Essay

AI Chief of Staff. AI Executive Assistant. AI [paste any human job title]. All going to fail.

Giving a human shape to AI is a mistake. Not because AI can't do what humans do, but because any human title you give is a ceiling on something that doesn't have one.

A well-orchestrated AI can consider data points that no single brain can hold at once. Work for more hours no human is capable of. Think from entirely different angles just by adjusting its own context. And act across every corner of an organization simultaneously.

No human does that. No human title captures it.

We give familiar titles to AI to make it easy for people to understand what our products do. Every one of them undersells it.

How many of your friends who bought a new car mentioned its horsepower? Today this unit tells almost nothing. Who has a clue what 200 horsepower practically means? Saying your AI is a "chief of staff" is like explaining a Tesla's capabilities in a horse's terms.

And there's a deeper problem: AI will never hold personal responsibility the way a human title implies. Dressing it in an existing profession doesn't change that.

AI shouldn't mirror humans. It should leave that frame entirely.

The Interaction.

Avoid direct communication with AI as much as possible.

ChatGPT made chatting mainstream because it is the most intuitive way for humans to communicate a task. But chatting is slow and you can barely transfer your brain's context through a chat window.

To take all the juice out of AI, you'd need to feed it far more context. So instead of a human giving every instruction, AI should act on what already exists - user's digital footprint, company's data, tools employees have been using for years.

AI must do the work as a backstage hero. The human stays in the loop only to greenlight or give feedback. The output shows up inside the tools people already use, in formats they understand. Not in another SaaS tool. Not in raw text. Least friction. Flattest learning curve possible, meeting users where they already are.

The best products will blend into users' existing habits.

There is a secondary benefit to it. No shiny "AI tool" replacing you means no enemy for modern-day Luddites. Users still feel like they are behind their own work. That is by design.

Where is all of this going?

Most of an organisation's work can be distilled into small tasks.

Emailing, scheduling, hiring, buying, negotiating, building, researching - a million things that make a company run. AI with the right prompt and context already solves individual tasks well. That's less of a bottleneck now. Orchestrating all of them is.

True impact for organisations will come not with the arrival of a mythical AGI - whatever we imagine it to be - but with the ability to orchestrate hundreds of agents towards a company's goal.

Coordinated groups outperformed individuals. Every great achievement in history was done by groups. As Naval put it: "People who don't organize into tribes get wiped out by people who do" [1].

Similar is true for agents. Early research supports it - coordinated architecture matters more than the intelligence of a single agent [2].

Just as Phil Jackson taking over an NBA team changed its trajectory. Or Ray Dalio starting a fund drawing everyone’s attention. A new generation of superstar company operators will emerge. Not today's CEOs. AI-native generalists. The fastest-learning people, running dozens of agentic operations and getting done what wasn't possible before.

Courage. Taste. Agency. Breadth and depth of mind being the most important qualities to orchestrate it all.

The transition will be gradual for existing enterprises, but existentially important. Startups will be able to act at the scale of today's largest companies and catch up at unprecedented speed.

Fast and Slow.

Chillax. AI is not taking your job. Not yet. The world is going to move fast. But much slower than most of us anticipate.

Outside of San Francisco tech bubble the adoption is lagging. The "AI is about to replace us all" hysteria is clickbait. And while it will certainly get there, it will take far more time than it feels from inside the bubble.

Global unemployment barely changed over the most hyped years in tech history. 4.9% in 2024 and 2025, expected to hold through 2027 [3].

Changes will be incremental. AI replacing humans first on the surface level. Then going deeper, layer by layer, until a company is run by AI agents.

Not overnight. By compounding small wins until we create the last job that needs to be invented - The Company Orchestrator.