Clause-level citations
Every answer opens the passage it relies on — check the authority in one click.
Law firms and in-house teams live in documents nobody has time to re-read. Marella answers from your matters, precedents and playbooks — and proved itself on exactly this problem with BAiSICS.
Marella — Legal
PrivateWhich of our precedents cover delayed practical completion?
Three precedent files apply. The closest matches your amended JCT position on liquidated damages — clause 2.29.
¶ Precedent bank · 2.2901
Hours in data rooms and bundles for answers a system should surface in seconds.
02
Scans, faxes and messy PDFs defeat most tools before analysis even starts.
03
A summary you can't cite is a liability. Legal work runs on authority, not vibes.
Every answer opens the passage it relies on — check the authority in one click.
OpenKit's document pipeline was built for large, complex, poor-quality files — the BAiSICS system surpassed GPT-4-class tools on exactly these.
Information barriers by design: departments and roles keep matters separated.
A contract-review agent that knows your positions, flags deviations and drafts in your house style.
For BAiSICS, OpenKit built legal document analysis with source verification that cut per-document review from 2 hours to 10 minutes and saved £200K+ in annual billable hours — with accuracy ahead of the AI legal tools they compared.
Access is scoped by department, role and document set, and every access is logged. Privileged material can be restricted to named teams — and excluded from any wider corpus.
Yes — handling large, complex, poor-quality documents was the core of the BAiSICS engagement, where the system outperformed the legal AI tools it was tested against.
Agents can produce structured outputs — summaries, clause tables, briefs — using your templates, always with references back to source.
A working demo on material like yours — not a slide deck. Bring your hardest documents.
Private by design · ISO 27001 · Built in the UK