AI assistants and agents
Many businesses already have the answers, in manuals, emails, and earlier cases, but finding them takes time and the same questions come back every day. AI assistants and agents can take over part of that work. We help you find out where they make sense and put them to use.
The answers exist. Finding them takes time.
- Knowledge sits with a few people. New colleagues interrupt experienced ones, and knowledge leaves when someone retires.
- The same questions every day. Colleagues and customers ask about things that are already written in manuals and earlier emails.
- Drafts start from a blank page. Quotes, reports, and replies take hours, even when they look much like the last one.
- Unclear where to start. There are many AI tools, and it is hard to know which one fits and what is safe to use with your data.
How we can help
Finding where AI fits
We look at your tasks with you and find where an assistant or agent would save time, and where it would add little.
Assistants built on your own knowledge
We can develop assistants that answer from your manuals, emails, and earlier cases, and show where each answer comes from.
Agents that carry out tasks
We can develop agents that prepare quotes, reports, or plans in your systems, within the limits you set.
Safe use of data and access
We can set up access rights, data handling, and approval points, so the AI only uses what it should.
Keeping them useful
We can follow how they are used after launch, add new material, and improve the answers.
Five levels of autonomy for AI agents
Researchers at the University of Washington describe five levels of autonomy for AI agents, defined by the role a person plays. The framework is a useful way to decide how much an agent should do on its own in each task.
YouDecide and direct each step.
The agentCarries out what you ask.
YouPlan and work side by side with the agent.
The agentShares the planning and the tasks.
YouGive expertise and preferences when asked.
The agentLeads the task and asks you for input.
YouApprove risky or agreed cases.
The agentWorks on its own, and stops for your approval.
YouMonitor and step in if needed.
The agentWorks fully on its own.
Source: K. J. Kevin Feng, David W. McDonald, and Amy X. Zhang, Levels of Autonomy for AI Agents, University of Washington, 2025.
How we work with you
We start by looking at your tasks and your material with you. From there we agree where to begin, often with one task, and build on what works. Learn how we work.
Questions we often hear
- Which data does it use?
- Only the material and systems we agree with you, with the same access rights your people have.
- What about wrong answers?
- AI can be wrong, so answers show their sources and a person checks before anything is used. We test with real questions from your business before launch.
- Which AI do you use?
- We choose the model and platform that fit your data, your security needs, and your budget, for example from OpenAI, Anthropic, Google, or Microsoft.



