We should offer maintenance contracts."
• Decision question: "Do we launch the maintenance service in Q3, pilot it with three clients, or postpone it a year? The CEO decides by 30 June."

A quick test: could someone say yes or no to it, and does it have a date? If not, it is still an ambition. You can follow along with the sample firm or with an anonymized version of a decision of your own.

An alternative is an option you would actually fund and carry out if it won. Options that are only there to make the favourite look good are not real alternatives. Neither are options you already know you would reject.

Always include do nothing or defer. Leaders often treat waiting as free. It rarely is. The cost of inaction is what you give up or risk by not acting, such as lost revenue, a competitor moving first or a client leaving.

For the sample firm, the real alternatives are:

1. Launch the maintenance service in Q3 for all clients
2. Pilot it with three existing clients
3. Postpone it for a year

Suppose the firm turns away about four maintenance enquiries a month, and each contract would earn around $1,200 of margin in its first year. Postponing for 12 months then gives up roughly 4 × 12 × $1,200 = $57,600 of first-year margin. Postponing may still be the right choice, because it also avoids hiring and setup costs. But the cost of waiting is now on the page instead of assumed to be zero. These figures are assumptions. You will learn to check them in the next lesson.

A constraint is a limit that any alternative must respect. If constraints stay in your head, the people preparing options (and any AI tool) will propose plans you could never approve.

State four kinds every time:

• Budget: how much money is available, and over what period
• Deadline: when the decision must be made and when results are needed
• People and capacity: who is available to do the work, and what they would stop doing
• Existing commitments: promises already made to customers, the board or staff

For the sample firm:

• Budget: up to $150,000 for the first year
• Deadline: decision by 30 June so a Q3 launch is possible
• People: two qualified technicians, both currently on cleaning contracts
• Commitments: the board was told there would be no new borrowing this year, and the largest client's contract guarantees current cleaning service levels

The last two often get missed. Moving the technicians could put the client guarantee at risk. That is exactly the kind of conflict a brief should expose early.

The decision owner is the one person who makes the call and answers for it. Name a single person, not a committee or "the leadership team." When ownership is shared, accountability usually goes missing.

Keep three roles separate:

• Preparers gather evidence and draft options. Examples are the finance lead, the operations manager, or an AI tool.
• The decision owner weighs the options and decides.
• Implementers carry out the decision once it is made.

One person can hold more than one role, but the brief should say who holds which.

AI tools sit firmly in the preparer role. They can draft, summarise and lay out tradeoffs. They do not hold decision rights, and they cannot be held accountable. In the sample firm, the operations manager and an AI tool prepare the brief, the CEO decides, and the operations manager leads the pilot if one is chosen.

Naming the owner also tells everyone else whose questions the brief must answer.

A decision brief is a short document, often one or two pages, that sets out everything the owner needs to make the call. It contains eight parts:

1. Decision question: what is being chosen
2. Context: why this is on the table now, in a few sentences
3. Alternatives: the real options, including doing nothing or deferring
4. Constraints: budget, deadline, people and capacity, commitments
5. Decision criteria: the tests you will use to judge the options
6. Owner: the one accountable person
7. Decision date: when the call will be made
8. Open questions: what is still unknown or unverified

Decision criteria are often the weakest part. For the sample firm, they might be: pays back within 18 months, no drop in cleaning service levels, and no new borrowing. Writing criteria down before you compare options makes it harder for an attractive story to win on presentation alone.

Open questions are not a sign of weakness. They tell the owner where the brief is still thin.

A prompt is the instruction you give an AI tool. A useful drafting prompt gives the AI your framing, constraints and the structure you want. It also asks the tool to flag anything it had to assume. For example:

You are helping me prepare a decision brief. Decision question: Do we launch our maintenance service in Q3, pilot it with three existing clients, or postpone it a year? Owner: CEO. Decision date: 30 June. Constraints: up to $150,000 in year one; two technicians currently on cleaning contracts; no new borrowing; service levels guaranteed to our largest client. Criteria: payback within 18 months, no drop in cleaning service. Draft a brief with these sections: decision question, context, alternatives, constraints, decision criteria, owner, decision date, open questions. Do not invent figures. Mark every assumption you make as [ASSUMPTION].

The last two instructions matter most. Without them, AI tools often fill gaps with numbers that sound plausible but have no source.

Practise with a sample or anonymized decision. Before you enter board, acquisition or employee details into any cloud tool, check your company's policy.

The AI's draft is a starting point, not an answer. Read it the way you would read a pack from a new analyst, and look for five things:

• Invented facts: market sizes, growth rates or competitor details you did not supply. Delete them or move them to open questions.
• Alternatives you would not fund: remove them, and restore any real option the AI dropped.
• Missing constraints: AI drafts often soften existing commitments, such as the client service guarantee.
• Ownership drift: phrases like "the team will decide" should name the owner.
• Flagged assumptions: keep every [ASSUMPTION] tag for now. Checking them is the next lesson.

For the sample firm, the draft might claim "strong regional demand for maintenance." Unless you have evidence for that, it becomes an open question: "How many enquiries did we turn away in the last 12 months?"

When you have finished editing, the brief should read as your thinking, with your name as owner.