On premises
Models and data can remain inside the organisation's infrastructure.
AI · HUMAN IN THE LOOP
AI should remove routine work. People remain in control.
We place AI where it can take over rekeying, searching and first drafts. It does not pretend to be infallible: it shows its sources, asks questions and waits for a person before an important step.
Discuss my AI scenario
Practical scenarios
Start with work saved—not with a model name.
We connect AI to a specific workflow. Its data sources, allowed actions and accountable reviewer must all be clear.
01Drop the document in
AI extracts supported data and prepares it for the next step. A person reviews the result before the case starts moving.
02Find what you need
The assistant searches only content the user may access and returns an answer together with the supporting sources.
03A form that knows the context
Known information is filled from context. The assistant asks clearly only for what is genuinely missing.
04Explain this workflow
It turns an OOD configuration into a readable explanation and flags branches or edge cases that still require review by the process owner.
05From policy to first draft
A specialist prepares a policy and form. AI creates the first OOD configuration draft; a person refines roles, exceptions, integrations and tests.
Where AI runs
On premises, cloud, hybrid—and increasingly at the edge.
We choose the architecture according to data sensitivity, performance, latency and operational ownership. One answer does not fit every workflow.
In the cloud
For work where a managed service or a more capable model is the right fit.
Hybrid
Sensitive content is prepared locally and only controlled context leaves the environment.
On the endpoint
Our direction for selected tasks where latency, privacy or offline operation matters.
HUMAN IN THE LOOP
AI proposes. An accountable person decides.
For important steps, we retain the sources behind a proposal and what a person accepted or changed. AI does not replace the process owner, controller or system administrator.
We expect the greatest benefit in preparing the first draft: work that is repetitive and time-consuming, yet still needs expert refinement.
Research and development
Digital document management
Together with T-MAPY, we are developing natural-language processing for capturing structured information and supporting document workflows.
Supported project details →Validate first, then automate
We measure a pilot by accuracy, source traceability and work actually saved. Only then does it become a product feature.