A lot of what gets sold as AI consulting is a model recommendation with a slide deck around it. That’s not strategy, it’s a purchasing decision dressed up as one.

The useful version of this work looks more boring: mapping which parts of a workflow genuinely benefit from automation and which don’t, defining what happens when the model gets it wrong, deciding where a human needs to stay in the loop and why, and being honest when the answer is that AI isn’t the right tool for a given problem yet.

None of that requires hype. It requires treating an AI feature the same way you’d treat any other system component: scoped, tested against failure cases, and justified by the problem it solves rather than by the fact that it’s AI.