A good AI strategy consultant ranks your first projects on the work itself, not on which tool is newest. Score each candidate on the hours it costs, how often it runs, how clean its inputs are, how cheap a mistake is to catch, and whether one person owns it. Then do the top one first and leave the other two on the list. Ask to see the scoring, because you should be able to check the order yourself.
TL;DR
Because a long list feels like progress and can turn into a pile of pilots. A Harvard Business Review article in November 2025 argued for the opposite of the scattershot approach: "pick one area and go deep." Three is enough to compare. It shows the consultant thought about alternatives, and it gives you a clear second and third when the first one is running.
The list will change. Once the first project is live, you'll learn things about your data and your team that reorder the rest. That's fine. The ranking is a starting order, not a contract.
Some consultants weight these differently, and that's fine as long as they tell you how. The weighting is part of the advice.
Something like this. The tasks and scores below are an illustration, not a client example:
| Candidate task | Hours a week | Runs | Inputs | Mistake cost | Owner | Rank |
|---|---|---|---|---|---|---|
| Drafting replies to common customer questions | 6 | Daily | Clean | Low (read before sending) | Yes | 1 |
| Turning call notes into proposals | 5 | Weekly | Mixed | Medium | Yes | 2 |
| Month-end reporting from three systems | 8 | Monthly | Messy | Medium | No | 3 |
Notice the task with the most hours isn't first. Monthly reporting saves the most time on paper, but it runs twelve times a year, its data lives in three places, and nobody owns it. It's a better third project, once the data is cleaned up, than a first one.
How impressive the demo would be. Which tool the consultant knows best. What a competitor says they're doing. Whether the project sounds like "an AI strategy." None of those predict whether the task will run reliably in your business three months from now. An OpenAI guide for businesses on choosing use cases makes a similar point from the vendor side, recommending "quick wins with strong impact and low effort" as the place to start.
The other trap is ranking by what's easiest to sell. A consultant who also sells a particular platform may rank the projects that platform handles at the top. Ask whether the order would change if that tool didn't exist.
If the answer to the second question is "a guess" for most of the hours, that's normal at the start. It does mean the first job is getting a real before number, so the ranking can be checked against it later.
Expect it to move. Running the first project teaches you things no scoring session could: how clean your data really is, how much checking the output needs, and how the team takes to a new way of working. Re-score the other two after a month with what you've learned. A second project that shares data or people with the first often moves up, because some of the groundwork is already done. One that depends on a process nobody owns may drop, and that's useful to know before you spend on it.
Keep the before and after numbers from the first project next to the scores. They're the best evidence you'll have for whether the ranking method works for your business.
When every candidate fails on inputs or ownership. If the information each task needs is scattered and nobody owns the process, the honest first project isn't an AI build. It's cleaning one list or writing down one process. That can feel like a disappointing answer to pay for, but it's cheaper than automating the mess. There's more on that in when a good AI strategy consultant tells you to skip AI.
At AI Marketing Labs, consulting starts by looking at how your business actually runs, finding the places where AI would pay off, and picking the one worth doing first, working one-on-one with Rick Kranz. If you're still deciding whether you need a strategist at all, settle whether you need a strategy consultant or someone who builds first.
Yes. List your candidate tasks, score them on the same factors, and start with the top one. A consultant earns the fee when the scores are hard to judge, like how clean your data really is or what a build would involve.
How long should ranking the first projects take?For a small business, a session or two once the consultant has seen the work. If it stretches into weeks of discovery with no list at the end, ask what's holding it up.
What if my team disagrees with the ranking?Listen to the person who does the task. They often know about inputs and exceptions that never made it into the scores. Adjust the ranking with them rather than around them.