How I work

Artificial intelligence has been advancing at a tremendous pace over the last few years. That meteoric rise has allowed experts and creators to produce higher quality content in a simpler and faster way. Creating itself has become easier, so the key to success has moved elsewhere: to expertise in a particular domain and to intent, a clear direction for what we want to create. I call the skill of setting that direction well and conveying it to our tools intent engineering.

The idea was never the obstacle

For me, educational solutions never got stuck on the idea. I dreamed about many of them for years. The obstacle was the work: for a single such solution to come about, an enormous number of person hours had to go into it. AI agents today make it possible to create that work in a new way, more cheaply, faster and more simply.

Perhaps most importantly, I can now iterate on my work very quickly. I come up with something, see what it looks like, implement it and try it out in practice, then move on to a new version, and so on until I reach the final product. That is why it is significantly better than it could have been before this era of AI.

A team that isn’t made of people

I also use AI to run my entire business. I have put together a system in which permanent “workers” work alongside me, agents that keep track of what is happening, assign the work and run my business. Alongside them there are workers who get a single task, carry it out and hand back the result, and the permanent agents manage them. I set the direction, review and decide.

A system like this now allows a single creative individual with good intent to create something they could only have dreamed of a few years ago, or that only a team of people or an agency could have created.

Guardrails and rules

For this to work, AI needs clear boundaries. Every claim needs a source, so a figure without a source goes no further, however convincing it sounds. The boundaries are mechanical, not merely agreed: nothing goes out without my decision, not a message, not a post, not an offer. Every mistake becomes a written rule that the next agent must follow, and done means verified, not just written.

Why I talk about this

This is exactly what I teach teachers and companies. AI does not replace expertise. It amplifies it. So first you need to know what you want and how you will check that it is right, and only then let the tool do the work.