About
Blue Abraham builds token factories for high-stakes document intelligence.
A founder-led consultancy, operating from the Nordics, working with mid-market CTOs and Heads of Data on AI that uses their proprietary documents as the primary source of truth.
What I do now
Architecture for AI on your own documents.
Blue Abraham engages with Nordic mid-market companies — roughly two hundred to two thousand employees — on AI that has to work on their proprietary corpora. Contracts, technical documentation, internal knowledge, structured scans, decision support. The material that does not exist in the general web and cannot be replaced by a vendor’s chatbot.
Most engagements start with an architecture review: one to three weeks, fixed price, written output. From there, some companies take the review back in-house. Others continue into a working session, a build engagement, or an advisory retainer. The full menu lives on the services page.
I operate solo. Where a build engagement needs more than one pair of hands, I bring in specific expert collaborators by name, under my architecture leadership. There is no bench, no fake team, no off-shore delivery unit behind the landing page.
Where this comes from
Aviation, on the technical side.
Before Blue Abraham, I spent my working life on the technical side of a Nordic aviation company — building and running the systems that turned aircraft documentation into something an engineer could actually use. Structured extraction from text-heavy manuals on one side, OCR and classification of scanned material on the other. Both ran in production across multiple model generations.
That work is documented on the two delivered use-case pages — the technical documentation engine and the scanned-page tagger. The company was later acquired; I can speak on record about the architecture and the outcomes. Some vendor and customer specifics stay out of public copy by default.
The methodology that Blue Abraham sells now — the token factory, four owned layers, one swappable model bus — is the generalisation of what actually held up in aviation over the long run. Not a slide, not a thought experiment.
What I believe
Proprietary documents are the moat, not the model.
Frontier models will keep changing. Which one is best this quarter is not the interesting question. The interesting question is what a company owns that no vendor can replicate: contracts, operational history, regulated documentation, structured internal knowledge. That is the asset AI multiplies — or leaks, depending on the architecture.
The build-versus-rent conversation is usually framed as a cost question. It is not. It is a question of where the proprietary signal ends up after the system has been running for two years. Rent it and the signal accumulates in the vendor’s prompt history and fine-tuning pipeline. Build it — as a token factory, with owned context, owned prompts, owned evals — and the signal stays on your side of the boundary.
Technical sovereignty is the word I use for that posture inside the methodology pages. Not as a slogan, not as a series name, just as a short term for the thing being bought: architecture where the artefacts a regulator, a board, or a future owner would ask to see are artefacts the company can actually produce.
The EU regulatory context — DORA, NIS2, GDPR, the AI Act — matters here in a concrete, non-decorative way. A token factory produces the same artefacts those regimes ask for: residency, auditability, vendor independence. Blue Abraham does not sell compliance. It sells architecture that makes the compliance conversation shorter.
Working together
Start with an architecture review.
One to three weeks, fixed price, delivered in writing. Most engagements begin there. From that review we decide together whether to go further — a working session, a build engagement, or a monthly advisory retainer.
The fastest way in is email: hello@blueabraham.com. Or use the form on the contact page. I reply within two business days, in English.