This is the first concrete playbook for implementing ChatGPT-5 at work.
It’s the third post in my series on a model that ~8% of the planet got on Thursday. The first two covered a broad review and a bumpy rollout.
Why work? Because ChatGPT-5 is different—trickier—than previous LLM rollouts, and I think AI implementation teams need to approach it differently. It has the best potential I’ve seen for solving business problems, but only if we find the right use-cases and the right way to deploy it.
Right now, ChatGPT-5 is “for work,” but no one can explain how in concrete terms. This piece is my early roadmap for leaders making rollout decisions today. My conclusion so far: the model’s differences will force teams to work differently in ways we haven’t had to before—and that isn’t well-documented anywhere.
At launch, Sam Altman opened by calling out work benefits, then spent most of the time on coding, healthcare, and more coding. That’s fine, but if you’re trying to level up AI at work, the real question isn’t whether it writes better Python—it’s how it changes the way we work in the messy, mixed-discipline reality of a company.
We need practical detail. Field notes from implementation. That’s why I write these—to get past the headlines and into the dirty work of what makes this model different, why it’s different, and what that means for teams.
Step back and you see the bigger frame: we’re all on the same accelerating intelligence curve. Each step demands we rethink not just what’s possible, but how we train, structure, and judge output. ChatGPT-5 is another step—capability is up, but access is trickier. The difference between good and bad usage is now strategic impact, not just output polish. Done well, it can beat the best humans at certain tasks (first time I can say that!); done badly, it produces work no intern would submit. That’s a scary range if you lead a team.
What follows are my first field notes—drawn from early tests, executive conversations, and watching teams wrestle with rollout. I’ve captured where it shines, where it fails, and the new failure modes it introduces. I’m also sharing an early implementation compass—10 use-cases I’d put it to work on tomorrow, where it’s already better at specific tasks than a human or any previous model.
ChatGPT-5 is in no danger of taking jobs wholesale, but it is an extraordinary tool that has its place at work. Read on to find out how!
This is an Executive Circle briefing, a Sunday newsletter exclusively for Founding Tier Members. You can learn more via this 60 second video explaining what’s in each tier, and you can change your plan here. Enjoy, and back to regular programming Monday!
