The context had to be rebuilt before a meeting
The most useful information about a customer was spread across different places. Some was in company systems; some was held by the people managing the relationship. Before a meeting with a customer’s senior leaders, the Business Partner had to piece the picture together.
It was also a continuity problem: when someone left the team, some of their knowledge could leave with them. Saving documents was not enough to preserve the context built through the customer relationship.
Why three PowerPoint attempts had failed
The first approach was to collect information in presentations that needed manual updates. The project account describes three unsuccessful attempts: there was no owner for updates, no sustainable routine and no maintainable format.
The problem was not writing a good dossier once. It was keeping it useful as the customer relationship changed. Adding another document did not solve the work required to keep it current.
Answer questions instead of writing a dossier
The next approach changed how people contributed. The system asks the Business Partner questions and uses the answers to build a structured knowledge base alongside CRM information and public sources.
The Business Partner remains the source of relationship knowledge. The task changes: they share what they know by answering questions rather than also having to structure and format a presentation.
| Source | Contribution to customer knowledge |
|---|---|
| CRM | Information already recorded in company systems. |
| Public sources | Information available outside the organisation. |
| Business Partner | Knowledge gained through the customer relationship, gathered by asking questions. |
A knowledge base that can be consulted and updated beyond a single meeting
The system described makes it possible to gather, find and update information without treating every meeting as the start of a new dossier. That is what Company Brain means in this Intarget case: giving shared form to knowledge that previously depended mainly on individuals.
The reported result is a change in how information is managed. No adoption, time-saved or commercial-impact figures are published; the case does not use them to claim effectiveness.
What this project teaches anyone building a Company Brain
Before choosing technology comes an organisational question: how does what people know enter the system? If contributing means writing and maintaining a demanding document, the risk is repeating the presentation problem.
The case suggests designing information retrieval and contribution together. Better search is useful, but there must also be a practical way to add what is missing. Update responsibilities still need to be defined; AI does not remove them.
The CRM remains an important source. This kind of knowledge base extends its context with information and contributions that do not fit into customer-record fields alone.
Who led the project
Yempik built this project for Intarget’s Innovation Hub. On the Yempik side it was led by Simone Bova, AI Engineer and co-founder. We share it to explain the project’s design choices, while respecting the client’s confidentiality.
Simone described the method behind this work in “How to put an AI agent into production: not just a demo,” published on Yempik on 2 July 2026.