The Future of Work
The future of work isn’t one person managing a team of people. It’s one expert building and managing a team of agents.
That holds across nearly every job in the corporate world. In the field I know best, product management, it’s the product manager becoming a product engineer. This product engineer now runs a team of agents rather than coordinating designers, engineers, QA, security, and DevOps teams to turn ideas into products.
That future isn’t coming. It’s already here.
The shift happened in November 2025, when Opus 4.5 landed. It wasn’t one feature. It was the combination: better coding, stronger tool use, longer context, memory, and subagents. Earlier models could help with the work. This one could start carrying the work.
Most people just haven’t done anything real with it. The average employee uses AI to draft emails faster, brainstorm ideas for a presentation, or to replace Google.
They haven’t been exposed to the real power of AI, and they’re judging all of AI by the small slice they’ve experienced.
They prompt. They don’t build systems. Prompting is asking once and taking what you get. A system is the setup that does that work the same way every time, without you there to re-explain it.
That isn’t on them. Working well with these models is a real skill, and building a skill takes time. Companies hand everyone AI and expect magic, without giving anyone the time to learn it or a real tool to learn it with. The tool most employees do get is usually the stripped-down kind, a fraction as capable as something like Claude Code. They meet that watered-down version, decide AI is overhyped, and never explore what it can actually do.
What can it do? It can operate as an entire team. All with you as the manager.
But building a system that actually does the work starts with knowing your domain cold. What gets done, and how it gets done. Know that, and you have enough to start building with AI.
The magic is in the depth. The better you understand how your field really works, the better the system you can build, because you can tell good work from bad.
Take that jump from product manager to product engineer. You keep the product core, discovery and validation, finding the right problem. But now you also need the ins and outs of design and engineering. Enough to design the system itself, and to know what good looks like in the finished product. You can’t build something that produces good work if you can’t recognize good work when you see it.
The AI is just a tool, and a tool is only as good as the person holding it. Knowing your field is what makes it useful. It’s what tells you what to build, and what good looks like once it’s built. Hand that knowledge to AI and one person can do what used to take a whole team.
To build the system, you don’t need to know how to code. You work with the AI in natural language to build out a team of agents. An agent is just an AI with one job, like a person on your team.
You manage that team the same way. Give each one a clear role, the skills to do it, and a workflow that gives it structure but room to think.
With all of this though, the parts that matter most stay human.
In product management, it’s finding the right problem to work on. You still go out, talk to people, watch where they struggle, and work out how to help.
AI really picks up in the spaces after that. AI can synthesize your interview data, build personas and journey maps, and then design, build, and test solutions in a fraction of the time it took before. With the right system and guardrails, it can do it safely and securely.
Below are two examples. One is a writing team that produces in my voice, from my own knowledge. I used this to build out the Product Playbook linked in the navigation. The other is a sneak peek at a toolkit that builds production-ready macOS and iOS apps.
And all of this only costs $200/month.
None of this is out of reach. It’s a skill, and you can start building it now. The people who do won’t be the ones who get replaced. They’ll be the ones who thrive in the future of work.