AI document classification for a law firm mailroom in London in 2026

The verdict in three sentences

In a firm of 40 lawyers, filing 1,800 letters and documents a month takes more than one full-time assistant on a task with no legal value. An OCR and AI classification pipeline identifies the matter, document type and deadline, then files the document in the DMS, for a market budget of 6,000 to 15,000 EUR excl. VAT and 0.02 to 0.10 EUR per document. A realistic goal is 70 % automatic filing within 6 weeks, with UK or EU hosting compatible with legal privilege and client confidentiality.

The hidden cost of manual filing

The practice director of a firm near Chancery Lane sees scanned post, served documents, opponents' submissions, expert invoices and client documents by email arrive every day. Each one must be read, linked to the right matter, renamed and flagged to the lawyer if a deadline is running.

The most tangible gain is not always time: it is the procedural deadline spotted on receipt, whereas a misfiled letter can cost a missed appeal window.

Architecture, budget and privilege

The pipeline reads the document (OCR), extracts the matter number, parties, court and dates, then a language model proposes a classification with a confidence score. Above the threshold, the document is filed automatically. Below it, it lands in a validation queue for the assistant.

Confidentiality requires four conditions: UK or EU hosting, a model provider contractually committed not to reuse documents, an access log, and deletion of working copies after filing. A data protection impact assessment under UK GDPR is advised before go-live. At Kolonell, this project falls under the custom application grid of 4,000,000 to 7,000,000 FCFA (6,100 to 10,670 EUR excl. VAT), with maintenance of 75,000 to 120,000 FCFA a month (110 to 180 EUR).

Driving adoption in the firm

Involve the assistants from the start: they know the unwritten naming rules and the sensitive matters. Their freed time is usefully redirected to hearing preparation, billing and client follow-up. For lawyers, deadline alerts must reach the tool they actually use, often Outlook, not a new portal.

Mini case study

Illustrative example: Emma, practice director of a 40-lawyer firm in London, finds that 2 assistants spend 150 hours a month on filing. With 70 % of documents filed automatically and a quick check on the rest, the need drops to 35 hours, or 115 hours redirected per month and 1,380 hours a year. At a loaded cost of 32 EUR an hour, that is 44,000 EUR a year. The project costs 12,000 EUR excl. VAT, plus about 250 EUR a month in usage and hosting, or 3,000 EUR a year. Payback: about 4 months.

FAQ

Does the AI read client documents outside the UK or EU?

Not with the right architecture: UK or EU hosting and a contract forbidding reuse. This is a condition for protecting privilege and meeting Article 28 of UK GDPR.

What automatic filing rate can we expect?

Between 60 and 80 % depending on scan quality and how consistently matter references appear. The remaining 20 to 40 % go through a validation queue at 30 seconds per document.

Do we need to change our DMS?

No, classification plugs into the existing DMS through its API. Licences of 30 to 80 EUR per user stay unchanged.

Does the AI spot procedural deadlines?

It extracts dates and document type in 85 to 90 % of cases and alerts the lawyer. The final deadline calculation stays with the lawyer.

How long until we are operational?

Allow 6 weeks: 2 weeks of scoping and tuning on 500 anonymised documents, 2 weeks of pilot, then rollout.

Let's scope your project. Share your monthly volume, DMS and document sources: we price a classification pipeline in the 4,000,000 to 7,000,000 FCFA band (6,100 to 10,670 EUR excl. VAT), delivered in 6 weeks. Detailed quote within 48 h. WhatsApp +221 77 596 93 33.