Legal Aid Chicago

Legal Aid Chicago, which has served low-income clients for over 50 years, faced an access-to-justice bottleneck with over 78,000 calls annually and more than 6,500 monthly phone requests for service. Staff capacity was insufficient to screen, qualify, and refer everyone who needed help, leaving thousands of applicants unable to get assistance. Traditional intake consumed extensive human hours on initial screening and scheduling, contributing to staff burnout and inconsistent screening that missed opportunities and produced inappropriate referrals.

Haven deployed a voice-powered AI intake assistant on Legal Aid Chicago’s family law line, integrating with existing phone infrastructure without requiring infrastructure changes. The system handled conversational screening, eligibility assessment against Legal Aid Chicago’s criteria, automated referrals for non-qualified callers, overflow capture for future contact, real-time transcription, and case flagging for staff review. Intake staff used a dashboard to review eligibility determinations with reasoning, access transcripts, manage callbacks, track reviewed calls, and override AI decisions when needed. A three-week pilot began August 11, 2025, with mid-pilot redesign of screening questions and the intake process to improve accuracy and efficiency.

During the three-week family-law pilot, the AI triaged 417 calls (281 screened out, 136 screened in), handling 22+ hours of client engagement at an average call length of 3.24 minutes—equivalent to about three full-time intake staff days. The system produced 44 priority intakes for domestic violence–related divorce or custody issues, 200+ referrals for non-priority clients, and specialized handling for 20 urgent cases with court dates within a week. Haven reduced the rate of family law–related calls later rejected as out-of-priority from 46% under traditional triage to 21%, and a follow-up survey found an average caller rating of 4.1/5 (most rated 5; many rated 4).

• AI voice intake can keep legal aid lines open after hours and absorb volume that would otherwise be turned away, while preserving human capacity for complex advocacy.
• Mid-pilot redesign of screening questions and eligibility criteria materially improved intake efficiency and priority identification.
• Callers were often comfortable disclosing sensitive domestic violence details to an AI agent, and structured questioning can surface high-risk cases that phone-menu self-identification misses.