Data and AI
AI agents
A software assistant that uses a large language model (LLM) to chain several steps across your tools and see a task through, within a scope you set.
Builds onData science

What it is
An agent is a program that uses a large language model (LLM) to choose what to do next and to call tools: read a document, query a database, draft a message. Because the model can get things wrong, the actions it is allowed, its limits and the points where a person signs off are part of the design.
What we do
The team defines with you the tasks the agent takes on, the tools it may call and what still needs your approval, then develops the agent and measures its results on representative cases. A record of its actions is kept so that you can review them.
During the free scoping call, we look with you at the task in view, the tools involved and what must still be approved by a person.
What you receive
The deliverable is defined in the quote. You receive it at the close-out step, and a handover session then closes the project.
| Deliverable | Contents | Handed over at |
|---|---|---|
| DeliverableDocumented source code | ContentsThe agent’s code in a repository, with its documentation: the tools it calls, its limits and its approval points. | Handed over at05 · Close-out and support |
| DeliverableHandover session | ContentsA session to hand over the deliverable, walk you through how it works and its documentation, and answer your questions. | Handed over at05 · Close-out and support |
Same area
- 01ModelsModels trained on your data, to forecast, classify or analyse.
- 02Data pipelinesThe pipeline that collects, cleans and serves your data.
- 03LLM integrationA large language model connected to your tools or documents.
- 05AutomationRepetitive tasks handed over to a program.
- 06AI advice and scopingWhether AI helps your case, and where to start.