How we work
Make the technology
earn its place.
We build agentic services around the work a business needs to do. Commercial purpose, security and human judgement shape the project from the first conversation through to the way it is used.
01
Start with a useful business change.
We begin with the customer you want to reach or the recurring work you want to improve. Where does the journey get difficult? Which information is missing? What would a useful outcome look like? Those questions tell us whether the next step belongs in your website, a workflow or the knowledge behind it.
02
Decide what the service is allowed to do.
We map the information it needs, who owns it and which actions need a person’s approval. A public website, an internal knowledge system and an agent that can change a record have different requirements. Access, source traceability and recovery are part of the scope, with the controls matched to the work.
03
Build a small, complete piece of the journey.
A working slice makes the idea easier to judge. That might be a clear route from discovery to enquiry, an agent preparing a reviewable task, or a useful answer with links back to company evidence. Nicolai leads the work with the engineering, agentic and interface skills the project needs.
04
Check it in context, then extend it.
We review the experience, the boundaries and the handoffs with the people who will use the service. Tests follow the actual risks, and release decisions remain distinct from a promising demonstration. Documentation, ownership and the next improvements develop alongside the software, so the business can keep working with it.
Before we begin
A little clarity.
What do you mean by an agentic service?
A service in which AI can work with context, use connected tools and help move a task forward. That might mean an assistant finding the right information on your website, preparing a piece of business work for review, or using company knowledge to answer a question with evidence. The useful part is the work it helps someone do, and the boundaries around its access and actions.
Can you work with the systems we already use?
Yes. We start by understanding your website, tools, information and operating constraints. A useful first project often connects or improves what is already there. The scope depends on the access available, the quality of the underlying information and the responsibilities people need to retain.
How do you approach security?
We make access, data boundaries and decision authority part of the scope. That means deciding what a person or agent can read and change, where approval is needed, how sources and actions are recorded, and what recovery should look like. The controls and verification follow the risks of the particular service; we do not promise that adding AI removes the need for judgement.
What does a project usually cost?
Most projects are around £3,000–£5,000 a month, and up from there if you need greater speed or the work is more complex. We can talk through the scope, the right pace and a useful first deliverable on the call.
Who will I work with?
Nicolai leads BTR and stays close to the commercial and technical decisions. Engineering, agentic and interface specialists contribute where the work needs them, with clear responsibilities and a shared understanding of what the business is trying to achieve.
Do I need a finished brief?
No. A customer journey that needs improving, work you keep repeating or information that is difficult to use is enough to begin. Choose a complimentary 20- or 40-minute conversation, or send a little context first. A person will help you work out a useful starting point.
Your starting point is enough
What would you like the business to do better?
Bring a customer journey, a repeated piece of work or a question about your company’s knowledge. A complimentary 20- or 40-minute conversation gives us room to work out a useful next step.