Some of the least transformed companies I know now describe themselves as AI-first. Usually what they mean is simpler than the slogan. They added AI to support, or sales, or coding, or search. The workflow improved. The company did not.

Cars24 began as an engineering company. Then transactions grew, and our attention moved to operations: inspections, logistics, refurbishment, and financing. I let it happen. The company drifted from the reason it had started. Around 2023, it became clear that the old way would not take us forward. We had to rebuild how the company worked while the live business kept running.

A founder I respect told me recently his company was AI-first now. I asked what had changed. He said they had integrated GPT into support. I asked what else. He paused, then said they were still figuring that out. I do not say this to mock him. Most of us built companies for a world in which context was expensive to gather, expensive to move, and expensive to act on. That is the distinction that matters. An AI-first company uses AI. An AI-native company rebuilds around what AI changes.

A company is a machine for carrying context

That sounds semantic until you look at what a company actually is. The economist Ronald Coase asked, in 1937, why firms exist at all, why we do not just buy every task on the open market. His answer, which won him a Nobel Prize, was that using the market has costs, and a firm exists to economize on them by replacing the price mechanism with the directing authority of a manager. Hierarchy, in other words, was never simply ego or bureaucracy. It was a way to coordinate information cheaply enough to act. Jay Galbraith later made the same point in organizational language: structure is fundamentally a way of processing information under uncertainty. And Mel Conway saw the consequence back in 1968: organizations ship their own communication structures. The org chart does not just process the product. It shapes it.

So every scaled company is a context machine. It takes signals from the edge, compresses them, routes them, interprets them, and turns them into decisions. The org chart is not just a power map. It is a context architecture. It decides who gets to see reality, when they see it, and how much of it survives the trip. As a company grows, no one person can see enough of the system to decide well, so you add layers. A layer summarizes what is happening below, escalates what matters, filters noise, translates decisions back down. It is not elegant, but it kept large organizations coherent. It is also how they became slow.

A layer is now a design choice

Every layer adds latency and loses fidelity. Facts get polished on the way up, decisions get generalized on the way down, and by the time a signal reaches someone with authority to act, it is cleaner, flatter, and less useful than the reality that produced it. That was a tax companies had to pay. AI changes the economics under it, not by replacing judgment, but because a large share of what layers did was context work: summarizing, routing, triaging, drafting first-pass analysis, turning messy operational reality into something another human could absorb. That work can now be done faster, more continuously, and on routine tasks with comparable or better fidelity than the human chain it used to travel through.

Once that is true, the old shape has to be questioned, and most companies are asking the wrong question. It is not whether to use AI. Everyone will. It is what kind of company makes sense once the cost of carrying context drops this sharply. What happens to spans of control, to weekly cadences that exist mainly to move information, to permission chains built around who can see context, to decision velocity when the state of the business is continuously legible instead of periodically narrated. These are design questions, not tooling questions. The AI-first company adds software to an old shape. The AI-native company rethinks the shape.

In the old company, context travels in batches, through meetings, decks, review documents, escalation chains, and people wait for the round trip to finish before they move. A surprising number of recurring rituals turn out to be design debt once shared context is cheap. Status meetings, chase emails, escalation decks, and permission loops were all compensating for the same thing: too few people able to see enough of the system to act. Once context is cheap, meetings should default to decision, not narration.

Our monthly P&L review is the clearest example I have. It used to run three hours, and it began by waiting for a business-finance person to explain how revenue had moved and why the margin shifted. Now everyone walks in having already put those questions to an agent that can read the P&L. The diagnosis is done before the room meets. So the meeting is half an hour, and almost all of it goes to what we do next and who owns it. It has become a stand-up. The hours we used to spend establishing what happened and why have collapsed, and the time moves to the only part that needed humans in a room, deciding and committing. A company that makes in two days the decision it used to make in two weeks does not become the same company with better tooling. It gets more learning cycles per quarter, corrects faster, and compounds faster.

There is a simple test I hold us to. If AI lets a year of work happen in a month, the business must learn and improve faster too. Otherwise the projects were activity, not progress. I will not call Cars24 AI-native because of what we adopted. I will believe it only when we are visibly faster on the things that count: customer experience, unit economics, growth, and how quickly we can do new things.

This is not the end of management

The lazy version of this argument says AI replaces managers. It does not. What AI replaces is the coordination tax around management: the status collection, the reformatted update for the next layer, the meeting held because no one had a shared picture without sitting in a room for an hour. That was never the highest form of management. It was overhead. The real work, judgment, hiring, standards, coaching, conflict, the hard call under uncertainty, is still deeply human. Strip away the coordination theater and what remains is leadership.

It also does not mean flatten everything. Hierarchy does more than move context. It assigns accountability, aligns incentives, and protects quality, safety, and coherence, and AI does not obviously collapse those. The point is narrower and sharper: the part of hierarchy that existed only to relay information is the part now in question.

And none of it works in a low-trust company. If people hoard context, edit bad news for the next layer, or wait for permission because being wrong is punished harder than being slow, AI will not make the company smarter. It will make it faster at producing polished summaries of the wrong story. Install better tools inside the same old permission structure and all you have built is a better-equipped waiting room. The technical shift lowers the cost of carrying context. The cultural question is whether truth is allowed to travel at all.

More shared context means more agency, not more agents

A traditional organization keeps pushing decisions upward because that is where context accumulates. The store issue escalates because headquarters has the picture. That is usually a context problem, not a character problem. AI-native companies can change it by lowering the cost of shared context, so the person closest to the problem can see the same system state that once existed only at the center. The point is not more agents. It is more agency.

This is the change I have felt most directly at Cars24. The thing that mattered was not the tools we added. It was asking which routing work still needed a human layer and which work could now be done directly, with shared context around it. Many companies will fail here. They will buy the tools and keep the old permission structure, summarize meetings that should not exist, make reporting cleaner while keeping authority trapped at the same level, and call it transformation when what they achieved was software modernization.

Start with the shape

The shape of the company is the AI strategy. Everything else is tooling. Start there and the agenda changes. You ask which meetings should disappear, not which copilots to buy. You ask why an update still moves through three people before anyone acts. You ask the hardest question of all: if you were building this company from scratch today, with AI available from day one, would you build the same layers, the same rituals, the same distance between signal and action. If the answer is no, the gap between the company you have and the company you would build is the real work of transformation, and it is slower than people admit. It touches authority, incentives, reporting rhythms, hiring, and status. It takes years, not quarters. Anyone telling you they became AI-native in six months is selling theater.

This is not only my read. McKinsey's 2025 State of AI survey tested 25 factors and found that redesigning workflows had the single biggest effect on whether a company saw bottom-line impact from generative AI. Redesign, not adoption, is what pays. The next decade will not be divided between companies that use AI and companies that do not. Everyone will use AI. It will be divided between companies that kept their old shape and companies that rebuilt for a new one. One group will still be batching context through layers, only with better software. The other will be running on a different nervous system.

That is the difference between AI-first and AI-native. One is adoption. The other is redesign.

Notes and Sources