Friday, October 23, 2026
Supersonic Tsunami
Action Guide

Your Company Just Handed You an AI Agent. Here's How to Become the One Who Runs It

AI tools are reaching whole workforces at once. The workers who come out ahead will be the ones who use the tools daily, make their work easy to verify, and own the results.

By Bryan· 6 min read
Office worker marking up an AI-generated report with a pen while colleagues plan at a whiteboard behind her

This week Uber expanded an enterprise AI agent from a small pilot to all 30,000 of its employees. That is not a research lab or a startup. It is a large company deciding that an AI agent belongs on every desk, all at once.

If that hasn't happened where you work, it probably will soon. And the background is hard to ignore. Analysts now estimate AI is eliminating more than 10,000 jobs a month, with 9% of jobs at risk. Bill Gates is telling people to prepare for a long, turbulent stretch of disruption. Elon Musk calls it a "supersonic tsunami" for desk jobs. I think his direction is right even if his timing may not be (more on that here).

So the useful question isn't whether to use the agent you're handed. It's what role you claim once everyone has one. This is a practical guide to claiming the right one.

The work is moving up a level

The clearest early signal comes from inside an AI company. The Financial Times reports that Anthropic is hiring relatively fewer junior staff while adding more managers and meetings as agents take on more of the work. Doing the work is getting cheaper. Directing it, checking it and deciding what's worth doing are not.

Managers are already feeling the strain. In one survey, more than half of managers said it has become harder to evaluate their employees' work since adopting AI. Executives aren't seeing miracles yet, either. A KPMG report finds CEOs aren't yet seeing transformational change from AI.

That gap is your opening. When a company can't tell good AI-assisted work from bad, the most valuable person in the room is the one who makes the work trustworthy.

Five steps for the next 90 days

1. Use it on real work, and do it in the open. Don't confine the agent to toy tasks, and don't hide that you use it. Pick two or three recurring parts of your job, such as a weekly report, a first draft or a data pull, and run them through the tool every time. Within a month you'll know where it saves hours and where it quietly gets things wrong. That knowledge is rare right now.

2. Make your work easy to check. If managers struggle to evaluate AI-assisted work, fix that for them. For anything important, note what the AI produced, what you verified and what you changed. It takes a few minutes. It also turns you from "someone who might be replaced by the tool" into "the person whose sign-off I trust."

3. Own an outcome, not a task. Tasks are what get automated. Results stay with the person accountable for them. A recent example shows the pattern: AI uncovered long-hidden flaws in PHP, a widely used programming language, and it took a global human effort to patch the systems that depend on it. The machine found the problem. People had to own the fix. Ask your manager which result you can be responsible for, end to end.

4. Rebuild the ladder you can't count on. If companies hire fewer juniors, the old way of learning judgment, years of grunt work beside senior people, gets thinner. Don't wait for it. Ask for stretch assignments, sit in on decisions above your level, and find a mentor who will tell you why they'd reject a draft, not just that they would. In the companies I've built, the people who grew fastest were the ones who asked to see how decisions got made. Everyone is moving up one rung on the ladder because of AI, and I encourage people to hire more and see who is really AI native and will survive the first 90 days.

5. Don't learn alone. Some of the most encouraging energy is coming from peer groups. Wired profiles a fast-growing community that brings midlife women together to learn and apply frontier AI tools. Find or start that kind of group at work or in your town. Comparing notes with others is the fastest way to get good.

Why this is more than survival

One Fortune analysis argues that AI won't show up in productivity numbers until companies shift from cost-cutting to building. When that shift comes, the people who already know how to direct and verify AI will be the ones handed the new projects. Every earlier wave rewarded the people who learned to run the new machines. The difference now is the pace, and that is why starting this quarter matters more than starting next year.

What to do this week

  • Career: Pick one recurring task and run it through your company's AI tool every time this week. Keep a short log of what it got right, what it got wrong and what you fixed.
  • Career: Ask your manager one question: "Which outcome could I own completely?"
  • Business: If you lead a team, don't just roll out the tool. Define who reviews and signs off on AI-assisted work, and give those people the time to do it well.
  • Civic: Public retraining money exists. The U.S. Labor Department is putting nearly $82 million into rural job training for high-growth industries. Ask your representatives, whatever their party, to expand programs like it, along with apprenticeships.

The agent on your desk can be a rival or a lever. Which one it becomes depends largely on what you do with it in the next few months. Start using it, make your work easy to trust, and take ownership of a result. That is a role no tool can fill.

Sources

CareersHopeful

This post was drafted with AI assistance and reviewed and approved by Bryan.

Keep Reading

All posts →