1. Coase cuts both ways
Masad opens with Ronald Coase. In 1937, Coase asked why firms exist. His answer was transaction costs. Organizing work inside a company is often cheaper than buying it on the market, because the market is expensive to use: finding the right person, briefing them, checking their work, managing the handoff.
Replit's argument is that agents cut the coordination cost inside the firm. Fewer layers and fewer handoffs.
But the same agents cut the cost of using the market.
Giving an outside expert full context used to take weeks. Reviewing their output took more time. Plugging it into your systems took a project. That friction is the reason companies hired a full-time CFO when they needed CFO-level judgment two days a month.
When agents carry the context and the handoffs, a company can buy senior judgment by the day and still get the depth of an insider.
2. What the self-driving company keeps for people
Read Replit's definition closely. People set the goals, bring judgment and taste, and stay accountable. Agents execute with the company's knowledge and systems, inside clear rules on when to act and when to ask.
That human slice is what a fractional CFO, CHRO or COO sells.
Take the monthly close. Agents can ingest bank statements, build the cash flow, draft the management accounts and a first board pack. Someone still has to look at the numbers and say this margin story won't survive the board. Someone has to sign. That person has seen many closes go wrong, and the company rarely needs them full-time.
The self-driving company may turn out to be the best client independent experts ever had.
3. The fear, taken seriously
The obvious objection: if agents run execution, companies need fewer people, and that includes advisors. A good share of fractional work today is execution sold at a senior rate. The reporting pack. The policy rewrite. The benchmark deck. Agents will absorb that work, and fast.
I think that's right. And I think it's healthy.
What stays is the work that depends on having been there. Reading a CEO in the first meeting. Calling a bad hire early. Knowing which forecast is optimistic before the data proves it.
Replit says its own people feel promoted by the change. I expect experts who sell judgment to feel the same. The ones who sell hours of execution are exposed.
4. The expert needs a habitat too
The strongest idea in Replit's piece is what they call the Agent Habitat. Today most agents live in chat windows, personal accounts and one-off scripts, holding whatever access someone gave them. Replit's answer is a governed environment: agents use the company's knowledge, ask for approval at defined points, and improve every time a person corrects them.
Now look at a fractional executive. The method lives in their head, a few templates, and a folder of past client decks. Every new client starts almost from zero.
Large companies will spend heavily to build their habitat. A company of one can't. Yet the expert needs the same things: their expertise captured, their workflows running with approval points where their judgment sits, and a system that learns from each correction.
That is the gap we work on at Bridge AI. I call the result the self-driving expert. One person sets direction and signs off. Agents carry the research, the first drafts, the data pulls and the follow-ups. Each assignment leaves a better workflow behind for the next one.
Replit borrows a line from Satya Nadella that fits experts even better than companies: "separate the supply of intelligence from the authority over it." Models will change every few months. The expert keeps authority over the method, the client relationship and the sign-off.
5. Whose judgment?
Replit ends on the idea that software is now learning to run the world. If so, every company will have to decide whose judgment its agents work under. The largest firms will keep most of it in-house. For the rest, more and more often, it will be an independent expert with a system behind them.
Are you a fractional executive ready to leverage AI for strategic impact? Explore how Bridge AI's platform empowers self-driving experts to deliver high-level judgment and transform client outcomes.
Source: [The Rise of the Self-Driving Company, Replit (Amjad Masad, 26 Oct 2026)
](https://replit.com/blog/rise-of-self-driving-company)
