Vera · supervisor

Hi, I'm Vera. I look after the AI colleagues that Neural Logic builds.

Below is my colleague Frida, with four archives of one company open in front of her.

Scroll, and watch her work.

Vera

One question. Sixteen thousand documents from four archives.

Frida reads exactly as fast as you scroll. Stop, and she waits. Every document is connected to the case it belongs to; nothing is skipped in silence.

Vera

Seconds later: the complete file.

In date order, quoted word for word, with the original one click away. And the one thing she could not find, said plainly.

FridaFile reader
Case
Sample data0.0 s

16.863 documents in four archives. Nobody has time to read them all.

Contracts, mails, invoices, protocols, scans. The answer to almost any question is in there somewhere. Finding it by hand takes hours, and nobody can say for certain what is missing.

Demonstration on sample data. The archives and the case are invented; the mechanism is real.

Supplier contract · Nordvik Components

27 documents touch this case5 decisive · 1 not foundin date order
  1. 21.02.2022
    Offer, Nordvik Components
    Mail archive
  2. 03.03.2022
    Mail: draft agreement
    Mail archive
  3. 14.03.2022
    Framework agreement, §4 pricing
    “…prices are fixed for twelve months from first delivery…”
    Data roomContracts / Nordvik / 2022
  4. 28.03.2022
    Order confirmation 2022-041
    Shared drive
  5. 12.04.2022
    Delivery note
    Shared drive
  6. 19.04.2022
    Invoice 3902
    Accounting
  7. 30.05.2022
    Quality report, batch 7
    Shared drive
  8. 08.09.2022
    Mail: delivery dates Q4
    Mail archive
  9. 17.11.2022
    Price list 2023
    Data room
  10. 24.01.2023
    Invoice 4108
    Accounting
  11. 06.03.2023
    Mail: price query
    Mail archive
  12. 22.06.2023
    Meeting minutes, supplier review
    Shared drive
  13. 04.09.2023
    Mail: proposal to index prices
    Mail archive
  14. 02.11.2023
    Amendment to §4
    “…adjusted once a year by the producer price index…”
    Data roomContracts / Nordvik / 2023
  15. 09.11.2023
    Mail from Nordvik: “signed copy to follow”
    “…the countersigned amendment will follow by post this week…”
    Mail archivepurchasing@ · inbox
  16. 13.12.2023
    Credit note 2023-19
    Accounting
  17. 15.01.2024
    Minutes, purchasing review
    “…Nordvik confirms the indexed price from January; the signed amendment is still outstanding…”
    Shared drivePurchasing / Nordvik / 2024
  18. 07.02.2024
    Order confirmation 2024-006
    Shared drive
  19. 26.02.2024
    Delivery note
    Shared drive
  20. 11.03.2024
    Mail: delivery delayed
    Mail archive
  21. 29.04.2024
    Quality report, batch 12
    Shared drive
  22. 21.05.2024
    Invoice 4471
    “…unit price 38.40, as per amendment of 2023…”
    AccountingPurchase ledger · 2024
  23. 03.06.2024
    Payment posted, invoice 4471
    Accounting
  24. 18.06.2024
    Mail: reminder, signed copy
    Mail archive
  25. 09.07.2024
    Price list 2024/25
    Data room
  26. 22.08.2024
    Invoice 4530
    Accounting
  27. 05.09.2024
    Mail: annual review
    Mail archive
  28. Not found
    A signed copy of the 2023 amendment.
    “The mail of 9 November 2023 promises it and the reminder of 18 June 2024 asks for it again. It is not in the contract folder and not in the mail archive. If it exists, it is not in the material I was given.”
    Looked inData room · Mail archive

What just happened

Done in 4.2 seconds. By hand, in a real case: over an hour.

16.863 documents read, 27 connected to the case, the five that decide it quoted word for word, and one gap named instead of skipped. Sample data and an invented case; the mechanism is real.

Two roles, one boundary.

Frida, who reads your documents, is chosen for where she processes: a European model, Mistral, run in Europe under a European contract. Your documents do not leave that boundary.

I never see a document or a name. I see the shape of the work: what kind of question, how many documents, which word was missing. So I can be chosen for how well I think: the strongest model available, wherever it runs. Every night I read how Frida worked and improve her before the morning.

One boundary, two roles, and a reason for each. It is the same for every colleague Neural Logic builds.

Vera · supervisor

Reads how the work went. Never what it was about.

Improves the colleague overnight: new vocabulary, better habits, fewer misses.

strongest model availablesees no documentssees no names
no documents, no names cross this line
Frida · colleague

Reads your documents. Quotes them. Says what is missing.

Answers from a checked register, never from memory.

European model · Mistralprocessed in EuropeEuropean contract

Built to be trusted.

Five commitments that stand in every contract.

  • DecisionsPeopleThe colleague lays out the facts. A person weighs them and acts.
  • Writing to your systemsOffSwitched on only by invitation: where you ask for it, and where the rules allow it.
  • Statements about peopleNeverFiles on things: buildings, products, contracts, cases. No scores, no rankings, no predictions about people.
  • ProcessingEuropeUnder a data-processing agreement. Nothing is read before you have said so in writing.
  • Every quotationCheckedCompared word for word with its source before it reaches you. If it does not match, it is held back and you are told.
  • The archiveYoursConversion turns your files into a checked, addressable archive that works with any language model. If we part ways, you keep it all.

What we are working on.

Real work, in its real state.

  • FridaA document colleague for teams that manage assets and the paperwork around them: contracts, protocols, correspondence, invoices. The first contract is the proof on real cases.In preparation
  • VeraFrida's supervisor. Reads how the work went each night, never a document, never a name, and improves Frida before morning. The same role will stand behind every colleague we ship.In preparation
  • PiaA catalogue colleague for web shops that live off supplier data. Every day she reads the supplier's feed, keeps the shop's products true to it, and reports what changed and what she could not resolve. Running daily on a related shop of our own.Live, daily
  • NextA working colleague built on a prospect's own case within days, so the decision is made on evidence rather than slides. And Pia for shops that are not our own.Testing on ourselves

Where this fits.

Any business whose knowledge sits in files nobody has time to read. The documents differ; the mechanism is the same.

  • Property and asset managementleases, handover protocols, inspection reports, tenant correspondence
  • Law and accounting practicesclient files, filings, engagement letters, years of correspondence
  • Insurancepolicies, claims, expert reports, endorsements
  • Trading and wholesalesupplier contracts, price lists, order confirmations, invoices
  • Engineering and constructionbills of material, drawings, site minutes, change orders
  • Agencies and studiosbriefs, proposals, approvals, change requests

One company, two names.

Neural Logic is run by Thomas Möller, founder and director of Axon Trade ApS in Copenhagen. We work in Danish, German and English, and we would rather show you a working colleague on your own case than a slide about one.

Opens your mail program, addressed to info@neurallogic.dk, with the line above as the subject. Nothing is stored here.
CompanyAxon Trade ApS
Register no.CVR 45 92 07 63
Addressc/o LegalDesk.dk ApS
Njalsgade 21F, 2.
2300 København S · Denmark
Contactinfo@neurallogic.dk
RegisteredDanish Business Authority · founded 18 September 2025
BrandNeural Logic is a brand of Axon Trade ApS. Contracts, invoices and the data-processing agreement carry the company name.