Scaling Reference Checks with AI: Inside Bright Horizons' Voice Agent Success

Reference checks are often treated as a final step in the hiring process. They're the final compliance hurdle — obligatory, time-consuming, and so often incomplete that many teams have quietly built workarounds just to keep the process moving.

For Bright Horizons, a family services company that provides child care, backup care, and in-home care for families at every stage of life, references aren't optional paperwork. They're state-regulated, they vary in quantity and type by location, and they're the last line of defense before a caregiver enters a classroom with children who sometimes can't speak for themselves.

At IAMPHENOM 2026, Stephanie Worley, VP of Talent Acquisition at Bright Horizons, shared how a single question from a member of her team — "Why can't we use the Voice Agent for references?" — sparked a new use case for AI in hiring. What started as a casual observation became a live deployment, going active the day of the session. It's a story about applying AI where it actually solves a compliance problem, not just a convenience one.

Explore the highlights below!

Why Were Reference Checks Stalling Bright Horizons' Hiring Pipeline?

Bright Horizons requires between two and four professional references for each caregiver hire, and the number varies by state and locale. Their background check vendor completes the majority, but roughly 15% of reference checks fail to get completed through that channel. Sometimes, references don't respond, or newer workers simply don't have an extensive professional history to draw from.

To compensate, Bright Horizons built a dedicated TA coordinator team whose job was chasing down that 15%. It was a practical solution, but coordinators capable of finding and nurturing talent were spending their days on hold, waiting for callbacks that might never come.

The cost of an unresolved reference is real. For a company that must meet strict teacher-to-child ratios, every day a qualified caregiver waits for a reference is a day of lost care capacity. As Worley put it: 'We can't start you if we don't have your references because if licensing walks in that center door, we need to demonstrate that we have your reference.'

Speed without quality was not an acceptable tradeoff. The solution needed to deliver both.

How Can AI Voice Agents Automate Reference Checks?

Phenom's Voice Screening Agent was designed for candidate pre-screening — calling applicants, asking structured questions, and scoring responses against a rubric. But the underlying mechanics are identical to what a reference call required at Bright Horizons: a structured question set, recorded and transcribed, with the ability to flag concerning responses for recruiter review.

The reference version sends an email to the professional reference explaining that they've been listed by the candidate, that they'll be speaking with a virtual agent, that the call is recorded, and that they can schedule at a time that works for them or speak immediately.

The agent (named Joey by Bright Horizons) walks through a predefined question set that Worley's team built based on their existing rubric for caregiver references. Questions and drill-down probes are both configurable by role. If a reference indicates that a candidate's attendance was inconsistent, Joey asks a follow-up for context. Responses are transcribed, summarized, and made available to recruiters on the candidate profile under an evaluations tab, alongside flagged questions and the ability to replay specific moments of the call. A recruiter doesn't have to listen to the full recording to identify whether the reference raised concerns; the system surfaces the flags.

What Results Can AI Deliver in Reference Check Automation?

Worley's team went into this with a clear baseline. They knew exactly how long coordinators spent chasing references each week, how long the post-offer-to-background-check phase typically took, and where time to offer and time to fill stood. That groundwork is what makes the projected impact concrete rather than speculative:

50% reduction in time spent chasing references for completion
3.5 hours saved per week per Talent Acquisition Coordinator
Reduced time to complete compliance screening

"That number probably only has room to go up," Worley said. "If this really works for us and we deliver it at scale, that number can significantly increase." The rationale? References in early childhood education are familiar with the process — they've given references before. When they receive a clear, transparent communication from a known brand explaining exactly what's expected and offering immediate or scheduled participation, the experience is better than an unexpected cold call from a coordinator they don't recognize.

What's Next for Bright Horizons After Automating Reference Checks?

Worley named candidate screening as the next use case, using the Voice Agent not just for references but to screen caregiver candidates directly. For an organization where the people being hired care for infants and children, every application of AI is approached with a deliberate question: where does this create value without compromising the quality of care? Reference checks were an answer that met that test. Candidate screening for teaching roles is next.

The broader lesson Worley offered teams thinking about AI deployment applies well beyond reference checks. "Where is the high value? What are you going to get from this result? And where do you already have strong processes in place that can just be amplified by AI?" The Voice Agent didn't fix a bad reference process at Bright Horizons. It automated a rigorous one — and that distinction, Worley noted, is everything. "If you don't have that, you need that first."

See how Phenom voice agents can work for you. Book a demo

Joanna Keel