The 9 dimensions whereAI agents operate

01What a dimension is, and why there are nine

A company gets requests all day. A customer asks about their delivery, a supplier sends an invoice, the director wants to know how much sold yesterday. In an office of AI agents, every request has to reach the agent it belongs to, and that is what sorting it by dimension is for.

A dimension is a question about the company: what it governs, what it contributes, what standard it sets, how it grows, how it treats the customer, what it improves, how it allocates what is scarce, what it delivers, and who it answers to. Our model uses nine because they are designed so that any request falls into one, and only one. That rule lets the coordinating agent, which we call the CEO agent, send every matter to the right agent without guessing. Anthropic describes this pattern as routing: classifying what comes in and sending it to the specialized task it belongs to¹. What does not fit any dimension does not stay floating: it goes to a person.

02The nine dimensions, one by one

The table shows the question each dimension answers and a real request that falls there.

03The mandate matrix: one owner per cell

The dimensions cross with eight agents. Seven carry the mandate of a leadership role: CEO (coordination), COO (operations), CMO (marketing), CFO (finance), CTO (technology), CLO (legal), and CRO (revenue). The eighth, Compliance, does not depend on any area: it watches all of them and reports directly to the CEO agent. The Institute of Internal Auditors explains it in its three lines model: the independence of whoever audits from whoever manages is what gives it objectivity and authority². Above all of them is the general director, who is a person and has the final word.

Each cell says what falls to that agent in that dimension, with a single owner. It is the rule of a RACI matrix, where every task has exactly one person accountable for the final result³. Read by column, the matrix is each agent's job description. Read by row, it says who answers each type of request.

Coordination, operations, marketing, and finance

Technology, legal, revenue, and compliance

Two columns deserve an explanation. The CRO exists apart from the CMO because whoever creates demand should not be the one who closes it: that way each one is measured by its own result. And Compliance has veto over every area because its job is to check that what got executed is what someone authorized.

04How the matrix works on a normal day

The example uses an invented company to show how three requests travel. It does not describe any client.

What the company gains shows up in the example. The night complaint gets handled before the customer gets angry. The invoice does not wait for someone to enter it. The director gets the checked figure in minutes, without asking three areas for it. And every decision that matters, like paying or confirming the crew, reaches a person with the information already put together.

There is a less visible gain. A mid-size company rarely has a director dedicated to each of these eight areas. The matrix gives every area the full mandate of that role, written and applied the same way every day, while decisions stay with whoever leads today.

05How we install the matrix in your company

Omni Agent is an AI consultancy based in Mérida, Yucatán. We install, connect, and maintain agents that run on top of models that already exist, mainly Anthropic's. The matrix is how we organize that work, and it is installed in three phases.

1. Discovery. We interview whoever runs each area and map the business rules, the criteria nobody has written down, who talks to whom, the tools they use, and how many times a month each thing happens. You come out with the map of the cells that apply to your company and the agent that answers in each one.
2. Contract. If you decide to move forward, it is signed once Discovery ends, for twelve months.
3. Implementation. Every cell turns into instructions, permissions, and connections: what the agent can read, what it can prepare, and what needs a person's authorization. Agents run on infrastructure dedicated to your company.
4. Maintenance. With use, each cell's rules get refined and stay in a memory every agent shares. If someone on your team leaves, what the agent learned stays.

Not every company needs all eight columns from the start. The CEO Agent package installs one agent for whoever leads. The Microswarm installs five, one per director. The Omni Agent package installs the full synthetic office, with a swarm per area. The matrix is the same across all three; what changes is how many columns get turned on.