ResolveAI — Intelligent Customer Support Suite

A SaaS company with 45,000 active users was spending $180K/month on a 22-person support team handling repetitive inquiries about billing, account management, and feature how-tos. Average response time was 4.2 hours. ResolveAI automated the entire tier-1 layer — the AI handles FAQs, account lookups, and guided troubleshooting instantly, and escalates to a human agent only when confidence is low or sentiment turns negative.

• 1

  The knowledge base had 2,400 articles with overlapping, sometimes contradictory content — raw retrieval returned confusing answers.

• 2

  Customers contacted support via chat widget, email, and an in-app 'Help' button — each channel had different context available.

• 3

  Sentiment was a critical signal: a frustrated customer mid-cancellation needed a human immediately, not another bot reply.

• 4

  Agent handoff had to transfer the full conversation context so customers never had to repeat themselves.

We built a unified intent + retrieval pipeline. Incoming messages are first classified by an intent model (fine-tuned on 18 months of historical tickets) into 87 intent categories. For self-service intents, a RAG pipeline retrieves from the de-duplicated knowledge base (cleaned and re-embedded during onboarding) and generates a GPT-4o response grounded in cited sources. For account-specific intents (billing, usage), the AI calls authenticated REST APIs to pull live data and weaves it into the response. A real-time sentiment classifier runs on every message; a rolling score below threshold instantly opens a human escalation ticket with the full transcript pre-loaded. All three channels (chat, email, voice-via-Twilio) share the same pipeline with channel-specific formatting adapters.