Public sector

AI for casework, reports and answering citizens, without the data leaving your organisation.

Thousands of grant applications sharing one deadline, an incoming register running weeks behind, and the person who knew how to process that file a year from retirement. Inferana puts open models to work inside your organisation, with the data staying put.

The rules that apply to you, point by point

What each regulation in your sector requires, and which part of that the deployment solves.

ENS

Royal Decree 311/2022 (Spain)
What it requires

Categorisation of the information system and the security measures that correspond to its level.

How we solve it

Deployed inside your own infrastructure, the service falls under your system's categorisation and its measures.

Automated administrative action

Act 40/2015, art. 41 (Spain)
What it requires

Define the responsible body, inform the citizen and allow human review of the decision.

How we solve it

An audit trail of every query and access control per user and team, to document what was produced and who reviewed it.

AI Act

Regulation (EU) 2024/1689
What it requires

Annex III: high risk in access to essential public services and social benefits.

How we solve it

Open models with published model cards and licences, with access control and traceability of every interaction.

GDPR

Regulation (EU) 2016/679
What it requires

Legal basis, minimisation and control over where citizens' personal data is processed.

How we solve it

Processing in the EU or inside your own infrastructure: prompts and responses are never stored.

Your most common use cases

The same ones you already use, specific to your sector.

See all use cases

Work comes in through the register every single day.
The team does not grow.

The call closes on Friday, the register has been running behind since January, and the person who knew how to process that file retires in March. What you hear next is always the same:

What happens today

The call closed with thousands of applications and three people to review them.

Head of grants

The incoming register is weeks behind and the backlog keeps growing.

Registry office

The person who knew how to process those files is retiring and nothing is written down.

General secretariat

We answer in two languages and the deadline runs the same in both.

Citizen services

I have files stalled since March waiting for someone to find the precedent.

Procurement section head

My system is categorised and that tool sits outside the scope.

Information security officer

The tender requires the data to stay in Spain or the EU.

Data protection officer

With Inferana

Every application and its annexes come in as structured data, with the terms of the call checked one by one.
Incoming records are classified on arrival and routed to the department that fits, with the reason for the routing.
Search over the case files you have already resolved, with the precedent and the source reference in every answer.
Open models that work across languages, with the draft grounded in your own templates and your own regulations.
Background, applicable regulation and a draft resolution gathered on the open file, leaving the signature as the only step left.
Deployment inside your own infrastructure, under the categorisation and the measures of the system you have already declared.
Inference on European infrastructure or inside the organisation, with the processing location written into the contract.

Sector FAQs

Flat rate. The cost does not run per query or per token, so the peak of a call and a dead August cost the same. What is worth sizing beforehand is the infrastructure if the deployment is yours, because there the ceiling is the hardware you have.

With verifiable facts for the usual criteria: a European provider, processing located in the EU or inside your own infrastructure, and the subcontracting chain declared in writing with its dependencies. Assessing capacity and awarding the contract remain yours. We supply the technical documentation that backs it up.

The models are open and the API is OpenAI-compatible, so the exit does not depend on us. The weights stay where they are deployed, integrations point elsewhere by swapping the key and the endpoint, and data and indexes are exported in open formats. No ransom fee, no lock-out window.

Article 41 of Act 40/2015 requires you to define the responsible body, inform the citizen and allow human review. That part belongs to the public body. We provide the record of every query and access control, so you can document what was produced and who reviewed it. The sensible use is drafting and proposals, signed off by a person.

Open models work across languages, with quality that varies by language and by task. We test with your own texts before committing to anything, and answers are grounded in your documentation. We do not assume identical performance in all of them.

Yes, with the same key. There the query does leave the perimeter, going out to the model provider. Citizen data and case files are handled with the open models inside your own infrastructure. Access is configured per team.

Access control per user and per team, with an audit trail of every query. The information security officer, financial control and audit can review who asked what and when, and export it for an audit or for a freedom of information request.

No. The API is OpenAI-compatible: swap the key and the endpoint and your code stays as it is. That also covers anything already integrated through that route into your case management system or your citizen portal.

Tell us what you want to deploy and we will show you where Inferana fits.

In the demo we go through your case: which models you need, where they run and what it takes to meet the regulation that applies to you.

We reply within one business day.