Human OS for AI

A program manager at a global bank reaches for her trackpad and minimizes a window.

The window is the company’s custom AI platform, twelve million dollars to build. Behind it, her personal ChatGPT. Twenty dollars a month. She does her real work in the second tab. So does most of her team.

She sat through an AI day, completed a 30-day sprint, reported to a program sponsor. The sprint ended. Nothing replaced it. If this doesn’t sound familiar yet, it might soon.

Executives have something to prove in 2026. That AI investments were sound. That returns are coming. $300 billion spent on enterprise AI last year. Three-quarters of initiatives failed to deliver, three years running. Only 12% of CEOs report real returns. The mandate sounds the same everywhere: be more efficient, cut 20 to 30 percent of operating budgets. Implied: go figure out how to do it. None of them have built the human system yet.

Digital transformations produce the same split: the mandated system and the workarounds where people get things done. AI made the gap visible far quicker than any ERP or CRM rollout. AI touches every role, every task. Leaders realize it. But they respond with the same change management playbooks they’ve been running with for twenty years.

Energy surges at launch. AI days, a sprint, a program sponsor tracking adoption. For a few weeks, the momentum is real. Then the sprint ends and nothing sustains it. Nobody designed it to last. The analyst drifts back to the $20 GPT tab. The official platform, built for everyone, fits no one. Usage becomes compliance. Compliance isn’t adoption.

This is where AI programs stall. The energy that launched an AI program had no system underneath it. So the organization does what organizations do with unsolved problems. Whoever’s already performing well gets asked to fix it.

A creative leader gets recruited to modernize a brand. Within months, the new energy she brings to the team stands out. Leadership hands her the AI mandate on top of her day job. She’s not an AI strategist. She doesn’t have the budget to bring one in or the bandwidth to become one. What she has is a title, hundreds of people waiting for direction, and no system underneath her.

She moves anyway. The problems she sees won’t let her sit still. Her reputation is what she does, not what she protects. When AI arrived, she didn’t see a threat to her position. She sees AI as a device to extend her thinking, not replace it. That’s why she moves.

Not everyone does. Two leaders, same company, same ask. One heard “AI will automate 30% of your function” and went quiet. The other started exploring before the presentation ended. The difference was what they’d built their career on. The first spent fifteen years becoming the person who knew the answer. The second rose in the company becoming the person who found it.

Most knowledge workers spent decades accumulating expertise as a fixed asset. AI makes this intelligence available to anyone. The scarce asset is now judgment: strategic, creative, imaginative. The people who keep sharpening human instincts are the ones moving now. They aren’t waiting for the organization to catch up. Their livelihood depends on it.

What’s needed is a Human OS for AI. A set of new practices where teams learn AI through the work itself, growing into a working platform. Something a team can move on by Monday. Something that doesn’t require a hero.

Then comes the fork. One path: write a plan, launch programs, report metrics, move on. There’s corporate speak for this: launch and leave. The sprint ends, the deck gets filed, leadership moves to the next thing. The other path is unplanned. Find who’s already performing and build the system around them.

Most take the first path.

I built a Human OS inside my previous company. Then I left to build Andus Labs around the idea.

In both cases, an enduring lesson emerged: you can’t roadmap something emerging as quickly as AI with fidelity. That turned out to be the advantage. In the lab, we didn’t plan what people would build. We created the conditions for them to build what the work demanded.

People got immersed in what was changing. Live problems, solved in front of them. On day one, they saw a sandbox: an AI policy and three GPTs for text, image, and summarization. They started with small hacks on work they’d done for years. Then they took leaps. Twelve months later: dozens of task-informed agents spanning strategy, creative, red-teaming, and scenario planning, working alongside client teams in a third space between the office and Zoom calls. Nobody planned that arc. The work produced it.

The sessions didn’t end when people went home. They got better. People on the hook were the ones doing the doing. People joined a community, compared notes, helped each other, unencumbered by reporting lines, silos, or geography. It gave them something no roadmap can: confidence. Without anyone naming it, an operating system emerged.

We run studies on how people feel about AI entering their workplace. One dividing line holds: fear and optimism. The cause is consistent. The people who got experience built confidence. Those who didn’t run scared.

One former colleague told me something I haven’t forgotten. She was a program manager, not typically first in line for talent investment. She said: you didn’t announce anything. You didn’t tell us to check out Co-Pilot because AI’s the future. You invested in us.

We didn’t reduce or automate. We expanded and grew together. That only happens through work. And work is the only thing that makes an operating system real.

The argument for delaying investment in people is easy to make. AI gets better every quarter. Maybe the tools get so intuitive that readiness becomes irrelevant. Maybe the people problem solves itself. Nothing in the field supports it. Tools get easier. The judgment required to use them well doesn’t.

Teams stop questioning AI output within weeks. The first override gets debated. The tenth gets waved through. By the fiftieth, nobody remembers what the check was supposed to catch. When AI drifts, and it will, nobody sees the cliff. The people who know why things work and catch what the machine misses are the organizational immune system. Gut it and it doesn’t grow back.

The fork isn’t next year. Maybe you already took it, or maybe it’s next week. Most won’t recognize it as one. It will look like a budget meeting or a quarterly review or a quiet conversation about whether the AI program is working. When that moment arrives, AI either earns its keep or gets filed next to the last three programs that promised “transformation.”

The movers in your organization are already working with AI. With tools you didn’t choose, in workflows you can’t see. Some are burning out. Some are updating their resumes. The window to build around them is open. It’s also closing fast.

The bank’s program manager is still working in two tabs. She was ahead the whole time. Find the people like her in your organization on the move. Don’t hand them a mandate. Build the system underneath and support above them. The operating system starts there. So do the returns.

Chris Perry · Andus Labs · 339 sources · 300+ diagnostics · 30 years in the field