When electricity first reached industrial production, the results were disappointing.
Factories replaced their steam engines with electric motors and continued operating exactly as before. Same floor plan, same production lines, same division of labor. The power source had changed; nothing else had. The productivity gains were real but modest. Engineers at the time were puzzled. The new technology was clearly superior. Why wasn't the output reflecting it?
The answer came later, when a different category of factory emerged. These factories didn't just swap the power source. They redesigned the entire production system around what electricity made possible: centralized power distribution, flexible floor layouts, specialized equipment that could now be placed anywhere, new roles for workers, new production sequences. The output improvement wasn't marginal. It was transformational.
Most companies are deploying AI the same way the first generation of factories deployed electricity. They're keeping the floor plan.
The Copilot Is the Steam Engine Swap
A copilot for sales. A chatbot for support. A summarizer for legal. A writing assistant for marketing. Each of these tools makes an individual faster. None of them change how the company works. The org chart is the same. The handoffs are the same. The decision-making structure is the same. The meetings are the same. The reporting lines are the same.
You swapped the steam engine for an electric motor. The production line hasn't changed.
"If you talk to people a year ago about how AI was useful, they talked about productivity, like co-pilots, making engineers 20% more productive. I think that is actually a broken way of thinking about AI. That's like taking the old way of working and adding a more powerful engine onto it." — Tom Blomfield, YC General Partner
The insight that moves a company forward isn't "what tasks can we do faster?" It's "what does this make possible that wasn't possible before?"
What Redesigning the Factory Actually Looks Like
When you redesign around electricity, the floor plan changes because the constraints change. You're no longer tied to a central power shaft. Equipment can go anywhere. Work can flow differently.
When you redesign around AI, the operational architecture changes because the constraints change. You no longer need a human in the loop for every decision. You no longer need a full-time employee to handle every recurring process. Work that required a person to read, decide, and act can now run in a loop that closes without human intervention.
What that looks like in a real company:
- A sensor layer that reads the environment: emails, support tickets, product telemetry, CRM data, code changes, cancellations.
- A policy layer that defines what the system can do, what it must ask a human about, what it must log.
- A tool layer that gives the agent the ability to act: query the database, send the email, update the record, open the pull request.
- A quality gate that catches errors and surfaces exceptions for human review.
- A learning mechanism that takes what didn't work and feeds it back into the top of the loop.
When you can run every step of that loop with minimal human intervention, the system improves while you sleep.
Most companies are nowhere near this. They're in phase one: adding AI tools to existing jobs and calling it transformation.
Copilots, assistants. Individuals 20% faster. Org chart unchanged.
Agents close operational loops. Loops run without human intervention.
The company gets better overnight. Humans in supervisory capacity only.
The copilot phase. Each person is faster. The company works the same way. Output scales with headcount. The floor plan hasn't changed.
Individual workflows are rebuilt for agents. Loops close without human intervention. Headcount requirements drop for specific functions. The floor plan is changing.
The system monitors its own performance, identifies failures, and patches them overnight. The company gets better while you sleep. A new floor plan entirely.
Most companies are in phase one and calling it AI transformation.
The Prerequisites for Redesign
Redesigning the floor plan is not just a mindset shift. It has prerequisites.
You need someone who understands how your company actually works at the process level. Not the org chart version. The real version: where decisions happen, where handoffs break, where the tribal knowledge lives that no system has captured. Raphaël Dabadie (YC P26) called this the "real company" problem. You cannot redesign what you do not understand.
Howie Liu, CEO of Airtable, spent a year working through exactly this. In restructuring Airtable's entire organisation around AI, the challenge wasn't the tooling. It was identifying which workflows genuinely required human judgment and redesigning everything else around agents. The floor plan, not the power source, was the hard part.
You need the data infrastructure to route context to the agents doing the work. An agent that can't read your live systems is running blind. The redesign starts with the MCP layer, the data connections, the permission architecture. None of that exists until someone builds it.
You need someone accountable for the live system. The redesign is not a consulting engagement. It's an operations change. Operations changes require an owner who stays in the building after the consultant exits.
These three things don't appear in AI strategy decks. They are the actual work.
The Question to Ask
The question most CEOs are asking: "How do we add AI to our current operations?" This is the steam-engine question. The answer is a copilot rollout and modest gains.
The question that produces 10x: "What would our operations look like if we designed them from scratch, knowing that AI agents can close operational loops without human intervention?"
Most of your current org structure exists to move information between humans who each hold a piece of it. If the information can move without a human intermediary, the structure built to move it becomes optional. Not all of it. But more than you think.
The floor plan is not sacred.
Where to Start
Redesigning the whole factory at once is not realistic. The operator model is function-by-function: pick the one function where the redesign is most clearly justified, build the loop, ship it to production, and learn what the next function needs from that experience.
That first function tells you where the real constraints are. It surfaces the data gaps you didn't know existed. It shows your team what a closed loop actually feels like to operate. And it gives you something concrete to point to when the board asks what AI transformation looks like at your company.
The First Build is 2 weeks, one function. Not a proof of concept, not a prototype. A production loop that closes without human intervention. That's the first section of the new floor plan.
Dan Shipper at Every.to calls the destination after automation: the point at which the interesting design problem is no longer what AI can do, but how humans operate at the supervisory layer above it. Most companies are years from that problem. The floor plan redesign is still ahead of them.
Which function do you redesign first?
The Diagnostic is free. One conversation, 30–45 minutes. We map which of your current operations is most ready for a redesign right now, what the loop would look like, and what it takes to have it running in production within 2 weeks.
Book the Diagnostic →