Going AI-Native: migrating the work to AI in the right order

By now almost every company has "adopted AI." They bought the chatbot subscriptions, added an AI line to the roadmap, told everyone to use Copilot. And then, mostly, nothing happened. MIT put a number on it: 95% of enterprise GenAI pilots deliver no measurable P&L impact. Not because the models are weak. Because buying a tool was never the same thing as changing how the work gets done.

There's a name for the gap now, the GenAI divide: high adoption, near-zero transformation. Everyone crossed the easy line, hand people a chatbot, and stalled at the hard one, rebuild the work around it. That's the tell. Going AI-native isn't a purchase you make. It's a migration you run, and it goes in an order almost everyone gets backwards.

The order is mindset, then work, then tools, then people. The companies that actually cross don't start by asking which AI to buy. They start with the first-principles question that genuinely hurts: if our customer never opens our software again, do we still create the value? Answer that honestly and everything behind the UI has to move, the processes, the tools, the roles, the org chart.

Bolt AI onto an old process and all you've built is the old process, faster. A faster horse.

This is the difference between AI-enabled and AI-native. Enabled is a feature stuck to the side of the way you already worked. Native is the work redesigned around what the machine can now do, with agents driving the primary workflow instead of assisting it. One is a coat of paint. The other is a move, and most companies flinch from it, because the hard part was never technical. It's human.

AI-enableda coat of paint

AI-nativea move

Where AI sits

Bolted to the side of the old workflowAt the center, running the workflow

Who drives

Your people; AI assists themAgents drive; your people direct them

The process

The one you already had, fasterRedesigned around what the machine can do

If the customer never opens your software

You stop creating the valueThe value still gets made

What you actually built

A faster horseA different animal

The columns aren't two points on a slider. They're two different bets. The left one keeps the org chart and speeds it up. The right one rebuilds the work, which is the part that hurts, and the only part that moves the number.

Make it concrete. Take a support team of twenty answering the same hundred questions all day. The AI-enabled move gives each rep a chatbot in a sidebar: they answer a little faster, you still staff twenty, and the P&L barely notices. The AI-native move redraws the flow. An agent handles the hundred repeat questions end to end, the humans move to the twenty hard exceptions a day the agent escalates, and the team that needed twenty now needs five, doing the work only people can. Same tool, same model. One kept the org chart and sped it up. The other changed what the org chart was for, and only that second move shows up in the P&L. (An illustration, not a case study, but it's the shape of every real crossing.)

I've written in The Small Team what the far side looks like: a few people, heavily leveraged, doing a department's work. This series is the road there from an ordinary company, including the parts nobody sells you, the roles that shrink, the tools you retire, and the ditch most companies stall in halfway across.

Going AI-Native

Mindset FirstSoon

Every migration starts in the head, not the stack. Until 'AI helps my people work faster' becomes 'AI does the work, my people direct it,' nothing you buy will move the needle.

Redesign the WorkSoon

Bolt AI onto an old process and you get the old process, faster. The 95% of pilots that fail skip this step; the few that work rebuild the workflow around what the machine can do.

Retiring the ToolsSoon

The per-seat SaaS stack is becoming something an agent just operates. The UI turns into a backend. What to rip out, what to keep, and how to tell the difference.

Moving the People

The hardest migration is human. Roles shrink, roles grow, and you move a team from doing the work to directing it, without pretending nobody's afraid. → Moving the People

The HandoverSoon

How much to let the system own, how fast, and how you keep a hand on the wheel while you take yours off it.

Stuck HalfwaySoon

Most companies stall mid-crossing: AI on the edges, humans still the bottleneck in the middle. The cost of both worlds, the benefit of neither.

Six pieces, in the order the migration runs. You don't have to start at the top, though: the symptom tells you where you already are. Adoption up but the P&L flat means you never did the mindset. Tools used daily with nothing structural changed means the work was never redesigned. The technology plainly ready but the room gone quiet means you've reached the people, where most crossings stall.

Wherever you enter, it routes back to the same place. No tool you buy will do the work the mindset won't.