implementation
Why the last mile is the hardest
A model that scores well in a test environment is not yet a solution. It has to get into the systems where people already work, within the permissions that apply there, and with a person who checks the result where it counts. That last point is not optional for us but a requirement: AI supports, people decide. That keeps responsibility where it belongs.
What you get
- 01
Connections to your sources
Integration with the systems where the data and the users already are, within the access rights that apply there.
- 02
Human in the loop set up
Established in advance where a person looks, decides or approves, and how that fits into the existing process.
- 03
Roles and permissions
Who may see what follows the permissions you have already set up. No separate shadow administration.
- 04
Handover to operations
Documentation, monitoring and a point of contact, so the solution keeps running after delivery.
What we build
View our workIntegrations with your systems
Connections to SharePoint, Teams, Outlook, Google Drive and more, within existing permissions.
Agent orchestration
Agents that drive tools and sources, with a human in the loop at the right points.
Sovereign hosting
Running in a Dutch cloud or on-premise, deployed on Kubernetes.
How we go about it
Our way of working- Step 1
Map the work process
Where in the working day does the solution land, who uses it and what does that person need to be able to check.
- Step 2
Connect and inherit permissions
We connect to the sources and interfaces that already exist and inherit the existing access rights.
- Step 3
First version with a small group
A working version with the people who use it daily, with the benchmark alongside to track quality.
- Step 4
Measure adoption and adjust
We track who uses it and for what, and adjust the solution to what people really do.
Work in this line
All cases
Planning bus, tram and metro more intelligently.
The Hybrid Intelligence Planning Engine, built with three transport operators. After a successful first phase, it is on its way to daily use.
Unlocking a mountain of documents for supervision.
A modular, AI-driven agent system for supervision, built with the NVWA Innovation Lab: thousands of documents made searchable, with a citation for every answer.
Handling objections with agents.
A proof of concept with the Ministerie van BZK: a sovereign multi-agent setup that supports a civil servant in handling objections against parking fines.
Frequently asked questions
Does this also work with our older systems?
Usually it does. We connect to whatever is available: an API, a database or an export. In the data exploration we quickly see where it pinches and what that means for the planning.
Do we have to buy a new platform?
Not necessarily. For a single application we build bespoke and integrate into what is there. If you want an AI environment for the whole organisation, we implement soev.ai, our own platform, on your sources, roles and way of working.
Who is responsible for an AI answer?
The person who decides. We set up where someone checks and approves, and every answer shows where it comes from. With HIPE a planner can check and explain every choice.
Grab a coffee with Daan. We think along, no strings attached.
We would love to hear what AI challenges you run into within your organisation. Together we figure out what the first step would be.