Sector · Public sector

The technology is not the hard part. The procurement is.

Most agencies already know what an AI assistant can do. The questions that stall the project come from security and procurement: where the data sits, who processes it, who has access, and what gets deleted. Here are the answers, with the documentation behind them.

The problem

The objections come from security, not from communications.

The communications team is usually convinced long before the project stalls. What stalls it is a risk assessment with questions no vendor can answer with a brochure: which subprocessors handle the content, what remains in the logs, and how deletion is documented.

  • Customer data is not used to train public foundation models
  • Role-based access, activity log and SSO in the GPT Module
  • Data processing agreement and subprocessor list on request
Employee
what is the deadline for responding to a freedom of information request?
MotherX
Requests must be answered without undue delay. Our routine sets one to three working days as the norm. Source: Access request routine, point 4.
Employee
and if we have to refuse?
MotherX
A refusal must cite its legal basis and state the right of appeal and the appeal deadline. Source: Access request routine, point 6.
Documented

What the risk assessment asks about.

These are the current public descriptions of the solution, not proof of certification. For a formal assessment they should be checked against the data processing agreement and security annex.

🇪🇺

Storage in the EU

MotherX states that customer data is stored in the EU and that the solution is built for GDPR.

🚫

No model training

The data is not used to train public foundation models. Necessary text fragments are processed by AI subprocessors through APIs.

🔒

Encryption

AES-256 at rest is stated, and TLS over HTTPS for documents, questions and API calls.

🧱

Tenant separation

A tenant ID separates organisations in storage, vector index and logs, and queries are scoped to the right tenant.

👥

Role-based access

Separate roles for uploading, crawling and viewing logs.

📜

Activity log

Uploads, crawling, questions and administrative changes are logged.

🔑

SSO

The GPT Module is stated to support Microsoft 365 and Google Workspace. Check plan and setup.

🗑️

Deletion

Data and conversations can be deleted. The exact process and deadline are confirmed in the agreement.

How it works

From assessment to production without detours.

01

Define purpose and sources

What the assistant should answer, and which documents and pages it answers from. The scope is what governs the risk.

02

Review the documentation

Data processing agreement, subprocessor list and security annex. We send the current versions so they can go into the risk assessment.

03

Pilot a limited area

One service or one subject area first. That gives real numbers on accuracy before you widen the scope.

04

Go live with roles and logging

Access is set per role, and the activity log gives traceability on what has been uploaded, changed and asked.

Insight

See what employees and citizens actually ask.

The same questions arrive at the service desk, in the mailbox and internally in the subject departments. The dashboard collects them, so you can see which routines and pages are not clear enough instead of guessing from individual cases.

  • Questions ranked by volume and subject area
  • Questions that got no good answer
  • Evidence for which routines need updating
Most asked · last 30 days
access request deadline112
who signs on behalf of the agency88
archiving email from a private account no answer64
expense claims and rates51
FAQ

Questions from procurement and security.

Where is the data stored?

MotherX states that customer data is stored in the EU and that the solution is built for GDPR. To clarify every processing location and subprocessor you should use the current data processing agreement and subprocessor list. We send both on request.

Is our data used to train ChatGPT?

No. The data is used to find relevant context and generate answers in your own solution, not to train public foundation models. Necessary text fragments and questions may still be processed by AI subprocessors through APIs, which is not the same as training. The subprocessor list shows who is involved.

Do you have ISO certification or SOC 2?

The technical security mechanisms are documented, but should not be read as proof of certification. If you ask about certification, penetration testing or audits, we send the current security documentation and answer concretely on what exists today rather than implying more.

Can we get a private cloud or a dedicated instance?

A private cloud can be agreed on Enterprise. It is a project to be scoped, not a standard delivery, so scope and price are set in dialogue.

Who is the data controller?

Normally you are the controller for your use of the solution, while MotherX acts as processor under agreement. The division of roles should be confirmed legally before it goes into a contract, and the data processing agreement is the starting point.

30 days free

Start with the documentation.

We send the data processing agreement, subprocessor list and security documentation, so the risk assessment can start on real ground.

Try for free