A practical setup for accountants, lawyers, and any business handling sensitive client data. Hours saved each week, with personal data staying exactly where it belongs.
All AI use is split into two zones, decided by one question: does the task touch a specific client or person? If yes, local. If not, cloud is free. The whole value rests on the setup being hard to get wrong, with two clearly separated shortcuts on the desktop, one simple rule, and proper team training.
An AI model installed on your own machine. Data never leaves the device. No cloud, no third party, no transfer abroad. Used for case material, client documents, postings, finances, anything with CPR numbers, names, or financial details.
The strongest models on the market, where Danish language is sharpest. Used exclusively for content without personal data, always under a data processing agreement. For general correspondence, marketing, newsletters, templates, research, and translation.
A local layer checks every request before anything is sent onward. CPR and account numbers are caught automatically and forced local. The Cases folder never goes to the cloud; Marketing does. When in doubt, the layer keeps the task local. Robust, and what we recommend.
Names and numbers are replaced with placeholders before sending, restored when the answer returns. Gives cloud-level language quality on sensitive content. But masking does not catch indirect signals, and pseudonymised data is still legally personal data. Use with care.
We deliver hardware, local model, cloud agreement, setup, and operations as a single package, not a software licence you have to figure out yourselves.
We go through your everyday workflow and find the tasks where AI saves the most time, and decide which belongs to which zone.
We install the local model on suitable hardware and set up the cloud agreement with a data processing agreement. Two clearly separated shortcuts, so the right choice is the easy one.
AI is placed in Slack and on the website, and your team gets short, concrete training in the rule and daily use.
We keep models updated, monitor security, and adjust as you find new tasks.
An informal call about your everyday workflow, and the two or three tasks where you will gain time fastest.