AppRocket
All legal-firm AI surfaces
Agent template

Intake Triage Agent

Inbound queries routed to the right practice group with conflicts pre-flagged in under 30 seconds.

6–8 weeksOverride band: 8–18% (healthy)

Outcomes

  • Senior-attorney triage time reduced 60–80% on routine inbound (firm-tunable)
  • Conflict surprises caught at intake, not at engagement letter
  • Inbound-to-engagement-letter cycle time compressed measurably
  • Audit trail every inbound query, every routing decision

How this surface works in production

Intake is the AI surface where mid-market firms get the highest leverage and the fastest payback. A senior attorney spending fifteen minutes triaging an inbound query that should have routed to a junior partner is fifteen minutes the firm cannot bill at senior-attorney rates. A junior associate manually checking conflicts before the matter has been scoped is junior time spent on work the conflicts team would do better.

Our intake-triage agent is deployed inside your existing case management system (Clio, Centerbase, Litify, or whatever you run) and handles the routing decision attorneys should never have been doing manually. Inbound queries from your web forms, monitored email aliases, and phone-transcript services flow into a triage queue. The agent classifies the matter into your practice taxonomy, runs a preliminary conflict screen against your conflicts database, drafts a one-paragraph matter summary, and routes the query to the right partner with all of the context attached. The receiving attorney scans the transcript in under 30 seconds and either approves the routing (one click) or overrides it (one click, with the override fed back into the eval set).

Eval discipline is non-negotiable on this surface. We deploy with a regression test set drawn from your firm's last 12 months of inbound queries, refreshed monthly, with the agent required to clear a firm-tunable accuracy threshold before any production traffic flows. Hallucination is bounded: the agent does not draft legal advice, does not commit to scope, and does not promise the client a specific attorney by name without explicit approval.

What this surface does

    01

    Multi-channel intake

    Web form, monitored email aliases, phone-transcript services, and partner-referral inbox unified into a single triage queue.

    02

    Practice-group classification

    Matter classified against your specific practice taxonomy (litigation / corporate / IP / regulatory / etc.) with attorney-tunable confidence thresholds.

    03

    Pre-flagged conflict screen

    Preliminary conflict screen against your conflicts database surfaces material conflicts before the matter is routed.

    04

    Attorney-scannable transcript

    One-paragraph matter summary plus key entities, key dates, and confidence scores rendered for 30-second attorney review.

    05

    Eval framework + override loop

    Continuous eval against monthly-refreshed regression set; every attorney override fed back into the model regression suite.

Architecture, in plain English

Foundation model selection: Claude for the long-context classification reasoning, GPT for structured-output and tool-use heavy paths (entity extraction, conflict-database queries). Explicit fallback chain documented in the architecture decision record. Retrieval: hybrid sparse+dense over your matter history and practice-group taxonomy, with metadata-aware ranking. Agent orchestration: deterministic state machine — classify → conflict screen → summarize → route → human-in-the-loop checkpoint — rather than open-ended agent exploration. Observability: AppRocket eval stack with weekly attorney-graded regression runs and per-practice-group threshold tuning.

Intake Triage — frequently asked

Ready to deploy intake triage?

Start with the AI Readiness Audit — 2 weeks, $15K, founder-led. We will scope this surface (and any others) against your firm specifically.

Start the audit