PM - AI First Helpdesk
Manifest Ai
Software Engineering, Data Science
Bengaluru, Karnataka, India
Posted on Feb 14, 2026
What You’ll Build
AI as Primary Resolver
→ High confidence, low risk? AI auto-resolves without human.
→ Medium confidence? AI drafts, agent approves.
→ Low confidence or high risk? AI escalates with full context.
→ Angry customer? Human takeover with AI briefing.
You’ll define the confidence thresholds, risk parameters, and escalation triggers.
Core Capabilities You’ll Own
- Contextual intelligence that pulls order history, past conversations, product specs, and customer sentiment before responding
- Living knowledge system with product info, policies, and issue-to-resolution mapping
- AI-guided agent mode: pre-diagnosis, suggested replies, policy guardrails, next-best-action
- Intent and sentiment detection that adapts tone automatically
- Omni channel context stitching — one customer, one conversation, across all channels
- Feedback loops that make AI smarter weekly
You Should Have
- 2+ years in product, ideally in support / CX / helpdesk tools
- Shipped AI features where AI made decisions, not just suggestions
- Comfort with LLMs, RAG, prompt engineering, confidence scoring
- Strong judgment on when AI adds value vs. creates friction
- Ability to think deeply about failure modes, not just happy paths
Metrics You’ll Own
→ AI Auto-Resolution Rate
→ AI Confidence Accuracy
→ Human Takeover Rate
→ CSAT (AI vs Human)
→ Knowledge Gap Rate