05 / CUSTOMER SUPPORT
RESOLO
Move past basic chatbots. Deliver tenant-aware AI agents that route confidently and hand off to humans without the seam showing.
- OUTCOME
- Custom orchestration engine with confidence tags and explicit handoff markers. pgvector knowledge layer, Redis-backed escalation, billing tied to plans.
- STACK
- TypeScript · pgvector · Redis
- Confidence as a routing input
- Every classification the orchestration engine makes carries a confidence value, and thresholds for acting, flagging, or escalating live in per-tenant configuration rather than in a prompt. That makes routing policy something an operator can state and tune, and makes threshold changes empirical rather than intuitive.
- Handoffs you can replay
- Transfers between agents are explicit, typed events rather than an emergent property of a conversation. A session reads as a sequence of decisions with their confidence attached, which means a held refund can be explained after the fact instead of investigated. Escalation to a human is a first-class destination, reached deliberately.
- Tenancy down to the billing line
- A pgvector knowledge layer keeps each tenant's context isolated, Redis backs the escalation queue, and usage is attributed per organization so billing follows plan limits automatically. Across 100,000+ interactions the operational question is rarely whether a single answer was good — it is whether the system can account for all of them.
- ROUTING
- CONFIDENCE-TAGGED
- INTERACTIONS
- 100K+
- QUALITY
- 4.9 / 5
CAPABILITIES SHIPPED HERE
- MULTI-AGENT ORCHESTRATION
- Confidence-tagged routing. Explicit handoff markers. Agents that know when to defer — to a specialist, a tool, or a human.
- HUMAN-IN-THE-LOOP
- AI proposes as structured previews. A hardened executor applies — with validation, encryption, and audit trails throughout.