{"stable_id":"88bf4a45891fbb14","slug":"justice-connect-legal-problem-classifier","company":"Justice Connect","workflow_name":"Free-text legal-problem classification","function_code":"customer_service","pattern_codes":["intake_redesign","human_quality_loop"],"changed_assumption":"People seeking legal help should describe their problem, not diagnose its legal category before they can ask for help.","evidence_strength":"reported","publication_tier":"showcase","freshness":"current","reviewed_at":"2026-08-23","updated_at":"2026-08-23","source_quality_summary":"1 independent, 1 primary; publication outcomes are reported.","caveat_summary":"JusticeBench reports the operator's staged comparison. The retained non-AI path strengthens the comparison, but no independent statistical evaluation was published.","freshness_caveat":null,"workflow_summary":{"before":[{"actor":"Help-seeker","action":"Selects a legal category from a fixed list, often choosing the wrong area or 'Something else'.","actor_type":"control"},{"actor":"Intake staff","action":"Manually reclassifies uncategorized and misclassified applications.","actor_type":"human"},{"actor":"Referral partner","action":"Reviews the referral and redirects matters that do not match its expertise or mandate.","actor_type":"control"}],"hinge":"People seeking legal help should describe the problem, not diagnose its legal category before they can ask for help.","after":[{"actor":"Help-seeker","action":"Describes the legal problem in ordinary language without selecting an area of law.","actor_type":"system"},{"actor":"Legal-problem classifier","action":"Maps the free-text description to likely legal categories and confidence scores for routing.","actor_type":"system"},{"actor":"Intake staff","action":"Reviews edge cases and works from cleaner, pre-categorized applications.","actor_type":"human"},{"actor":"Pro bono legal experts","action":"Compare model classifications with expert judgments and feed disagreements into refinement and retraining.","actor_type":"system"}],"decision_mode":"shared","decision_marker":"AI classifies; experts govern exceptions"},"before":[{"order":1,"actor":"Help-seeker","action":"Selects a legal category from a fixed list, often choosing the wrong area or 'Something else'.","handoff_to":"Intake staff","control":"The applicant must understand the legal nature of the problem."},{"order":2,"actor":"Intake staff","action":"Manually reclassifies uncategorized and misclassified applications.","handoff_to":"Referral partner or internal service","control":"Human triage and service eligibility rules."},{"order":3,"actor":"Referral partner","action":"Reviews the referral and redirects matters that do not match its expertise or mandate.","handoff_to":"Another legal service","control":"Partner-specific service criteria."}],"after":[{"order":1,"actor":"Help-seeker","action":"Describes the legal problem in ordinary language without selecting an area of law.","handoff_to":"Legal-problem classifier","control":"Disclosure, opt-out, and a retained non-AI path."},{"order":2,"actor":"Legal-problem classifier","action":"Maps the free-text description to likely legal categories and confidence scores for routing.","handoff_to":"Intake staff and referral partners","control":"The model classifies and routes; it does not provide legal advice."},{"order":3,"actor":"Intake staff","action":"Reviews edge cases and works from cleaner, pre-categorized applications.","handoff_to":"Legal service","control":"Human review remains available for uncertain or consequential cases."},{"order":4,"actor":"Pro bono legal experts","action":"Compare model classifications with expert judgments and feed disagreements into refinement and retraining.","handoff_to":"Model governance team","control":"Standing expert-review loop."}],"decision_rights":"The model proposes intake categories and routing. People may use the non-AI path, and humans retain review of edge cases and all legal-service decisions.","exception_path":"Users can opt out to a non-AI route. Low-confidence or consequential edge cases receive human intake review, while expert disagreement is recorded for model refinement.","removed_work":["Requiring help-seekers to identify their own area of law","Broad manual re-triage of 'Something else' and misclassified applications","Referral roundabouts caused by mismatched intake categories"],"outcomes":[{"metric":"'Something else' selections during intake","baseline":"23% among users on the retained non-AI path","result":"9% among users offered the classifier","period":"Staged rollout, initially to about 30% and then about 50% of users","scale":"Justice Connect's production online Intake Tool","attribution_caveat":"JusticeBench reports the operator's staged comparison. The retained non-AI path strengthens the comparison, but no independent statistical evaluation was published.","evidence_label":"reported"}],"executive_lesson":"Move classification burden off the requester and into the intake system; keep experts on exceptions and model governance.","anti_pattern":"Automating intake while forcing every user onto the new path removes the comparison group and hides whether routing actually improved.","questions_for_leaders":["Are requesters being asked to classify a problem they do not yet understand?","Which cases require a human route regardless of model confidence?","How does expert disagreement reach retraining and control changes?"],"collections":["human-still-decides","queue-eliminated"],"bundle_version":"1.0.0","bundle_fingerprint":"sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93","canonical_url":"https://brianletort.ai/transformations/justice-connect-legal-problem-classifier"}