Operator
The operator of datapraxis.ee is Datapraxis OÜ, Pärnu mnt 22, Kesklinna linnaosa, Tallinn, Harju maakond, 10141, Estonia.
- [email protected]
- Phone
- +372 675 7540
- Office hours
- Monday to Friday, 9:00 – 18:00 (Tallinn time)
Responsible AI at Datapraxis OÜ
We write and consult about practical AI adoption for Estonian teams and public-sector offices. This page describes how we handle data, who approves the output, what we log, and when a question belongs with your own data protection contact instead of with us.
General information, not legal advice. Sector rules and special-category data need qualified Estonian counsel.
Scope
The operator of datapraxis.ee is Datapraxis OÜ, Pärnu mnt 22, Kesklinna linnaosa, Tallinn, Harju maakond, 10141, Estonia.
Third-party tools you test on your own are outside this page. Their terms, retention and training practices are theirs, not ours.
Read each provider's own notice before you paste work into it — particularly for anything naming people, health, or payroll.
Operating rules
Four decisions we ask every team to write down in one page, before anyone connects a data source. They are the same four on a customer-service inbox, a contract reading task, or a shift-scheduling draft.
On this page they read as notes. In practice each one takes a meeting and one owner.
Use the smallest set that still tests the question. Drop duplicates, old drafts, and any field the task does not need. For support pilots, exclude identity numbers and health details unless the task truly requires them.
Name one person who approves external output. Define the uncertain cases that pause for manual handling. If corrections spike, pause the pilot and fix the source material before turning it back on.
Limit accounts with export rights. Use separate test folders rather than live customer stores wherever the tools allow it. Note retention periods for logs and drafts, and brief everyone who will press approve on what to check.
We date every article and correct errors from the source. When a programme or a product changes its stated function, we update the page and note the change. Readers can report an issue by email with the page address and the sentence.
Oversight and stop rules
For every pilot we name one person inside your team who approves anything that leaves the building. That role is not the AI, and it is not the vendor. Drafts that the approver flags as unclear go into a manual queue; the pilot resumes only after the source material is corrected.
On public-sector workflows the final decision stays with your staff. The AI contribution is noted in the record — what was suggested, what was changed, what was approved — so the file can be reconstructed later without opening a laptop.
Retention is set per pilot, in writing, before the first run. We keep the log inside your environment where the tools allow it.
Oversight is a working posture, not a policy paragraph. Someone reads the output at the end of the day, marks what needs a second pair of eyes, and closes the laptop.
Boundaries
Sector rules, employment questions, and special-category data need qualified advice in Estonia. Bring three things to that conversation and it stays short: your pilot scope, your data list, and your oversight plan.
Our pages help you prepare those three items. They do not replace review by your data protection contact or counsel.
Before the pilot starts
Read the two policies this page sits beside.
Questions
Daily habits
These are the everyday behaviours behind küberturvalisus in a small Estonian team. None of them need a large tooling budget. Review them at the start of each pilootprojekt and again when it closes.
Not everyone needs to pull a full table out of the system. The pilot works fine with read-only access for most of the team and one named exporter.
Where the tool allows it, separate the pilot from the customer stores. Copies and duplicates are simpler to delete when they never touched a live folder to begin with.
Drafts and logs need a date when they go away. Write the retention period into the pilot brief so nobody has to guess later.
Anyone who presses approve should know what to look for. A short list of the three most common fixable mistakes beats a long policy.
Run through the same short checklist when the pilot starts and when it ends. Habits tend to drift quietly in between; a second pass catches that.
If you find an error in an article, write to us with the page address and the sentence. We correct from the source and note the change on the page.
We do not use tracking beyond what our cookie notice describes. For details, see the Privacy Policy and Cookie Policy.
If a pilot is on your desk and the governance side is still open, send us the scope, the data list and the oversight plan. We will read them the way this page describes — as preparation, not as legal review — and tell you plainly which parts belong with your own adviser.
Datapraxis OÜ · Pärnu mnt 22, Tallinn 10141 · Mon–Fri, 9:00–18:00