Four shifts and one outcome board
Four shifts and one outcome board. Together they answer the questions any Board should ask of an organization: who is accountable, where the work happens, who does it, and how people are grouped. McKinsey finds 88% of companies now use AI somewhere, and only about a third have begun to scale it. This is what scaling it actually requires.
Shift one. Technology leadership governs instead of advises
Today's CTO or CIO earns a seat at the table by translating technology for the business. When the business builds for itself, what gains value is the thing nobody currently owns: the control plane, meaning standards, identity and permissions, visibility into every agent, and where the money and the work go. It is the only seat that becomes more centralized, which is what makes it safe for everything else to decentralize.
Shift two. Engineering dissolves into the business
There is no central engineering organization. The people who build sit inside the teams that own outcomes, and those teams own the outcome end to end: not a ticket, but "fix accounting" or "make returns same-day." The instinctive objection is shadow IT, and the answer is shift one: it is not shadow if it runs on the governed platform.
Shift three. The builder profile changes
The scarce person is no longer the senior coder. It is the person who knows the domain, can direct agents to build, can judge the result from the user's seat, and knows when to pull in a deep engineer. Many already work for you and have never written a line of code; quick training gets them most of the way, and the rest is permission.
Shift four. The org chart follows the domain architecture
Every company already operates on a set of things (products, customers, orders, suppliers, money), and a conventional org chart cuts each across five functions, so the customer has five partial owners and no real one. The accelerated organization gives each thing one owner. In practice this is alignment, not overhaul: the top of the chart barely moves, and the org chart is redrawn last, after the working example shows what needs to move.
Three things that are already true
Speed is no longer the constraint: the thing that used to take a quarter to get built can now be built by almost anyone, in days. Your organization is the constraint: every company I have worked with was designed to ration the capacity to build, and that supply is no longer limited. And you are already changing, badly. Two-thirds of office professionals at companies over $500MM in revenue say they have used AI tools at work that they believed were not permitted, and 81% believe leadership plays by a different set of rules. Departments are buying and building their own agents with their own credentials, and nobody can say what they are doing, what they cost, or what data they reach. The choice in front of you is not whether to change, because that has been decided for you, but whether to govern the change.
How it runs on a Tuesday
The unit of the company is an ownership team: small, single digits, owning a domain of outcomes end to end: the thing, the data behind it, the interface the rest of the company uses, and the results. It has its own spend envelope (people, tokens and cloud together) and no roadmap. It has a queue. Ownership is measured on outcomes, always with a balance metric that protects the rest of the company: fixing accounting cannot add a day to the monthly close, and cutting support handling time cannot reduce customer satisfaction.
The connective thread between the teams is the outcome board. It looks like a Kanban board, but each card is one outcome, big enough to matter to the executive team and small enough for one team to ship. "Fix accounting." "Same-day returns." Each card carries its owner, the pieces of the company it uses or creates, its spend, its metrics, and its evidence of being done, which means live and in use. When finance's "fix accounting" needs better data from the supply chain, that becomes a card of its own with the dependency visible, so nobody runs a program office to know who is waiting on whom. The outcome board is the strategy review, the status report and the budget conversation, and it replaces all three.
Every day the team answers one question: what did you build yesterday? Show it, don't describe it. There are no sprints, no roadmap reviews and no steering committees to hide behind. One rule makes it compound: a team's work is not done until the rest of the company, people and agents alike, can build on it. Every shared thing lands in the catalog, the company's own list of everything it has built, who owns it and who uses it, and the catalog, not the org chart, becomes the truer picture of how the company works.
What makes it safe
Decentralizing building across a whole company would be chaos without the one part that becomes more centralized: the control plane. It owns no business outcomes. It owns five things: standards for how a piece of the company is built and exposed; one permission system, inherited from what IT and HR already run, applied to people and agents alike; a log of what every agent did, why, and on whose authority; spend per outcome, per team and per agent, visible next to salaries; and a check that nothing starts which already exists.
Governance lives in the platform, not in a committee. The old model governed by gate (the architecture review, the change advisory board, the steering committee) and was slow because a human had to approve before anything moved. This model governs by default: permissions are inherited, actions are logged, spend is metered, everything built is registered, and review happens on evidence after the fact. It also inverts the security problem. On the platform, identity, logging and spend are the ground everything stands on, so the risk becomes anything that is not on the platform, and for the first time that risk is visible. Agent-led systems live on the same outcome board as human teams, with the same owner and the same metrics. There is no separate world for the machines.
What the CEO does now
Most of what fills the executive calendar disappears. There is no annual technology budget to negotiate, because spend lives in domain envelopes and is visible daily. There is no prioritization committee, because the outcome board is the prioritization. There are no status decks, because the status is the evidence of done. What replaces it is more strategic and not less work: the CEO and the executive team decide what goes on the outcome board and in what order, read the catalog to understand how the company actually works, and govern often and with purpose, as the body that makes the hard calls against the vision and the guardrails. In every transformation I have run, that is the part the executive team ends up valuing most.