Harlow & Reyes LLP (fictional)Professional services2024
AI intake for a 40-person law practice
A family and employment law practice was losing prospective clients in the two days between the first call and the first consult. Ten weeks to build an intake system the paralegals actually trust.
- AI adoption
- Operations
- Hiring
- -71%
- First-call to consult booked
- 22% → 34%
- Consult to signed client
- 26
- Paralegal hours freed weekly
Context
Harlow & Reyes is a two-office practice with eleven lawyers and a busy front desk. Most new matters start with a phone call. The call was answered well. What happened next was not.
Notes went into an inbox, a paralegal re-typed them into the practice system when they had a spare hour, a conflict check ran the next morning, and the prospective client heard back in two to three days. By then, a quarter of them had signed with someone else.
The challenge
The managing partner wanted "AI for intake". The paralegals wanted the partner to stop saying that in meetings.
Both were right. The bottleneck was not the call. It was the re-typing, the conflict check, and the fact that nobody owned the two days in between. AI could help with the first two. The third was a decision.
Approach
- Weeks 1–2
Map the two days
Shadowed the front desk and both paralegals. Timed every step from the first ring to the consult confirmation. Wrote the map on one page.
- Week 3
Structured intake
Replaced the free-text notes with a short structured form. Turned out that solved half the re-typing without any AI at all.
- Weeks 4–5
Assisted summaries and conflicts
Built an assistant inside the case system that drafts the matter summary from the form and pre-fills the conflict check for a paralegal to approve. Every draft is reviewed by a human before anything is filed.
- Weeks 6–9
Trust and ownership
Ran it in parallel with the old process for four weeks. Paralegals scored every draft. Errors were fixed in the prompt and the form, not blamed on the tool.
- Week 10
Hand over
One paralegal now owns intake end to end, with a dashboard showing call-to-consult time by day. Hiring plan adjusted: the practice needed a second intake owner, not another lawyer.
Results
- -71%
- First call to consult booked
- 34%
- Consult to signed client
- 26
- Paralegal hours freed, per week
Call-to-consult went from two to three days to the same afternoon for most matters. The signed-client rate rose because fewer people wandered off, not because the lawyers got better at consults.
I asked for AI. What I got was a process with AI in the middle of it, and a paralegal who runs it better than I could.
What I'd do differently
I would skip the parallel run after two weeks instead of four. By the end of week two the paralegals were correcting the tool less than they corrected each other, and the extra fortnight was reassurance, not evidence.
I would also push harder on the hiring conversation earlier. The practice spent three weeks debating whether the freed hours justified a second intake role. The dashboard answered it in a day.