Running the enterprise from any AI — on a model it can actually read

Why This Exists

Enterprise software is being turned inside out. For thirty years the application was the product: you bought the screens, trained people on the screens, and the work happened in the screens. Now the work is moving into the assistants and agents people already use — Claude, ChatGPT, Copilot, Slack, Teams — and the application is becoming something AI talks to rather than something people click through.

That's what we mean by the headless enterprise: an enterprise where people choose the interface, and the work is done against shared systems underneath by people and AI together.

Every major vendor now agrees on the direction. Almost nobody agrees on what has to sit underneath it. The vendors are each opening their own records to AI — the CRM, the ticket queue, the HR system. That's necessary, and it isn't enough. An agent that can read a ticket still doesn't know whether the work is funded, which strategy it serves, what it costs, which control applies or who has to make the call.

Our view is simple: the headless enterprise needs an operating model AI can read. Strategy, money, portfolio, architecture, controls and the decisions between them, connected and current, with the same permissions and rules as the people using it. Without it, AI is confident guessing at scale.

This playbook has two halves. The first frames the opportunity and the research — where this is going, why most AI programs still fail, what the market is building and what's missing. The second is the playbook: seven plays for building a headless enterprise that is useful, governed and still yours in five years.

There is no single right sequence. Treat the plays as a set you'll revisit, not a waterfall.

A few words we use

• Headless: the work isn't tied to one application's screens. People and agents reach it from whatever interface suits them.
• Operating model: the connected picture of how the enterprise runs — what it's trying to do, how it funds and staffs the work, what it runs on, what rules apply and what was decided.
• Surface: anywhere people or AI meet the work — an assistant, a chat tool, an app, a dashboard.
• Skill: a written-down method for a job that repeats, which an agent or a person can run.
• Decision owner: the person accountable for a call. Agents prepare decisions; decision owners make them.