Skip to content

AI solutions

AI that lives inside the systems your people already use: recommendations, assistants and automation on your own data, with the access control and audit trail an organisation needs.

AI features inside custom platforms

The most useful AI is rarely a separate chatbot. It is the recommendation that appears at the right step of a workflow, the draft that is already written when a user opens a form, the search that understands the question. We design these features as part of the platform, measured against a baseline, and we keep a human in the loop where the decision matters. On the EAPSI national platform this means course recommendations driven by each citizen's self-assessment and AI-assisted generation of assessment questions, reviewed by the academy's staff before publication.

Assistants and agents on your data

Assistants that answer from your documents, records and systems, not from the open internet. We build retrieval over your content with the same role-based access the rest of the platform has, so an assistant never reveals what a user could not open themselves. Agents can take actions through your APIs with explicit limits, logging and approval steps.

Process automation with language models

Classification, extraction and summarisation of documents, emails and forms; routing and triage; back-office steps that used to be copy-and-paste. We start from a sample of real cases, measure accuracy, and only automate what the numbers support. Everything else stays with a person, assisted.

Where the models run and what it costs

We work with hosted models from the major providers and with open models deployed in the EU or on your own infrastructure, depending on data sensitivity and procurement rules. Every feature ships with cost monitoring per request and per user, so AI spend is a line you can see, not a surprise.

Related work

National Academy of Digital Skills (EAPSI), gov.gr

EAPSI My Academy

Learning platform and self-assessment tools for the national digital skills academy, with gov.gr single sign-on.

Read the case study Web Visit live site ↗

Frequently asked questions

Where is our data processed?

Where you decide. Hosted models in the EU, open models on EU cloud or on your own servers, or a mix. Data retention and training-use settings are part of the specification.

How do you keep the assistant from revealing restricted information?

Retrieval runs with the same permissions as the user asking. Documents a user cannot open are never passed to the model for that user.

How do we control the cost?

Every call is metered per feature and per user, with budgets and alerts. We choose smaller models where they perform equally well on your evaluation set.

When would you advise against AI?

When the task has a deterministic rule, when the error cost is high and there is no review step, or when there is not enough real data to evaluate. We say so during discovery.

Have a similar project in mind?

Tell us about your organisation and what you need. We reply within one working day with the next steps.

Discuss your project