Lead Product Manager, Voice AI

Salient builds AI agents for regulated financial services, automating loan servicing, compliance, collections, recovery, insurance claims, and disputes for banks, captives, and specialty lenders. The Lead Product Manager, Voice AI (omni-channel) will own what the agent can do, how well it does it, and how quickly a new lender can get it live. This is a hands-on role involving listening to call recordings, sitting with servicing teams, writing specs, and defining what good means for an agent that talks to people about their money. The team typically works around 60 hours per week, beginning at 8:00 AM, with four days spent collaborating in person at the San Francisco office.

Responsibilities

Own the roadmap and strategy for the agent across voice, text, email, and chat, including making a conversation that starts in one channel finish cleanly in another; define agent quality metrics (containment, resolution rate, escalation, handle time, CSAT, error and hallucination rates), build the evaluation loop, and move the numbers; turn the servicing workflow catalog (payments, payoffs, hardship, collections, title, and insurance) from bespoke deployments into configurable product; reduce time-to-live so new lenders go from signed to in-production faster every quarter; own guardrails and compliance posture in the product, including disclosures, audit trails, human handoff thresholds, and design decisions that keep the company inside FDCPA, TCPA, UDAAP, and state servicing rules; set the bar for discovery, specs, and instrumentation that the product org will inherit

Requirements

Shipped LLM-powered products to real users, not demos; can reason about prompting, evals, retrieval, tool use, and latency/cost tradeoffs, and knows where models break; owns outcomes, not artifacts, and moves fast while deciding with incomplete information; listens to hundreds of calls and sits with ops teams to understand the real workflow; rigorous about measurement and can tell a model problem from a product problem from a data problem; can read a trace, debate an architecture decision, and write a spec that doesn't need translating; knows how to design a system that's allowed to be wrong sometimes without being catastrophic; can say no to a large customer and keep the relationship; writes and presents clearly enough to stand in front of a lender's risk committee; Voice, IVR, contact center, or conversational AI experience, including telephony constraints, barge-in, latency budgets, transcription error, and call QA

Nice to have

Background in lending, servicing, collections, or adjacent financial operations; experience hiring or mentoring product managers

Benefits

Medical, dental, and vision coverage; 401(k); catered lunches