A law firm that processes client intake in minutes, not days
A mid-size commercial law firm was losing potential clients to slow intake processes. We built an AI-powered intake and document analysis pipeline that cut time-to-first-response from 3 days to under 10 minutes.
The challenge
A 12-person commercial law firm specialising in contracts and disputes was handling new client enquiries manually. A potential client would fill out a contact form, and a partner would review it — sometimes days later — before deciding whether to reply, request more info, or decline. The lag was costing them clients who had already engaged a faster-moving competitor.
What we built
We built a two-stage intake pipeline. Stage one: when a prospect submits a form, Claude reads their enquiry and any uploaded documents, classifies the case type, assesses complexity and fit against the firm's practice areas, and generates a structured summary. Stage two: an n8n workflow routes that summary to the right partner via HubSpot, creates a CRM contact, and — if the case is a clear fit — sends the prospect a personalised acknowledgement email drafted by Claude explaining next steps. Complex or ambiguous cases flag for human review before any outreach.
The outcome
First-response time dropped from an average of 3 days to under 10 minutes for clear-fit cases. Partners now spend their screening time on the genuinely ambiguous cases — roughly 30% of inbound — rather than on work the AI handles well. Consultation bookings tripled in the first quarter after launch.
The before state
Partners were effectively doing triage work that required legal judgment for maybe 30% of cases, and clerical judgment for the other 70%. A prospect asking about a straightforward NDA review sat in the same queue as a complex multi-party dispute. Response time depended entirely on how busy a partner was that day.
What we built
The document analysis component is a Python service that uses Claude to read uploaded contracts, extracts key facts (parties, dispute type, jurisdiction, claimed value), and compares them against the firm's practice area taxonomy. The classification output feeds directly into the n8n routing workflow. We also built a confidence-threshold system — below a set confidence score, the case is flagged rather than auto-responded, protecting the firm from misclassification.
How it runs today
Partners open HubSpot each morning to a queue of pre-assessed leads, each with a one-paragraph AI summary and a recommended next action. Clear fits have already received a tailored acknowledgement. The firm has expanded to two new practice areas since launch, partly because the intake system made taking on more volume tractable.