The Problem
HireRight’s customer service team was handling thousands of inbound support calls and emails daily — most of them repetitive queries around background check status, document submission, and process timelines. The volume was growing, costs were rising, and the team was stretched thin.
The opportunity was clear: an AI-powered chatbot could deflect the majority of these queries, reduce load on human agents, and improve response times for customers. The challenge was doing this in a regulated environment where accuracy, compliance, and trust are non-negotiable.
The Approach
Rather than jumping to implementation, I started with deep discovery — shadowing customer service reps, analyzing call recordings, and mapping the top 20 query types by volume and resolution complexity.
The key insight: 80% of queries fell into 6 repeatable intent buckets. That meant a well-scoped bot could handle the vast majority without needing to be a general-purpose AI.
I partnered with engineering, legal, and Customer Service to define the guardrails: what the bot could and couldn’t say, how to handle escalations, and what compliance language had to be present in every interaction.
The Outcome
- 300K+ conversations handled in the first year of operation
- ~95% deflection rate — fewer than 1 in 20 conversations required human handoff
- ~88% CSAT — customers rated the bot experience positively
- $75K+ in annual cost savings from reduced headcount and tooling consolidation
- Currently expanding to Voice bot capabilities
What I Learned
Building AI products in regulated environments requires more constraint-setting upfront than in typical consumer contexts. The compliance review cycle added 6 weeks to our timeline — but it also forced us to design better escalation paths, which ultimately made the bot safer and more trustworthy.
The highest-impact PM work wasn’t the AI decisions — it was getting legal, Customer Service, and engineering aligned on what “good” looked like before anyone wrote a line of code.