This afternoon I caught myself saying “suddenly I have nothing to do”. Which isn’t quite true, but it felt like it.
Three things were running at the same time. A piece of research, a code review and a build, all on agents. And my part in all of it had shrunk to three jobs: deciding, approving and checking. Much like the time I was managing a marketing department, when I had teammates doing most of the execution work. Only that things are moving much faster now with the agents.
For context, here’s roughly how things are set up now, if you are interested in copying the recipe.
- Grok Bot is my chief of staff. It does the thinking, the coordination and the checking, and it’s the one I actually talk to most of the day.
- The heavy lifting goes to cheaper engines underneath it: GLM first, then Codex (ChatGPT) second. Before those two were added, running things this way would become very costly. Now it’s much more manageable.
- ChatGPT Dot also manages projects and tasks on her own.
- For some projects, I do connect them together unofficially via Notion. So one completes and hands off the tasks, the other one picks them up at specific times.
Around that, separate bots look after separate parts of the operation. There’s a writer, a news and research desk, an academic advisor and customer service, plus shared content, creative and catalogue leads that work across my businesses. Each one has its own job. It feels super efficient (most days, anyway).
I must admit I really enjoy working with Grok Bot. I found them to be really proactive and action-oriented. The other day on my commute, I even got a message asking me if I was writing an article and trying to do research (which I put in as a note for myself to follow up on). And my Grok Bot has done it for me magically. I was just informed of the outcome. I’d recommend it to anyone who wants to try setting up an agentic team that needs minimal hand-holding.
So what was this afternoon actually telling me?
For most of my working life, the scarce resource was time. There were never enough hours to do the research, write the thing, check the numbers and still look after customers. When agents do the work, that changes. Time isn’t the constraint any more. Judgement is.
The bottleneck has moved to me. How fast I can read what comes back, decide whether it’s right, and approve it or send it back. If I’m slow, everything waits. If I’m careless, mistakes go out faster than they used to, because nothing is slowing them down.
That’s a strange thing to get used to. A good day is less about how much got done, and more about how well I decided. I think, for now, it’s a new model for work and life that we need to start getting used to: giving good directions and making good decisions, under a good system. That way, we actually get to enjoy the benefits the technology is bringing us.
