AI for Strategic Planning: AI Drafts, Leaders Decide

AI for strategic planning is most useful for the slow, mechanical half of the work: turning a written strategy into candidate measures and targets, spotting priorities that nothing measures, and checking that departments are not pulling against each other. It is not useful for the other half: choosing what matters, deciding what to stop, and putting a name next to each number. AI drafts; leaders decide.

That split sounds obvious. In practice, most leaders who use AI every day use it to write emails, and almost none of them have pointed it at their strategy. That is backwards. A better email saves a few minutes. A priority that never became a measure costs a quarter.

What AI for strategic planning is good at

Planning has a lot of work that is careful rather than clever. Someone has to read the strategy, translate each priority into something countable, check every department's draft against the others, and notice what is missing. That is exactly the kind of work current AI tools do quickly and reasonably well, as long as a person checks the result.

Planning task	What AI does well	What you still check
Drafting measures from the strategy	Reads a strategy document and proposes several candidate KPIs per priority	That each measure is something you can actually count, and that someone can move it
Drafting targets	Suggests a target range once you give it a baseline and a time frame	The baseline itself, and whether the ambition is right for this year
Spotting gaps	Lists priorities with no measure, departments with no link to any priority, measures with no owner	Whether the gap is real or a priority you chose to leave unmeasured
Consistency checks	Finds two departments measuring things that conflict, or the same thing defined two ways	Which definition wins
Stress-testing	Drafts scenarios: a key customer leaves, a supplier raises prices, a hire is late	Which risks are plausible enough to plan for
Summarizing inputs	Condenses last year's review notes, department submissions or interview notes	That nothing important was smoothed away

Notice the right-hand column. Every row has a check, and every check is a judgment. AI is doing the drafting so that the leadership team's time goes to the judgments.

Where leaders must decide

There are five decisions in planning that should never be delegated to a model, however good the draft looks.

What matters most this year. AI can list everything your strategy mentions. It cannot tell you which three things the company will be judged on, because that depends on what you know about your market, your people and your board that is not in the document.
What to stop. Every plan has more priorities than capacity. Deciding what will not be done is the hardest planning decision and the one most often skipped. A model will happily keep everything.
How ambitious each target is. AI can calculate what a ten percent improvement on your baseline looks like. Whether ten percent is timid or reckless is a leadership call.
Who owns each measure. One name per KPI. A model can suggest the obvious owner by title; it cannot know who has the capacity, the authority or the appetite.
What the board sees. Which measures go on the one page the board follows, and how they are explained, is a decision about trust and accountability, not about text.

If AI made any of these five, the plan would look complete and nobody would feel responsible for it. That is the failure to design out.

Three rules for using AI with your strategy

Rule 1: AI drafts, a named leader decides

Every output AI produces in planning is a draft until a named person accepts it. Write the name next to it. "The leadership team agreed" is not an owner; "Maria accepted the sales measures on the 14th, and changed two of them" is.

This is the same propose-versus-act line that matters in any AI governance framework. In planning, nothing AI produces acts on its own, which makes it one of the safest places to start using it.

Rule 2: Decide what goes in before anything goes in

Your strategy, your financials and your people's names are not equally sensitive. Before anyone pastes the plan into a tool, decide which tool is approved for internal and confidential information, and keep personal details out unless there is a specific reason to include them. Planning works fine with job titles in place of names. The draft measures do not need salaries.

Rule 3: Check every number against your own data

AI does not know your baseline. If you ask it for a target for on-time delivery without telling it where you are now, it will produce a plausible number, and plausible is the dangerous kind of wrong. Give it your actuals, or ask it only for the shape of a target ("from baseline X to Y by quarter Z") and fill in the numbers yourself.

Worked example: the four questions, with AI in the room

A good plan answers four plain questions, in this order: Where are we now? Where are we going? What could affect us? What must we deliver? Here is how the work divides when AI helps with each. The company is an illustrative 220-person logistics business whose strategy says, in a paragraph, "grow regional accounts and become the most reliable carrier in the region."

Question	What AI drafts	What the leadership team decides
Where are we now?	A summary of last year's results and review notes, and a list of the baselines the plan will need	Which baselines are trustworthy, and which need a month of measuring first
Where are we going?	Plain restatements of the strategy as three to five priorities	Which three priorities the year will be judged on
What could affect us?	Eight risk scenarios, from a large customer leaving to a driver shortage	Which three risks get a measure or a contingency
What must we deliver?	Nine candidate KPIs across the company and four departments, with draft definitions	Which four survive, the targets, and one owner each

In this example the AI's nine candidates include "number of regional accounts", "revenue from regional accounts", "on-time delivery", "delivery complaints", and five more. The leadership team keeps on-time delivery and regional account revenue at company level, drops account count (easy to grow with tiny accounts), and asks the AI for a consistency check across the department drafts. The check finds that sales is measured on new accounts while operations is measured on cost per delivery, and that nothing stops sales from winning accounts operations cannot serve profitably. That conflict gets resolved in the meeting, not in March.

The AI saved the team the drafting. The four decisions in the right-hand column were the actual plan.

When the drafts are ready, put them into a KPI template so every department's measures carry the same fields: definition, baseline, target, owner, and the priority they serve. That shared format is also what makes the AI's consistency check possible, because it can compare like with like.

When to do this

The natural moment is the few weeks of annual planning before your financial year starts, when the strategy is being restated anyway and departments are drafting their goals. Using AI then shortens the slowest part of the process, the back-and-forth between department drafts and the company priorities, so the leadership team spends its meetings on choices rather than on editing.

It also works mid-year. Give AI your current KPIs and your strategy and ask one question: "Which priority has no measure, and which measure serves no priority?" The answer is usually worth a meeting.

If you want to see how the method works end to end, how to create KPIs from the strategy you already have walks through it step by step. And if you would rather do it with help, Measure What Moves is a free working session built on the same rule: AI drafts, you decide.

Mistakes worth avoiding

Asking AI to write the strategy. It will produce something fluent and generic, because it knows what strategies usually say and nothing about yours. Start from your own words, however rough.

Accepting the first draft. The first list of measures is a starting point. The useful work is cutting it from nine to four.

Letting the tool's vocabulary replace yours. If every department's KPIs suddenly sound the same, the AI has smoothed out the differences that matter. Keep your own terms.

Skipping the check because the output looks finished. Polished text is not checked text. The better the draft looks, the more carefully someone should read the numbers in it.

Frequently asked questions

Can AI write our strategic plan?

It can draft parts of it, but it should not decide it. AI is good at turning your strategy into candidate measures, targets and risk scenarios, and at checking department drafts for gaps and conflicts. Choosing priorities, deciding what to stop, setting ambition and naming owners are leadership decisions that need to stay with people.

Is it safe to put our strategy into an AI tool?

It can be, if you decide first which tools are approved for internal and confidential information and keep personal details out. Use a company account on an approved tool, not a personal one, and remember that planning works perfectly well with job titles instead of names. Your AI governance policy should say which tool is approved for this.

How do we stop AI-drafted KPIs from being generic?

Give the tool your own strategy text, your real baselines and your department names, and ask it for several candidates per priority rather than one answer. Then cut hard. Generic KPIs usually come from generic prompts and from accepting the first list.

Where should a CEO start with AI for strategic planning?

Start with a consistency check on the plan you already have. Give an approved AI tool your strategy and your current KPIs, and ask which priorities have no measure and which measures serve no priority. It takes minutes, it changes nothing on its own, and it usually surfaces a gap worth fixing.