Each contract is read once and its key terms — parties, dates, renewal and notice, liability caps, payment and indemnity — are written into a searchable register, with every value linked back to the clause it came from. That is 5–10 minutes per contract instead of an hour, at 90–95% field-level accuracy on standard forms.
“Which agreements auto-renew before March?” is a reasonable question with an unreasonable answer: someone spends a day opening PDFs, and the answer is only true until the next contract is signed.
Somebody started a tracker. It covers the contracts that were live when they started, in the fields that mattered that quarter, and it has been drifting from the actual documents ever since.
A notice period that had to be exercised sixty days out is invisible until it is fifty-nine. The cost of that is not legal work — it is another year of a contract nobody wanted.
The DMS, SharePoint, a shared drive or the signing platform. Scanned paper is handled too — OCR runs first, and pages that come through badly are flagged rather than silently half-read.
Not a generic clause list: the fields your team actually tracks, defined with you up front. Parties, term, renewal mechanics, notice windows, liability and indemnity, payment terms, governing law, assignment, whatever your portfolio turns on.
Every extracted value carries the clause number, page and the sentence it came from. A lawyer verifies in seconds by looking at the quote instead of reopening the contract.
Low confidence, an unusual formulation or a clause that contradicts another in the same document goes to a person with the reason attached. Nothing ambiguous is written to the register as fact.
New contracts and amendments are processed as they are signed, and date-based fields drive alerts — renewals, notice windows, expiries — so the register acts rather than just records.
We build in your stack rather than moving you onto ours. The list below is what this solution most often connects to — other systems are a scoping question, not a blocker.
On your own templates and common market forms, field-level accuracy is 90–95%. On heavily negotiated or unusual paper it is lower, which is why every field carries a confidence score and low-confidence extractions go to a person instead of into the register. We measure this on your own contracts during the pilot, not on a benchmark set.
We use cookies for analytics — to see which pages bring enquiries. Nothing else, and nothing before you agree. Cookie Policy