Every week, after I finish writing one of these Insights, there’s still quite a bit of work to do before it makes its way to you. The piece needs to go into the website, the formatting needs to be right, the links need to work, and the graphic needs to be added. Then there’s the email, of course, which needs its own graphic, formatting, subject line, preview text, testing, and scheduling.

For the most part, I like writing these pieces, though that weekly deadline comes up pretty quickly. At the same time, after writing 185 of them, publishing each one on the website and scheduling the email had become the bane of my existence.

That was until I realized I could simply have an agent handle it. Once I finish a piece, I can give it to ChatGPT, which follows a publishing workflow I’ve taught it using the “record a skill” function. It prepares the article in Framer, adds the links, retrieves the graphic from the designer in Slack, adds the metadata, publishes it, and checks the page at various screen sizes.

Afterward, I review what’s been published, and once I approve it, I have it prepare, test, and schedule the email. I should reiterate that it does this autonomously within the workflow and approval checkpoints I’ve established. That means I drop the Word doc into Chat, tell it to publish, and it goes to work while I do something else.

As an entrepreneur who runs a small strategy consultancy, I’m sure you can see the appeal. This is work that I would have given to a project manager or an assistant in the past, but with AI, I no longer need to. The division of work is much closer to what I want it to be—and much closer to what it was when I had seven people working for me.

In What Do You Expect from AI Agents?, I wrote about managing agents and evaluating their work. This publishing process has helped me think about a decision that comes before any of that: what do I actually want to delegate?

Where to Start

I think this is the fundamental question in front of leaders right now: when you look at the work you do, what do you want to do more of, and what do you want to do less of?

There are lots of ways to think about using AI to manage work, but this is probably one of the most practical places to start. Even though I work with organizations on AI transformation all the time, it still took me a bit to recognize that this detailed process, with all its clicking through screens and moving information around, could be handled by an agent. But once I did, and once I recorded the skill, the time I got back became time I could spend on client work, writing, or any number of other things.

If you’re a CEO thinking about how agents could fit into your business, I’d start there. Find a recurring process you know well and would like to spend less time doing. Then work through what it would take to hand it off, including where you’d still want to be involved. Don't yet worry about how your team is going to use agents or how you're going to scale agents across your organization. Just begin with yourself.

Here’s a prompt you can use to get started:

Help me identify one recurring workflow I could delegate, partly or entirely, to AI. Ask me questions one at a time about the work I want to do more of, the work I want to do less of, how often it happens, and how much time it takes. Then ask about the steps, tools, and information involved.

Once you understand my situation, recommend one practical place to start. Explain what AI could handle, where I should remain involved, what access or setup would be required, and how to test it on a small scale before relying on it. Don’t assume every step can be automated; flag anything that needs to be verified.

You don’t have to build an entire company of agents to start working this way. You can start with one process you’re ready to hand off. But deciding what to delegate, what good work looks like, and what you’ll do with the time you get back is still your job. That’s the CEO part. And that's the part that I'm fairly certain is never going to change.