Make your archive machine-readable.

The work nobody has done yet

A PDF is an electronic document, not a machine-readable one. Neither is a scan, a mail thread or a folder named “final_v3”. They are readable by people, and opaque to everything else.

This is why AI projects in companies with real archives stall. The models are fine. The material is not: it has no structure, no addresses, no way to say which clause sits in which version of which contract. Anything built on top of it is guesswork wearing a confident tone.

What conversion actually is

Your archive is read once, properly. A scan becomes text. A contract becomes clauses. A mail thread becomes dated messages with their attachments. A protocol becomes decisions. Every piece keeps a link back to the page it came from, so any sentence can be traced to its original.

While that happens, two things are written down: a register of what each document contains — and what was expected and never arrived — and a glossary of your own vocabulary, the words your trade and your company use for things.

What you get, and keep

• The converted archive, in open formats, usable by any language model — ours, or the one you choose in three years.
• The register: which document holds which clause, which protocol is signed, which invoice belongs to which case, and which documents are missing.
• Your glossary: the terms your company actually uses, so a machine reading your files does not have to guess at them.
• A written account of what could not be read, and why. Unreadable pages, broken scans, files that were empty. Most archives contain a few, and nobody knows until someone reads them all.

All of it is yours, in your hands, not locked in a system of ours. That is not generosity; a client who cannot leave is not a client, only a captive.

Why this is worth doing on its own

Whatever you eventually use AI for, and whoever you eventually hire to build it, this work comes first. It is the slow part, and it does not get cheaper by waiting. Doing it once, on a small contract, with the result yours to keep, means it is behind you before you commit to anything larger.

There is a second return that clients rarely expect: you find out what your archive actually contains. The contract everyone refers to that was never signed. The four versions of a document that differ in one clause. The years where the correspondence simply stops. That is worth knowing whether or not a machine ever reads it.

How it runs

A small first contract on your real files, with a fixed scope you name: one property, one client's matters, one year of correspondence. Acceptance is judged on what it produces, not on a demonstration. You can stop after any phase and keep everything built so far.

The work happens in Europe under a data-processing agreement with Axon Trade ApS, the Danish company behind Neural Logic. Nothing is read before you have said so in writing, and only the archives you put in scope are touched.

What it is not

• Not a system to move into. Your people keep working exactly where they work now.
• Not a migration. We work from a copy. Nothing in your systems is changed, moved or written to.
• Not a scanning service. We work from digital copies; paper you already scanned is read as text like anything else.
• Not a promise that AI then works miracles. It is the groundwork that makes the question worth asking at all.

Where this stands today

Conversion is the part of our work that is furthest along: the same method runs every day on a catalogue of our own, where a colleague of ours reads a supplier's feed and keeps a live shop true to it. For document archives, the first client engagement is being prepared now. We would rather say that plainly than describe a plan in the present tense.