AI project management is NOT software project management

Most people walk into an AI project running the software playbook. Fixed scope, a clean spec, a Gantt chart, a launch date. Then the model does something nobody predicted, the data turns out to be a mess, and the plan falls apart in week two.

AI delivery is different in ways that matter.

The spec is a moving target

In software, you can pin requirements down and build to them. In AI, behaviour emerges from data and the model — you discover what's possible as you go. The job shifts from "build to spec" to "run experiments and converge."

Data eats the timeline

On most AI projects, the model is the easy part. Getting clean, representative, permissioned data is where 80% of the time goes. If your plan doesn't budget for that, your plan is fiction.

"Done" doesn't exist

A shipped feature stays shipped. A shipped model drifts — inputs change, performance degrades, and someone has to watch it. AI delivery includes the work that happens after launch, not just up to it.

So what do you do?

You manage AI projects as a series of bounded experiments with checkpoints, you put your effort into data and evaluation, and you set expectations with stakeholders accordingly. The full skills picture — and a 90-day path to build them — is in the guide below.