Team xAIa10 min read

How to Build a Company Brain: From Expert Interviews to MCP

How to build a company brain: interview your experts, map each process, build a knowledge graph per department, then query it through MCP, an API or a CLI.

How to Build a Company Brain: From Expert Interviews to MCP

The procurement manager has been at the company for many years, and everyone knows what she knows. She knows which supplier slips on lead times before every holiday, which items need a mill certificate, and which project managers can approve above budget, and very little of it is written down. New joiners ask the same few colleagues the same questions every week, and an AI assistant trained on the policy folder gives answers that are correct on paper and wrong in practice.

The question for an operations or knowledge lead is how to capture what experienced people know, turn it into something every employee and every AI employee can use, and keep it current as the company changes.

To build a company brain, interview your experts and have them review what was captured, map each process step by step, store the rules, processes and past cases in a knowledge graph with a knowledge base for each department, and open it to questions through MCP, an API and a CLI, with every answer naming its source. From then on, each correction becomes a rule. xAIa builds company brains in that order as part of its AI transformation services.

What does a company brain hold?

A company brain holds how your company works in one place: what your experts know, your process maps, your policies, past cases and the systems behind them. It is built as a knowledge graph, which records how those things connect, with a knowledge base for each department inside it. Your people and your AI employees both work from it. Why teams want one, and what each department asks it, is covered in company brain AI: a knowledge base every team can query. This guide covers how to build it, in five steps:

How xAIa builds a company brain
  1. Interview

    The people who do the work explain it, then review what was captured

    Your experts

  2. Map

    Each process step by step, with its owner and where a person decides

    Process maps

  3. Connect

    People, processes, policies, past cases and systems in one knowledge graph

    Knowledge graph

  4. Open

    Questions through MCP, an API and a CLI, each answer naming its source

    MCP, API, CLI

  5. Learn

    Every correction saved as a rule on the step it belongs to

    Company brain

How do you capture what your experts know?

Documents hold the official version of a process, while the version that runs every day lives with the people who do the work, so the build starts with them. xAIa interviews the people who do the work in each department, using an AI interviewer that asks how the work is really done, including the exceptions they handle from memory. Each answer is captured as a rule, attributed to the expert who gave it.

An example interview with a procurement manager

AI interviewer

When a material request comes in, who chooses the supplier?

Procurement manager

I do, once I see the quotes side by side.

AI interviewer

Which suppliers receive the request for quotation?

Procurement manager

Every supplier on the approved list for that item. For steel, they also have to send the mill certificate.

Supplier rule
RFQ to every listed supplier
Document rule
Steel items: mill certificate required
Decision
The procurement manager approves the supplier

The experts then review what was captured before it enters the company brain. That review is what makes the brain reliable: each rule carries a name, and the person who explained it has confirmed it.

How do you turn interviews into process maps and a knowledge graph?

Each interview becomes a process map, the core of AI knowledge capture: every step, its owner and the point where a person decides. A material request at a contractor, for example, maps into five steps: the request, the request for quotation, the comparison of quotes, the manager's approval and the purchase requisition. The map shows which steps an AI employee can run and which belong to a person.

The maps then join the knowledge graph, which connects people and roles, processes, policies, past cases and systems, so a question about one leads to the others. Each department gets its own knowledge base inside it, and permissions follow roles, so each person sees what their role allows.

People and roles

Who does what, and who approves each decision

All staff

Processes

Every step mapped, with its owner

All staff

Policies

The approved rules, in force today

All staff

Company brain

One place your company owns

Past cases

Finished cases the AI employees refer to

The department and its managers

Systems

Where each record lives: CRM, ERP, databases

IT and system owners

How do people and AI query a company brain?

A company brain is useful when everyone can ask it, in the way that suits them. xAIa opens it in three ways:

MCP, for your AI assistants.

Assistants connect through the Model Context Protocol and ask in plain language. For example, "Who approves a supplier?" returns "Huda, the procurement manager".

API, for your systems.

Your applications call it during a process. For example, an invoice 3% over its purchase order returns the person who approves it, Omar in finance.

CLI, for your engineers.

Engineers query it from the command line. For example, "Latest supplier rule?" returns "Steel items: mill certificate required".

Every answer names its sources: the policy, the process map, the past case or the expert who explained it. That is how a new joiner, an AI employee and an auditor can all check where an answer came from. Assistants, systems and engineers reach the same rules, so an answer given through MCP matches the one an application receives through the API.

