Every business I talk to is being asked some version of the same question by a board member, an investor, or their own anxious internal voice: where is our AI strategy? It is usually the wrong question, asked before a better one has been answered.

The wrong question

"Where can we use AI?" treats AI like a substance you sprinkle over an org chart until something shines. It leads to pilot projects with no owner, chatbots bolted onto broken support workflows, and a general sense that something innovative is happening while nothing measurable improves. The better question is: where does more intelligence, more judgment, more pattern recognition, more synthesis, actually create leverage in this business, right now?

That question has a real, specific answer. "Where can we use AI" almost never does.

Three places AI earns its place

In practice, I see it earn its keep in three areas. First, decision support, pulling signal out of noise a person would otherwise have to sift through manually, so a human makes a faster, better-informed call. Second, production acceleration, a genuine first draft, a summary, a first pass at a repetitive creative or analytical task, reviewed and finished by a person. Third, institutional memory, making a business's own accumulated knowledge searchable and usable, instead of trapped in someone's head or a folder nobody opens.

AI is a layer, not a strategy.

Where it does not belong yet

None of that works if the underlying process is broken. A confused intake process, automated with AI, produces confused outputs faster. Unclear ownership does not become clear because a model is now involved, and it just adds a new place to point fingers. AI is very good at accelerating whatever is already there, for better or worse. If what is already there is chaos, AI is an amplifier for chaos.

The order of operations

This is why I keep coming back to the same sequence with every business I work with: systemize first, understand and document how work actually moves. Automate what is repeatable and rules-based, so people stop doing work a workflow should be doing. Only then, intelligentize, bring in AI specifically where judgment, synthesis, or pattern-recognition genuinely creates leverage, on top of a foundation that is already solid.

Skip the first two steps and AI does not fix your business. It just gives your existing problems a faster engine.