MMOOZZGG.com

AI center of excellence

AI center of excellence: who in your company owns the result of AI

Mike Mozg7 min read

An AI center of excellence is the people and rules that answer for where AI works and how much money it brings. A company needs one because AI works across departments, and per BCG about 70% of the difficulties in rolling out AI come from people and processes, not technology. Below are four reasons to set one up, who is in it and how it shares the work with the IT department.

Why AI needs a center of its own

AI works across departments. Sales takes the request, the warehouse checks the stock, accounting issues the invoice. If every department rolls out AI on its own, the company gets scattered subscriptions and numbers that do not match.

In an industry survey of 700 companies in December 2025, 74% of Russian companies had no AI strategy. Without a strategy each department buys solutions one by one, and nobody answers for the overall result.

The center puts these solutions into one plan. It decides which processes to give to AI first, with what, and by which number you see the result. It reports to the head of the company, not to the IT department.

It is not a team of AI agents. People sometimes call it an AI department of people, but people work in it: a chief AI officer, heads of departments, IT, a lawyer. The team of agents that does the work is called an AI department on our site, and we explain the difference separately.

You do not start with a new unit but with one owner and one process. The order by weeks is described on the page about how to set up an AI center of excellence.

Departments work together

The first reason: one process from request to payment instead of six separate services. The sales agent, the support agent and accounting work with one customer and see one history.

Without a center each department picks its own contractor and its own subscription. The company pays for similar tools several times, and a customer who talked to the sales agent yesterday explains everything again to support today.

The center decides which data the departments share and in what order to launch processes so that each next one costs less than the previous. What this means for sales, support, finance and HR is shown on the page about how a chief AI officer works with departments.

It also brings the rules of departments into one place. If sales promise delivery in a day and the warehouse ships in three, the agent must not repeat the promise of sales. Only someone who works with all departments at once sees such a gap.

Fewer risks

The second reason: the risks of AI get one owner. He answers for data, access, 152-FZ (the Russian personal data law) and vendors. Without him each department decides on its own what it may send to a neural network.

A mistake of AI is seen at once, not a month later. The center agrees in advance which questions the agent answers itself, which it hands to a person and who checks its work every week.

Employees already use neural networks on their own, and someone sends contracts and customer lists there. The center writes short rules: what can go to outside services, what only to models inside the company and what to nobody. Every employee gets these rules.

Dependence on a vendor is also on the center. It knows from whom the company buys each solution and keeps a backup option in case the vendor raises the price or cuts off access.

Results you can measure

The third reason: every solution has a metric. What it was, what it became, how much money it saved. The metric is written before launch, not picked afterward.

If it does not work, we switch it off. A solution that gives no number by the deadline is closed, and the money moves to the next process in the plan.

This way AI stops being a cost line without a report. Once a month the head of the company sees one page: the metric of each solution, the savings, the cost and a decision on every row. Which numbers to set and how to report is covered on the page about metrics.

People who work with AI

The fourth reason: people learn to work with AI instead of fearing it. Then AI is used every day, not only in a presentation.

They learn on their own processes. A manager finds out what the agent does, what stays with him and how to accept the agent’s work. A course on neural networks teaches how to write prompts to a chat, and that is not enough when an agent is connected to your CRM and answers your customers.

New hires start work with agents already in place. The rules of the process are written in the system, so a newcomer starts from a ready base, not from questions to a senior manager.

Heads of departments learn to pick a process for AI and to judge the result by a number. For managers we run an intensive from $240 a seat, and the chief AI officer leads the training of departments inside the company.

Who is in the center

1. Chief AI officer: owns the AI roadmap and the result, and reports to the head of the company.

2. Process owners: heads of departments, each with a number. They say what AI must do in their process and accept its work.

3. IT department: connection to systems, accounts and access.

4. Lawyer and security: customer data, 152-FZ, contracts with vendors.

5. Training: programs for employees and managers.

In a small business the center is one owner plus a chief AI officer for a few hours a week. In a large one it is a unit of its own. The duties of each role and who reports to whom are on the page about the structure of the center.

You do not have to hire a chief AI officer full time: we offer the role from $999 a month, 5 hours a week.

The center and the IT department

The IT department makes sure systems work. AI changes how sales, accounting and production work, and that is a decision of the head of the company. So AI is given to the center, not to IT.

The numbers agree. In the MIT NANDA report for 2025, projects with an outside partner reached working use in about 67% of cases, in-house builds in 33%. This is survey data, not an exact measurement, but the difference is twofold.

The IT department stays necessary. It connects AI to the CRM, accounting and telephony, gives each agent a separate account and watches security. The center sets the task and accepts the result.

If the IT department has already started rolling out AI on its own, what it did is not thrown away. The center looks at what works, what is paid for out of habit and what IT lacks to turn a pilot into a working process.

Questions and answers

Does a small business need an AI center of excellence? It does not need a separate unit. It needs one owner of AI and a chief AI officer for a few hours a week: from $999 a month with us.

How this works is covered on the page about the center in a small business.

Sources and dates

Contact us