Hunt the muda, not the novelty

There is a quiet failure mode in most AI initiatives: the tool arrives first, and the search for a problem begins second. The result is a pilot that demos well and changes nothing.

Lean offers a sharper starting question. Not “where could we use AI?” but “where is the waste?” Map the value stream, find the muda, and the right intervention — sometimes AI, often something simpler — chooses itself.

Muda is the Lean word for waste: effort that consumes resources without creating value a customer would pay for. Naming it matters, because the moment waste has a name it stops being “just how things work” and becomes something you are allowed to remove.

Start at the value stream

Before a single model is wired in, we trace how value actually moves through the operation: where work waits, where it is redone, where judgment is duplicated. The waste is rarely where leadership expects it.

Only then does the tooling enter — embedded directly into the step that was leaking, not bolted onto the org chart.

Let the waste choose the tool

Once the muda is visible, the intervention almost picks itself. A queue caused by missing context wants better information upstream. A step drowning in routine cases wants automation so people reach the exceptions faster. A bottleneck of duplicated judgment wants a single source of truth.

Sometimes none of those is AI, and we say so. Starting from the waste rather than the technology means we are never committed to a tool before we understand the problem — and that is what keeps a program honest.