AI Is a Tool. Start With the Business Problem.

There is enormous pressure right now to “use AI.” It shows up in conference talks, in vendor pitches, and in the quiet worry of a business owner who suspects everyone else has figured out something they have not.

Here is the thing I keep saying to owners, and I will say it plainly: the goal isn’t to use more AI. The goal is to run a better business. AI is a tool — an unusually capable one — and like every tool it is only valuable when it is pointed at a problem worth solving.

A small example that made the point for me

I needed a simple graphic — the kind of thing I have made in Canva a hundred times. Five minutes, maybe less. Templates I know, a layout I have used before, done.

But because the tool of the moment was AI, I decided to do it the modern way. I wrote a prompt. The result was close but wrong. I refined the prompt. Now the type was off. I tried again, got something interesting but unusable, adjusted, regenerated, and eventually exported the image so I could fix it — in Canva.

Forty minutes for a five-minute task, and the output was no better. I had not improved a process. I had replaced a quick, familiar workflow with a longer, more frustrating one, and told myself it was progress because the word “AI” was in it.

That is a trivial example with a serious lesson: adopting AI does not automatically improve a process. Sometimes it improves it dramatically. Sometimes it is a slower, more expensive way to reach the same place. The only way to know is to be honest about what you were trying to accomplish before you picked the tool.

Three questions before you adopt anything

Whenever an owner asks me whether they should be using AI for something, I ask these three — in this order:

• What problem am I trying to solve? Stated as a business outcome, not a technology wish. “We lose leads because nobody follows up for two days” is a problem. “We should use AI” is not.
• How are we solving it today? There is almost always a current process, even if it is informal, undocumented, or living in one person’s head. Write it down. Note what it costs in time, money and errors.
• What specifically will AI improve? Faster? Cheaper? More consistent? Higher quality? Possible at a volume you cannot staff for? Name the improvement, and name how you will measure it.

If you cannot answer the third question in a sentence, you do not yet have a project — you have an impulse. That is fine. Impulses are how exploration starts. Just do not fund one as if it were a plan.

Decide what “better” means before you start

The step people skip is the measurement. Not a dashboard — just a plain before-and-after. Hours spent per week. Days to respond. Percentage of customers who get followed up with. Reviews requested per month. Cost per completed job.

Pick the one number the problem actually shows up in, write down where it is today, and then evaluate the tool against it in thirty or sixty days. If the number did not move, the tool did not help — regardless of how impressive the demo was.

Then, and only then, pick the tool

Once the problem and the measure are clear, the tool choice gets much easier — and it is frequently not AI. It might be a checklist. A calendar reminder. One clear owner for a task nobody owns. A phone call you are not making. A template. Hiring the right person. Removing a step entirely.

And sometimes it obviously is AI: drafting responses at a volume you could never staff, summarizing what customers keep saying, first drafts of routine writing, making sense of information you already have but never read. Those are real wins. They are just wins that come from matching the tool to the job, not from adopting the tool and hunting for a job.

AI is also genuinely good at helping you see the problem

One fair counterpoint: this is not an argument for using AI late. AI is often at its best before the solution stage — helping identify and analyze the problem in the first place.

Paste two years of reviews in and ask what customers complain about most. Describe your sales process and ask where prospects most likely drop out. Ask what questions a customer in your category would want answered before buying, then check whether your own website answers them. Ask it to argue against your plan. That kind of use produces clarity, and clarity is what makes the tool choice obvious later.

Strategy before tactics, again

This is the same discipline in a new costume. AI is a tactic. Tactics belong downstream of a decision about what you are trying to achieve and how you will know whether it worked — the argument I make at more length in Putting Strategy Before Tactics.

Businesses that will get real value from AI over the next few years will not be the ones that adopted the most tools. They will be the ones that were clear about their problems, so they could recognize which tools actually solved them. That clarity is not a technology advantage. It is a management one, and it has always been the harder half.