← All services

AI assistant for your documents

Ask it a question, cut a long contract down to a page, compare two versions, pull a table. Always with the source.

what we do

we take the documents you already have, contracts, case files, procedures, old proposals, and put them to work in four ways. you ask something in plain words and get the answer with the document it came from. you hand over a long document and get back just the fields you care about, on a template agreed once and applied the same way every time. you ask the same question of the whole pile and get a table, not a list of files to dig through. you put two versions of the same document side by side and see what moved between them, clause by clause.

how we work

we start from the work your team does by hand today, not from the technology. we pick the first archive and the first two things to get out of it, prepare it (scans included), then test on real cases until the results hold up. access rights are mapped onto the ones you already have, so nobody ends up seeing more than they saw before.

what you get

a distributor with 400 supplier contracts gets a table of the ones auto-renewing next month, with the notice period on each. an architecture practice takes an 80-page tender pack and pulls the deadlines and penalties onto a single page. a procurement team puts the version they sent next to the one the partner sent back and sees in seconds which clauses moved. in all three, the archive stops being the place where things get lost.

Frequent questions

It answers from your documents, not from thin air, and it shows the source every time. Same for summaries and tables — every row shows the document and page it came from. If the answer isn't in the archive, it says so instead of guessing. Before go-live we test it on real cases from your team.
No. It stays in your own space, and where we use a model provider we do it under a no-training agreement. For sensitive material — personnel files, contracts, medical records — the model can run on your own infrastructure with nothing leaving the company.
That's the normal starting point, not a problem. Scans go through text recognition, duplicates get stripped out, and if something genuinely isn't readable we tell you up front, not afterwards.
A first version over one archive takes 3–4 weeks. The full thing, with several sources and per-person access rights, 2–3 months.

Other services

Have something in mind?

Tell us what you're building. We usually reply within a day.