
The knowledge the company's people and agents work from
Usually the procedure sits in a document, the exception is known to a few experienced colleagues and the outcome of the process stays in an unconnected system. In the Company Brain we bring together processes, rules and decisions, which people and agents read within precise permissions, so agents produce higher-quality results for a fraction of the token spend.
Company Brain · What it contains
- Memory with evidence
- Facts and decisions with source and validity; what no longer applies stays, marked as superseded
- Ontology
- The company's concepts and how they relate: customers, contracts, guarantees and policies
- Rules and procedures
- Criteria, exceptions and limits that an agent applies and a person reads
- Process map
- How the work is really done, updated at each Mapping
- Context for agents
- The information a task needs, with permissions already applied
What it contains
01
Memory with evidence
Facts and decisions with source and validity; what no longer applies stays, marked as superseded
02
Ontology
The company's concepts and how they relate: customers, contracts, guarantees and policies
03
Rules and procedures
Criteria, exceptions and limits that an agent applies and a person reads
04
Process map
How the work is really done, updated at each Mapping
05
Context for agents
The information a task needs, with permissions already applied
The memory of decisions, with the reasons behind them
ERP systems record that an order was fulfilled and almost never why that customer was given priority. In the Company Brain we keep the reasoning next to the decision, so a choice can be understood and reviewed when conditions change.
A correction that becomes a rule, once it has been checked
The agent opens the claim with the rules in the Company Brain, identifies the policy edition and proposes the settlement, citing its sources. If the claims handler corrects it, the correction enters the shared base only once the rule's owner has validated it.
- The agent identifies the policy edition
- It proposes the settlement, citing its sources
- The claims handler corrects it
- The rule's owner validates it
- The correction enters the shared base
Illustrative example
Frequently asked questions
Is it installed like software?
No. We build it in your context, starting from the processes we work on, with the technologies you have, our components and the development your processes require.
Do we have to organise all our knowledge before we start?
No, it is born with the first project and contains what that process needs. What gets built, though, must be reusable for the processes that follow. It will take some time from the people who know the unwritten rules, and that time pays off for later processes too, which find the collected rules ready if we wrote them to be reused. This is the logic we call Compounding Intelligence.
What does an agent do if the sources contradict each other?
It does not choose for itself: it asks for clarification or passes the case to a person.
Do you use our knowledge to train a model?
No. Agents consult the Company Brain within the permissions that have been set, and we do not use your data to train third-party models.
See how it applies to one of your processes
We start from a procedure that today lives in the heads of the people who carry it out and show you how it becomes a rule that people and agents read the same way.