People I train often start with the same question: what should I use AI for? It is a reasonable question, but it is the wrong starting point. The better question is how to work with it, and my answer is short. Take the director's chair.

Why a director

A director does not do every job on a production, and does not ask the crew for permission. The director holds the vision, explains what is needed and judges the result. The crew brings the skill. The director brings the standard.

I have worked this way for most of my career. At my own animation studio I direct every production, from story to final edit. In Japan I directed curriculum and teacher training for a group of language schools. In my current role I am responsible for every piece of content the company publishes. When I started building with AI, I found it was the same job.

The change of role

Asking AI a question gets you an answer. Directing it gets you finished work, because you decide what finished looks like. The tool has not changed. Your role has.

Once you see yourself as the director, the range of work opens up. AI stops being a calculator or a search box and becomes a crew that can plan a campaign, draft a proposal, storyboard a video, organise a schedule or pressure-test an argument. The limit is usually the brief, not the tool.

What directing looks like in practice

  • Set the vision. Say what the result is, who it is for and what good looks like before anything is made.
  • Brief the crew. Give it the facts, examples and limits. AI knows a great deal about the world and nothing about your company unless you tell it.
  • Widen the brief. Go past the obvious uses. The tool that summarises a document can also build the plan, the tracker and the presentation that follow from it.
  • Call for another take. If the first draft misses, say exactly what to change. A director does not accept the first take by default.
  • Approve the final cut. Nothing goes out until you have checked it. You are accountable for the result, not the tool.

When the answer is "that can't be done"

AI will sometimes say a task is not possible. Often that is a limit of the first approach, not of the task. A good director hears "we can't do that" and asks for another way: break the job into steps, supply the missing information, try a different tool or change the format of the output. In my experience, there is usually a route.

The rule also works the other way. When the limit is real, such as a fact the tool cannot verify or data it should not see, a good director listens. Pushing past a real limit is how you end up with confident, invented answers. Directing well means knowing the difference.

Why it matters for teams

Most AI rollouts teach people which buttons to press. Few teach them to lead the work. The systems I build follow the director model: people set the direction and approve the result, and AI does the heavy lifting in between. Because people stay in charge of the outcome, they keep using the system after the pilot.

If you take one thing from this article, make it the change of role. You are not asking a tool for help. You are the director, and AI is your crew.

Start with one task this week. Write a proper brief, call for one more take and see what changes. I would love to hear how it goes.

Download the two-page guide: Take the director's chair (PDF), with the five habits, a brief template and a checklist to share with your team. For the mindset behind it, see Guide 2: Think like the director (PDF).

This approach grew out of my work on the ELIZA effect. The two-minute video is on my home page, and the article is here.