See it live

TRACE's live observability.

TRACE's live observability — per-partner scorecards, response-time distribution, automation rate, and token-cost tracking across the agent pipeline.

01 · Capabilities

What's inside.

Intercom Fin runs on Intercom's data. Assist runs on yours — your tenancy, your stack, your CRM as the source of truth. Built for ops teams that can't move their data.

Multi-channel — WhatsApp Business · voice (IVR + TTS) · web chat · email · Slack, one conversation across channels.
4-agent routing — Identifier classifies. Support resolves L1. Expert handles edge cases. HIL escalates only when needed.
T0 pre-filter — a rule-based deflection layer handles ~50% of traffic at $0 LLM cost.
Tiered LLM routing — T1 self-hosted OSS for 80% · T2 mid-cloud for 15% · T3 frontier for 5%.
PII redaction pre-LLM — Aadhaar, PAN, mobile, account numbers stripped before any model call.
CRM / ITSM connectors — Freshdesk, Zendesk, ServiceNow, HubSpot, Salesforce. Bidirectional.
Immutable audit trail — every external call, model invocation, and human decision written to a WORM store. DPDP-compliant.

02 · Architecture

Channels in. Agents in the middle. Systems out.

Inbound channels feed an intent + sentiment router. Routes hit the multi-agent core (Identifier → Support → Expert → HIL). The tools layer reaches your CRM, ITSM, and FAQ knowledge base. Outbound delivery fans across SMS, push, email, voice callback, and webhook.

Flow

whatsapp · voice · web · email · slack
  ↓
identifier (intent + sentiment)
  ↓
support → expert → hil
  ↓
crm · itsm · knowledge · webhooks

03 · Live in production

TRACE at GyanDhan.

support

TRACE at GyanDhan

Multi-agent system handling partner ops across 2,000+ active WhatsApp groups. Ops load went from 2.5 hours per day to 15 minutes per day.

Read the case →

What we won't do

Send your PII to a third-party model — Aadhaar, PAN, mobile, and account numbers are stripped before any external call.
Ship without a CLEAR scorecard — every deployment gets a baseline across five dimensions before it goes live.
Move your data out of your tenancy — self-hosted models run in your cloud; ~95% of calls never leave it.

Start with a Diagnose.

Two to three weeks. $6K–$12K. A written scorecard with topology recommendation, cost ranges, and remediation plan. No commitment to build.

Book a diagnosis

Customer ops across every channel.

Measured in production.