AI in the portfolio

Agents need context

A corridor in a data centre

Portfolio companies are being encouraged to adopt AI everywhere: assistants for every employee, agents for finance and customer service, copilots inside every system. The tools are capable, and many of them still go off the rails in the same way, because they do not know the business they are working in.

Capability without context

An agent asked to chase late invoices needs to know which customers have negotiated terms, which ones are disputing a shipment and who owns each relationship. An assistant asked about a process needs to know how the process really runs, not how a three-year-old procedure document says it runs. Without that context, even a strong model produces answers that sound right and are not, and teams learn to stop trusting it.

The brain as shared context

A Business Brain gives every AI tool the same governed picture of the company. Through MCP, an assistant or agent can query the brain directly; through retrieval, its answers are grounded in cited notes; and through the API, the brain feeds the dashboards and systems where people already work. Access levels apply to every one of those paths, so an agent sees only what the person it serves may see.

Agents that improve the record

The brain is not read-only. When an agent learns something new, from a meeting, a document or a change in a system, it writes back as a proposal with its source, and a person decides whether it becomes part of the record. Over time the agents make the brain better, and the brain makes the agents better, while the people who run the business stay in control of what it says.

A practical order of work

  • Capture the business first, so there is context worth connecting to.
  • Start with the questions leadership asks most often, and measure how well they are answered.
  • Connect the AI tools the company already uses through MCP before buying new ones.
  • Let agents propose changes through review, and expand what they do as trust builds.

The firms that deliver on the promise of AI across a portfolio will be the ones that give their agents a reliable understanding of each business. That understanding is what we build.