AI doesn't fix broken workflows. It scales them.

Every few weeks a client asks us some version of the same question. “We want an AI strategy. Where do we start?”

The honest answer is almost never the answer they want.

We start the way we have started every engagement for years. We sit down with the people doing the work and we listen. We watch the business as it actually runs, not as the deck says it runs. We document what we hear. We map it. Intake to outcome. Every step, every handoff, every owner, every exception path. We name where people make decisions and where a rule could replace a person but currently does not. We measure how long handoffs actually take. We do this for the workflows that matter, not all of them.

This is unglamorous, unfashionable and not what the client wanted when they asked about AI strategy.

It is also the AI strategy.

The reason is structural. AI is a very good attacher to a clearly defined pipeline. The model reads an input, transforms it, returns an output and hands it to the next step. The architecture assumes a pipeline exists. If the pipeline is undocumented, ambiguous or improvised owner by owner, the AI deployment is being asked to operate on a workflow nobody has actually agreed on. The result is what bolting any automation onto an undefined workflow produces. It runs faster, the errors compound faster, and the firm now has a more expensive version of the same confusion.

The AI transformation is not actually different from the work we have been doing for years. The technology has changed. The discipline has not. Sit down. Listen. Document. Map. Earn the right to recommend by understanding the business first. Process consulting has been teaching this lesson for forty years. AI has not changed it. AI has only raised the stakes, because AI runs at machine speed and the errors compound at machine speed too.

The data is catching up. BCG's 2025 study of 1,250 companies found that 60 percent were achieving little or no material value from their AI investments, reporting minimal revenue and cost gains despite substantial spending. Gartner's survey of more than 4,200 CIOs, technology executives and business leaders found that only 48 percent of digital initiatives met or exceeded their business outcome targets. Even the management literature now says it plainly: a recent MIT Sloan Management Review piece is titled simply “AI Won't Fix This.” The technology is not the failure mode. The organizational and operational fundamentals underneath it are.

The firms that get real value out of AI are not the ones with the most sophisticated model choice. They are the ones that did the unfashionable workflow work first. They drew the map. They named the owners. They cleaned up the handoffs. By the time they got to the AI question, the model selection was almost mechanical.

When we get the “where do we start” question, we start where we have always started. With the listening. With the workflow.

Not because the AI is unimportant. Because the AI is downstream.