Zeus - AI-assisted lawyer matching
An explainable AI matching tool that helps claims and legal teams assign the right external lawyer to each case in order to guarantee the best outcome on legal cases. The solution was designed and prototyped in 6 weeks, winning a $100K global innovation championship organized by Zeus.
Lawyer selection was manual, relationship-driven, and inconsistent.
Zeus's claims and legal teams needed to assign the right external lawyer to each case, but selection was manual, relationship-driven, and inconsistent, with no clear decision process or mapped outcome. Mismatches meant slower claims resolution, uneven case outcomes, and higher legal spend at scale. Zeus opened this up as an internal innovation challenge, and Hence Technologies had 6 weeks to design and build a prototype of a smarter matching solution.
The real question wasn't whether AI could rank lawyers.
I studied how legal assistants and paralegals were actually making these decisions today, largely by memory, reputation, and informal relationships, with no consistent criteria applied across the team.
- Working with the data team, we identified the factors that should drive a match: legal specialty, jurisdiction, track record on similar cases, current caseload, and cost.
- The open question was whether legal assistants would actually trust and act on a ranked list from a system they didn't build - not whether the AI could rank lawyers.
Design for explainability, keep a human in the loop.
- Designed the recommendation screen around transparency: every suggested match showed why it was ranked where it was, not just a score.
- Kept a human decision-maker firmly in the loop all the way through the design process by using them as expert testers during AI training. Legal assistants could filter, override, and adjust criteria rather than being handed a single "correct" answer.
- We then integrated the tool directly into the existing case workflow so it fit how paralegals already worked, instead of asking them to adopt a separate system.
The mockups walk through the full workflow as a demo story - cases arrive from three channels, get triaged into a case detail view, then routed through match criteria to a ranked recommendation list.
case synced
Email, Client Portal, and Phone Intake unified into one queue.
case detail
Full context, policy, and scoring criteria collected in one place.
match criteria
Specialty, jurisdiction, track record, caseload, and cost - weighted, with manual override.
ranked recommendations
Every match shows why it ranked where it did, not just a score.
human decision
Legal assistant reviews, can override, and assigns - AI recommends, doesn't decide.
Explainability drove adoption more than raw accuracy.
For AI-assisted tools, explainability drove adoption more than raw accuracy did - legal assistants trusted a slightly-less-perfect suggestion they could understand over a black box. A tight 6-week timeline also meant design and the data team had to work in lockstep from day one, not hand off sequentially.