Starting with the technology almost guarantees a search for places to force it.
The question creates theater
When leaders ask where the company can use AI, teams return with chatbots, copilots, summaries, and demonstrations. The portfolio looks active because the prompt rewarded visible uses of the technology.
Most ideas remain disconnected from a meaningful constraint, operating owner, production workflow, or measure of value. Pilots accumulate without creating capability.
Begin with friction
Ask why quoting takes five days, why customers repeat questions, why employees reconcile records, or why a decision requires three systems and a spreadsheet. Define the consequence and current cost.
Then examine the intervention. The answer may be AI. It may also be search, integration, a database rule, clearer policy, or deletion of unnecessary work.
The presence of an AI use case does not prove the presence of a worthwhile problem.
Test the reason for AI
AI earns its place when probabilistic interpretation, language, classification, or generation creates value that deterministic software cannot deliver responsibly at comparable cost.
The team should also explain acceptable error, evidence, review, recovery, and ownership. A clever capability without an operating model is still a demo.
Choose the problem first
Prioritize constraints with business importance, adequate data, an owner able to change the workflow, and an outcome the organization can observe.
“Where can we use AI?” is the wrong question. Ask what is limiting the business and select the least complicated responsible intervention.
