Zohreh Razavi
← back to work

SaaS company · 2026

Conversational AI Support Agent

An LLM-powered support agent that resolves customer questions from product knowledge, with guardrails and human handoff.

Client
SaaS company
Role
Project Manager
Timeline
Feb 2026 – Jun 2026
Domain
AI / Customer Support
Approach
Agile
Team size
5–9
Budget
Up to $250K

Objective

Deliver a conversational AI agent that answers customer questions from product documentation and support content, with safe fallbacks and human handoff.

What I did

  • ▸ Defined scope, success metrics, and guardrail requirements with stakeholders
  • ▸ Coordinated retrieval pipeline, prompt design, and evaluation with the AI team
  • ▸ Planned sprints and tracked delivery in the issue tracker
  • ▸ Set up human-in-the-loop handoff and escalation flows
  • ▸ Ran evaluation cycles for accuracy, safety, and tone
  • ▸ Coordinated integration with the existing support platform

Key deliverables

  • ▸ Conversational agent
  • ▸ Retrieval + prompt pipeline
  • ▸ Guardrails & handoff flow
  • ▸ Evaluation dataset & metrics
  • ▸ Support-platform integration
  • ▸ Live deployment

Outcome

Deployed an assistant that handled common questions and routed complex cases to agents, reducing response times.

Tools & tech

  • LLMs
  • RAG
  • Vector DB
  • Node.js
  • REST APIs
← back to work