How AI creates impact
Four methods we use to set up every AI project.
It takes four things at once: The right model, the right task, a robust implementation and clear accountability.
Our methods
Open Model Stack
Technology-agnostic: The right model for each task, without being tied to one provider.
Questions it answers
- Which model can do the task?
- Where is the data processed?
- What does it cost to run?
How we work
We test models from OpenAI, Anthropic, Microsoft, Google or open models with typical cases of the task. Then we check the contract and the place of processing and calculate the running costs.
Your benefit
The right model instead of a default. You remain free to choose your provider.
Example for insurers
Claims notifications with health data are processed by a model in the EU. Answers to brokers based on the policy terms are handled by a smaller, lower-cost model.
Guide: ChatGPT, Copilot or your own AI?Value Map
Every task rated by value, effort and risk. The result is a clear order of priorities.
Questions it answers
- Where do we lose the most time?
- What pays off first?
- Where is the risk too high?
How we work
One day on site with your team. Every task is rated against the same questions:
- Value: How often does it come up and how much time does it take?
- Effort: Are the necessary data and documents available?
- Risk: Can a person check the result quickly?
Your benefit
A written plan within ten working days: What comes first, what comes later and what does not.
Example for audit firms
From minutes to the audit manual, there are many ideas on the table. First comes the task that comes up often, takes a lot of time and can be checked quickly.
More on the potential analysisBuild & Run
One team from strategy to operation. We also run our own software.
Questions it answers
- How does the application connect to our systems?
- Who looks after it after the pilot?
- What happens when something changes?
How we work
Strategy, design and development at one table. We plan integration, permissions, updates and running costs from the start. After launch we keep looking after the application and adapt it when workflows change.
Your benefit
No handovers where knowledge gets lost. An application that holds up beyond the pilot.
Example for insurers
An assistant for broker queries needs to know which tariff generation applies to which policy. We settle that before the pilot, not after it.
Human Gate
Anything that goes out or prepares decisions is checked by a person.
Questions it answers
- What does a person need to approve?
- Who may approve?
- How do we document it?
How we work
Before launch we agree with you what needs approval and who approves it. Your team sees at a glance what the AI has prepared. Every approval is recorded with name and time.
Your benefit
Control over every important result and a record of who approved what.
Example for audit firms
AI flags anomalies in the journal. The auditor decides whether to follow them up. The decision is documented with name and date.
More on AI agentsIn each of our solutions
AI Strategy & Consulting
Assess tasks, set priorities, choose models.
- Open Model Stack
- Value Map
AI Design & Development
Develop the application, with approval built in from the start.
- Open Model Stack
- Build & Run
- Human Gate
AI Support & Evolution
Run and improve the application and review the model regularly.
- Open Model Stack
- Build & Run
- Human Gate
In your industry
How we use AI in specific industries.
Frequently asked questions
Which AI providers do you work with?
For example with OpenAI, Anthropic, Microsoft, Google or open models. Which ones we use is set out in the proposal.
Can we switch the model later?
Yes. We build applications so that the model can be replaced.
Is the Value Map the same as the potential analysis?
The Value Map is the method. The potential analysis is the day we apply it with your team.
Does the AI make the final decision?
No. Anything that goes out or prepares decisions is checked by a person.
Does the Human Gate replace a legal review?
No. We clarify case by case which obligations apply under the EU AI Act. This does not replace legal advice.