AI strategy workshops: what a good one produces
What should come out of a day spent on AI strategy, who needs to be in the room, and how to tell a useful workshop from an expensive awareness session.
The short answer
An AI strategy workshop is worth a day only if it produces decisions. A good one ends with a ranked shortlist of candidate processes, each with the hours it consumes, its data readiness, a named owner and the measure that will show success; a one-page strategy statement including what you are deliberately not doing; a list of data and policy items to fix first; and a first project with a start date. All of that should be usable by any competent partner, not just the facilitator. Getting there requires the right room: the decision-maker, the people who actually do the repetitive work, whoever owns the main systems and data, and someone who can speak to privacy and legal obligations. It also requires spending the first half of the day on the business rather than the technology, because a list of processes as managers imagine them differs reliably from how the work actually runs, and the difference is exactly where automation succeeds or fails.
The shape of a useful day
| Part | What happens | Output |
|---|---|---|
| Where the hours go | The people who do the work walk through their week; volumes, steps, exceptions | An honest map of repetitive work |
| Where the data lives | Systems, sources of truth, gaps, quality, who owns what | A data readiness picture |
| What the rules are | Which decisions have written criteria, which are judgement, what is regulated | Constraints and no-go areas |
| What is possible | Only now: what current tools do well, with real examples | Shared, realistic expectations |
| Candidates | Match the hours to the capabilities; discard what data or rules do not support | A long list |
| Ranking | Score by hours, readiness, risk, willing owner | A ranked shortlist |
| Decisions | Policy on data and tools, what is off limits, who owns what | Written decisions |
| First project | Chosen, scoped, measured, dated | A start |
Running it well
- Prepare: ask attendees to log where their hours went for a week beforehand. It transforms the first session.
- Start with the business, not with capabilities or vendor demonstrations.
- Get the people who do the work talking and let the managers listen.
- Write the exceptions down; they are where automation projects break.
- Introduce capabilities with real examples relevant to what you heard, not a general overview.
- Rank openly on hours, readiness, risk and willing owner.
- Make the policy decisions in the room: approved tools, data rules, what stays human.
- Leave with a first project: scope, owner, measure, date.
- Circulate the one-page output within two days, while it is still fresh.
What to do in the following week
Fix the smallest data or policy items yourself, because they block everything. Confirm the first project’s owner and measure. Take the baseline measurement, which takes two weeks, so start it now. Circulate the one-page strategy to everyone affected, not just attendees, because the people whose work will change should hear it from you. And put the quarterly review in the calendar, because a strategy without a review date is a document.
What this means for you
A useful AI workshop spends half its time on where your hours and data actually are, includes the people who do the work, and ends with a ranked shortlist, written policy decisions, named owners, measures and a first project with a date, all usable by any partner. Prepare by logging a week of work beforehand, fix the small blockers immediately afterwards, and judge the day by what you can act on, not by how interesting it was.
Frequently asked questions
Who should attend?
The decision-maker, the people who do the highest-volume repetitive work, whoever owns the main systems and data, and someone who can speak to legal and privacy obligations. Six to ten people. A workshop of only managers produces a list of processes as managers imagine them, which is reliably different from how they actually run, and the difference is where automation succeeds or fails.
How do we tell a good workshop from a sales exercise?
By what it produces. A useful one ends with your processes ranked by hours and readiness, your data gaps listed, your policy decisions made, and owners and measures assigned, all of which you could hand to any partner. A sales exercise ends with excitement about capabilities and a proposal. Ask beforehand what the deliverable is and whether it is usable without the facilitator.
Do we need one at all?
Not necessarily. If you already know which processes consume the most hours, have an owner willing to run a first project and know your data situation, skip the workshop and start. A workshop earns its place when the business is unsure where to start, when several departments have competing ideas, or when the leadership team needs to reach a shared view before committing money.