AI as staff, not software
Two frames for what AI is doing to work. The tool frame makes tools smarter. The staff frame makes roles unnecessary. Those aren't the same product, the same company, or the same industry.
Two frames for what AI is doing to work. Most commentary lives in one of them and doesn’t know the other exists.
The tool frame. AI makes the existing work faster. Copilot writes your functions. Cursor accelerates your IDE. Notion sprouts an “Ask AI” button. Every product grows an AI sidebar. The shape of the work doesn’t change — a human is still doing it. The AI is a better keyboard.
When the tech press says “AI is transforming knowledge work,” what they mean, almost always, is this. The human drives. The AI is the nicer dashboard.
The tool frame is the whole conversation in 2025 and 2026. It’s what the demos sell, what the benchmarks measure, what the pricing models charge for. Per-seat, per-token, per-query. Faster human. Same job.
The staff frame. The human doesn’t drive. The AI is the role. Kit doesn’t make household management easier — Kit is the household manager. Charlie doesn’t make the COO role more productive — Charlie is the COO. Synaxis doesn’t help a marketing team move faster — Synaxis replaces the marketing team.
This is harder to see. Tool-frame language is the only language most people have. “AI that does X” parses. “AI that IS the person who did X” stumbles. The tool frame has been legible for forty years; the staff frame is three years old and most of the words for it haven’t been invented yet.
What changes when the role disappears, not just the friction inside it:
- Measurement flips from interaction quality to outcome. Nobody asks how smooth the chat was. The question is: did the groceries arrive, did the newsletter ship, did the client sign. Tool products are measured by user satisfaction. Staff agents are measured by work completed.
- Trust flips from prompt to latitude. In the tool frame, you micromanage. Every query is a mini-instruction. In the staff frame, you evaluate once and let it run. You don’t prompt your COO every morning. You hired them.
- Compensation flips from per-token to retainer. You don’t pay a household manager by the minute. You pay for the role. Per-seat, per-call pricing is holding onto a tool-shaped revenue model because the product hasn’t made the role leap.
- Hiring flips from evaluation to commitment. Tool AI asks: is this a good fit for my workflow? Staff AI asks: is this a good fit for my org? One is a procurement decision. The other is a hiring decision, with everything that implies.
- Accountability lands on outcomes. Tool AI hands you a deliverable and you decide if it’s right. Staff AI owns the outcome. If the agent runs marketing, the agent owns the pipeline number, not the deck that describes the pipeline.
Tool-frame commentary can’t describe staff-frame products because the words don’t fit. “How productive is the agent?” is a tool-frame question. You don’t ask how productive your head of finance is. You ask whether the books are closed, the taxes filed, the audit came back clean. Outcome-shaped questions, not productivity-shaped.
There’s a phase model underneath this I’ve written about elsewhere. Phase 1 is the Augmentation, which is where the tool frame works. Phase 2 is the Reckoning, which is when tool-frame firms discover that a lot of what they were selling was production their clients can now get from the staff frame directly. If that sounds like a joke at professional services’ expense, it’s only partly. The geometry is in The chain was never a chain.
For anyone building an AI product right now, the question isn’t “how smart does the model need to be.” The question is which frame you’re building in. Tool AI is a crowded category that every major platform is going to absorb — the sidebar, the hotkey, the sibling app, the keyboard shortcut. Staff AI is an open category with a dozen early entrants and no incumbent. The winners in the tool frame will be three platforms. The winners in the staff frame will be whoever figures out what it means to hire an agent the way you’d hire a person, and then does it reliably.
Tool frame makes tools smarter. Staff frame makes roles unnecessary. Those aren’t the same product. Those aren’t the same company. Those aren’t even the same industry.
You can’t do both.
Nobody takes you aside anymore
Print taught a generation when to stop. What we lose when the machines absorb the constraints that used to form us.
Your AI agents need a water cooler
Coordination is a property of the room, not the org chart. What that means when your coworkers are agents.
On the death of the author and the birth of the detector
Why worrying about AI authorship is lazier, and more prejudiced, than it looks.
The work of being available now
A book on AI, judgment, and staying human at work.
The practice of work in progress
Practical essays on how work actually gets done.
Memory is (almost) solved. time is next.
AI can't tell if a memory is two minutes or two weeks old. The fix isn't making models feel time — it's cache invalidation: an as-of stamp on every fact, a clock in the context, and a freshness window for anything volatile.
Did the state change? A simple test for whether work actually happened
Either something exists now that did not exist before, or it does not. A simple test for whether work actually happened, and what changes when you build your systems so they can't record anything else.
How to manage content for multiple clients without flattening their voices
How to manage content for multiple clients without their voices blurring into one house style: a workspace and a voice profile per client, batchable stages, and approval buffers.
The agent-shaped org chart
Every real org has the same topology: principal, role-holder, specialists. Staff AI maps onto it, node for node, and the cost collapse shows up in the deliverables that were always just human-handoff overhead.
The chain was never a chain
On roles, fleets, and the Hegelian reversal waiting at the end of the AI transition. The sequel to Knowledge Work Was Never Work and Apps Are Irrelevant.
In the AI era apps are easier to build. And irrelevant.
I spent months building a meal planning app. This weekend I replaced it with two emails, a spreadsheet, and an AI model — and realized the stage I was racing toward wasn't the destination.