AI lease review for law firms.
BAiSICS · document analysis for law firms and property consultants
Reviewing a commercial lease by hand took a skilled professional two hours. The same document now goes through BAiSICS in ten minutes, and the reviewer can still check any answer against the paragraph it came from.
Overview
BAiSICS sells document analysis to law firms and property consultants, where a wrong answer becomes a position someone has already advised a client to take. They engaged OpenKit to build the whole thing: the brand and interface design, the marketing site, the subscriber platform, and the AI backend underneath it.
- An OCR pass tuned for skewed, poorly scanned leases with handwritten amendments, the documents generic tools quietly get wrong.
- Every extraction linked and highlighted to its exact location in the source, side by side with the original.
- A template engine each firm owns, plus a partner API so third parties can put the analysis behind their own product.
10 min
to take a commercial lease end to end, against the two hours the same document took a skilled professional by hand.
96%
of extracted fields matched the partner’s read, tested on the same set of historical leases they had already marked up.
£200K+
saved a year, as reported by the firms now running their commercial-lease work through the platform by default.
100
documents in a single batch, processed in parallel, and one bad file fails on its own while the rest of the run finishes.
Challenge
Why lease review defeats generic tools
Commercial leases are not standardised, they are often scanned badly, and the terms that matter are frequently amended by hand in the margin. Keyword search cannot cope with that scan quality, and a chat model that returns a confident paragraph with nothing behind it is worse than no tool at all to a partner who has to stand behind the answer in front of a client. That set what the platform had to do before a firm would put a live matter through it:
- A reading of the leases that defeat standard tooling, with an honest signal when the reading is uncertain.
- Every answer checkable against the source in seconds, at the paragraph the answer came from.
- Each firm deciding which fields matter to its own work, without needing a developer.
Approach
Inside the platform
A custom OCR pass turns the lease into citable text and grades how legible the document is, and the processing strategy changes with that grade. OpenKit benchmarked the extraction workflow behind it against historic leases that senior partners had already marked up, and it runs across several model providers, so BAiSICS is not tied to the availability of a single vendor.
The source sits next to the answer
Each extracted field carries a jump straight to the paragraph that justifies it, with the lease open alongside. A reviewer confirms a clause in two clicks, and the platform flags a low-confidence extraction for a person to check.
The firm sets the questions, and sees the cost first
A template loads its question set and the team edits the wording and the context note behind each question before anything is processed. The cost of the run shows against the credits the account holds, so putting a 200-page bundle through is a decision made with the number in view.
Templates each firm owns
Legal teams define their own field set and their own required-field rules, then run that template across any matching document type. Setting one up takes no code, and the fields are the ones a firm’s own work turns on.
A library that grows with the caseload
Each firm keeps its own templates beside a standard set for common document types, shared across the team so a template written for one matter becomes the starting point for the next. Firms run them across hundreds of documents without rebuilding the field set each time.
- The build: Custom OCR, Bespoke LLM workflow, Stripe billing, AWS UK region.
- OpenKit certifications: ISO 27001, ISO 9001 (UKAS-accredited), Cyber Essentials.
- Controls on this project: UK GDPR, UK data residency.
Result
What changed for the firms using it
The work started on legal document analysis, and the platform now runs in other document-heavy sectors, with each round of scope set by how the firms were using it.
Voice
In their own words
Having completed our second major AI agent development project with OpenKit, I can confidently say they're the real deal. Our system analyses large, complex and poor quality documents with remarkable accuracy, surpassing solutions like GPT-4 and other AI legal tools. The transparency is fantastic; you can instantly verify every AI output against source documents, which builds trust with our users.
Other engagements
What could your business do with AI?
Find out with experts who become part of your team. We uncover opportunities, get ideas working, and help your people build on the results. We reply within one working day.
Get in touch AI Engineering