Legal services firm
Evidence document coding, two to three times faster with the same team
- per document, from manual coding to AI-assisted review
- 60s → 20–30s
- throughput per person, on the same headcount and payroll
- 2–3×
- of documents still reviewed and confirmed by a person
- 100%
per document, from manual coding to AI-assisted review
throughput per person, on the same headcount and payroll
of documents still reviewed and confirmed by a person
The situation
The firm receives evidence documents from law offices and codes each one so it can be used downstream. Every document was read and coded by hand, about a minute each. At real volumes that meant a large share of the team's week went to repetitive reading, with accuracy resting on tired eyes.
What we did
We built an AI layer over the existing workflow. It reads each incoming document and proposes the coding; staff review the proposal and confirm or correct it. Nobody's job changed and nobody was replaced. The tedious first pass simply stopped being manual.
The result
Coding time fell from roughly sixty seconds per document to twenty or thirty. Each person now clears two to three people's units of work, with no new hires and no quality trade-off. The coders' attention goes to the judgment calls instead of the typing.
Time per document
Facing something similar? We’ll tell you straight whether AI can help.