Automate Legal Aid Intake with Voice AI

Legal aid offices often rely on a combination of a basic phone tree, voicemail, and manual call-backs managed by paralegals or administrative staff. Tools like Clio Grow or Lawmatics offer intake forms, but they are web-based. They do not solve the problem of a distressed client calling after hours or a staff member spending 15 minutes on the phone just to realize the caller's issue is outside the firm's practice area.

Consider a small legal aid office with 10 attorneys specializing in eviction defense. The office receives 50-70 calls a day. A paralegal spends their entire morning returning voicemails. They ask the same five screening questions repeatedly: "Are you a renter?", "Have you received an eviction notice?", "What county do you live in?". Half the calls are from people with non-housing issues or who live outside the service area. This is hours of skilled staff time spent on disqualification, not case work.

The structural problem is that standard Interactive Voice Response (IVR) systems are rigid. They force callers through a fixed menu ("Press 1 for..."). They cannot understand natural language or ask follow-up questions based on a caller's response. A generic scheduling tool like Calendly does not have the context to perform the multi-step logic needed for legal intake: check for conflicts, verify service area, and confirm case type eligibility before even offering an appointment slot.

The result is a bottleneck. Qualified clients in crisis wait days for a call back, while staff burn out on repetitive, low-value screening calls. There is no easy way to capture data from these initial calls, leading to incomplete records and more manual data entry when a client is finally onboarded. This intake friction directly limits the number of clients the office can serve.

The first step is to map your existing client intake workflow. We would document every screening question, decision point, and data field required to open a new matter. This audit identifies the exact logic the AI needs to follow. You would receive a detailed flow diagram and a technical specification for your approval before any code is written.

The technical approach would use a voice-enabled API connected to the Claude API for natural language understanding. A FastAPI application would contain the business logic for screening questions. When a call comes in, the system transcribes the user's speech, sends the text to Claude to understand intent and extract answers, and follows the logic defined in discovery. Supabase would store call logs and extracted data, creating a durable audit trail. This architecture allows for response times under 2 seconds.

The final system would connect directly to your case management software's API. A qualified caller who completes the screening would have a new client record and an initial consultation appointment created automatically. The system could handle over 100 simultaneous calls and costs less than $50 per month in cloud infrastructure to operate. You receive the full source code and a runbook for updating screening questions.