AI assistants and agents
AI that works on your company's knowledge
An assistant finds information in your documentation and prepares answers with sources. An agent can carry out an agreed action in your CRM, inbox or another system. The scope of access and the situations that need a human decision are settled before launch.
We can start with a single use case, or plan a solution that covers several related processes.

- 01
Answers from documents
The system searches the sources you name, prepares an answer and shows what it based that answer on.
- 02
Actions in systems
The agent fills in data, creates a task or changes a status, within the permissions it has been granted.
- 03
Handover to a person
Incomplete data, exceptions and decisions that carry responsibility go to a named person.
Where an assistant helps
The starting point is a task whose result can be checked. The choice of model and tools comes later.
-
Internal knowledge
Answers based on procedures and contracts
An employee asks a question, receives an answer and goes straight to the relevant passage of the document.
-
Customer service
Drafting a reply to a ticket
The assistant gathers information from your documentation and the contact history. An adviser checks the reply before it goes out.
-
Documents
Reading and organising content
The system extracts the agreed fields, compares documents and flags information that is missing or contradictory.
-
Operations
Completing the next step of a process
Once the rules are met, the agent updates a record, creates a task or sends a notification to the right person.
-
Sales
Preparing information before a call
The assistant collects the account history, what was agreed and the details of the enquiry, so the salesperson does not hunt through several systems.
-
Reporting
Collecting data for a recurring report
The agent pulls information from the named sources, puts it in order and hands the summary over to be checked.
Agreed before launch
The system is given access only to the sources and operations a given process requires.
- 01
Knowledge sources
We name the documents, databases and systems the assistant may use. We also agree how those sources are kept up to date.
- 02
Scope of permissions
We describe the data the agent may read, and the actions it may take on its own or only once approved.
- 03
Rules and exceptions
Typical cases follow a set path. Missing data, a contradiction or low confidence sends the case down the exception route.
- 04
Operational oversight
We log answers, completed operations and errors. The person responsible for the process receives alerts that need a response.
Scope of implementation
The scope depends on the number of sources, the integrations required, the permissions involved and how results are reviewed.
On our side
- a map of the task, its rules and its exceptions
- preparation of the sources and the integrations
- configuration of the agent's answers and actions
- testing on typical cases and on difficult ones
- monitoring, documentation and handover of the system
What we need from you
- a process owner who signs off the rules
- sample questions, documents or tasks
- access to a test environment or a copy of the data
- people who will check the results before launch
From task to system
Stage 1
Choosing the use case
We describe the task, the users, the sources and what counts as a correct result.
Stage 2
Trial on your data
We check the quality of the sources and how the system behaves on representative cases.
Stage 3
Integration and testing
We connect the systems required, grant the permissions and test the exceptions.
Stage 4
Launch
We give users access, switch on monitoring and document how the system is supported.
Common questions
Answers on knowledge sources, permissions, testing and responsibility for the result.
An assistant prepares information or an answer for the user. An agent can also carry out an agreed operation in a system, for example create a task or complete a record.
Contact
Discuss your AI use case
In the first conversation you describe the task, the data sources and the people who use the result. Afterwards we set out the scope needed for a fuller assessment.
- 01
Task
You describe an activity that today means searching for information or working across several systems.
- 02
Sources and users
We establish which data may be used and who will check or act on the result.
- 03
Further assessment
We set out what has to be verified before the scope of the solution and the approach to testing can be defined.