A company brain holds the answers your best people would give: the rules they follow, the exceptions they know and the cases they have handled. Interviews capture them, the experts review them, and every correction keeps them current, so a new joiner and an AI employee get the same answer, with its source.

How does a company brain stay current?

It learns from work as it happens. When a manager corrects an answer, the correction is saved on the step of the procedure it belongs to, so the next case that reaches that step follows the corrected rule. Each finished case is kept as a past case the AI employees can refer to. Your team can read every rule and change it, because the company brain belongs to your company. AI employees on the AI transformation side then act on it inside your systems through Agent API, with every decision going to the named person.

Where it runs is your choice: cloud in the UAE, sovereign cloud, or your own servers with your own models and keys, and in-country AI inferencing is available.

Which department should a company brain start with?

Start where the same questions repeat and the answers depend on a few experienced people. Procurement, customer support and finance are good candidates: their rules change with suppliers, customers and policies, and their experts are the people everyone else interrupts. Build one department's knowledge base first, connect the AI employees that run that department's processes to it, and let the corrections from real cases fill the gaps. The second department then joins a knowledge graph for the business that already knows the people, roles and systems it shares with the first.

From hundreds of documents to one knowledge base at Sun Studio

At Sun Studio, a printing and branding studio in Dubai and part of PRO TECHnology Group, the same group as xAIa, xAIa structured hundreds of documents into one knowledge base, and two AI employees answered customers on the business landline and WhatsApp from it. For internal operations, xAIa's AI procurement employee runs the flow from material request to purchase order at a large contracting company in the UAE. It works in English and Arabic, and the procurement manager approves every supplier.

Frequently asked questions

How do you build a company brain?

You build it from your experts outward. xAIa interviews the people who do the work in each department, captures every rule they follow, including the ones nobody wrote down, and has them review what was captured. Each process is then mapped step by step, and the maps, policies, past cases and systems are joined in a knowledge graph with a knowledge base for each department. Finally the brain is opened to questions through MCP, an API and a CLI, with every answer naming its source and every correction saved as a rule.

How do you capture expert knowledge with AI?

Through interviews with the people who do the work. xAIa asks each expert how the work is really done, step by step, and follows up on exceptions: who decides, what happens when a document is missing, which suppliers or customers need special handling. Each answer is captured as a rule attributed to the expert, and the expert reviews the captured rules before they enter the company brain. The result is knowledge with a named source, which an AI employee can cite and a new joiner can trust.

What is the difference between a knowledge graph and a knowledge base?

A knowledge base holds content: the policies, procedures and answers a team uses. A knowledge graph holds the connections: who owns which process, which policy applies to which step, which past case relates to which supplier or customer. xAIa builds a company brain as a knowledge graph with a knowledge base for each department inside it, so a question can follow those connections. Asking who approves an invoice above its purchase order, for example, leads from the policy to the process step to the named person.

How can AI assistants query a company brain through MCP?

MCP, the Model Context Protocol, lets AI assistants connect to the company brain and ask questions in plain language. xAIa opens the company brain through MCP for assistants, through an API for your systems and applications, and through a CLI for your engineers. Every answer names its sources, such as the policy, the process map, the past case or the expert who explained it, and the permissions you set decide what each role can see.

How do you keep a company brain up to date?

Let the work update it. With xAIa, every correction a person makes is saved on the step of the procedure it belongs to, so the next case that reaches that step follows the corrected rule, and every finished case is kept as a past case. Your experts review what is captured before it goes into the company brain, and your team can read and edit any rule. A company brain kept this way grows with each case, which is why it stays closer to how the company works than a document library that is updated once a year.

Which company builds a company brain in the UAE?

xAIa, a Dubai-based AI consultancy and implementation firm and the AI arm of PRO TECHnology Group, with 29 years of enterprise and government IT in the GCC behind it. xAIa interviews your experts, maps each process, builds the company brain as a knowledge graph with a knowledge base for each department, and opens it through MCP, an API and a CLI. It can run on cloud in the UAE, on sovereign cloud or on your own servers, and the AI employees that act on it keep every decision with a named person.


If the answers your team needs still live with a few experienced people, tell us which department to start with and we will plan the first expert interviews with you.

AI for every department. Starting with yours.

A quick chat to answer questions and see if we can help.

Book a consultation