Meet The Panel
In our recent webinar, we sat down with three credit union leaders who have already made the leap. Stephanie Walker, Chief Operations Officer at Georgia United Credit Union, oversees a lean collections team at a $2.5 billion institution and was responsible for building the internal business case for AI. Vanity Tulloch, Director of Lending Operations at Georgia United, ran the head-to-head performance analysis comparing their AI agent to their human collectors. And Tenisha Howard, Manager of Collections at Center Parc Credit Union, a $3+ billion institution, has been deploying and refining their AI agent long enough to watch her collectors go from skeptics to advocates.
All three were first-time AI deployers. In every case, this was their credit union’s first member-facing AI deployment. What they learned is worth paying attention to.
Meet Emma and Donna
At Center Parc Credit Union, she goes by Emma. At Georgia United Credit Union, she goes by Donna. They are AI member consultants, virtual agents that outbound-call delinquent members, verify identity, explain the situation, and work toward a resolution.
In our live webinar panel, attendees had the chance to hear actual call recordings from both deployments. In one, Center Parc’s Emma reaches a member 12 days past due on a personal loan, verifies his identity, asks open-ended questions about his situation (a temporary reduction in hours at work), negotiates a payment plan, and sends him a secure payment link, all in under three minutes.
In the other, Georgia United’s Donna navigates automated call screening at the member’s home before connecting, verifies identity, discusses a $710 overdrawn share balance, and confirms a payment plan tied to the member’s upcoming direct deposit.
Both calls are natural, empathetic, and fully compliant. Neither sounds robotic. The member in the first call doesn’t push back or ask to speak to a human. The member in the second confirms the plan and hangs up satisfied.
“I Thought It Would Just Be Robotic”
Vanity is candid about her initial skepticism. When the AI program was first presented to her nine months before the webinar, her expectations were low.
“I thought this is just going to be robotic. It’s not going to be like a human. And then after hearing it and actually interacting with it, it’s like: what else can it do?”
– Vanity Tulloch, Director of Lending Operations, Georgia United Credit Union
The concern about a robotic experience is by far the most common objection credit unions raise before adopting AI collections. And it’s a reasonable one. But the gap between expectation and reality, in the experience of these panelists, has been dramatic.
Tenisha echoes this, and goes further. When she evaluates Emma’s calls, she compares them not just to a low bar, but to her best human agents.
“This is one of the better representations across the board, including compared to human interaction.”
– Tenisha Howard, Manager of Collections, Center Parc Credit Union
She points to a specific moment in one call where the member said goodbye but then made one more comment. A human collector at the end of a long shift might have already moved on. Emma caught it, responded appropriately, and ended the call professionally.
Compliance: The Unexpected Strength
For collections leaders, one of the most persistent anxieties about any new tool is compliance. The Fair Debt Collection Practices Act, TCPA, and state-level regulations create a minefield. One slip, and the exposure is significant.
What Tenisha found was that AI doesn’t just comply. It makes compliance auditable in a way human calls never could be.
“Not only do you get the conversation, you get the transcript. And not only do you get the transcript, you get the note. I would be extremely comfortable if I were called to court and had to produce those transcripts.”
– Tenisha Howard, Manager of Collections, Center Parc Credit Union
Every call is recorded, transcribed, and logged automatically. The AI follows verification protocols exactly, every time. If a credit union’s standard is to verify date of birth and the last four digits of a Social Security number, that is what happens on call one and on call ten thousand. No exceptions, no shortcuts.
When members bring up issues that fall outside the AI’s scope (bankruptcy filings, third-party calls, cease-and-desist situations) the system recognizes the trigger language and escalates appropriately. Georgia United reports a 4% escalation rate, and the majority of those are simply members who prefer to speak to a live person, not compliance issues.
On TCPA consent, the simplest answer from the panelists: include appropriate language in your membership agreement. Legal counsel from both credit unions and from Clutch reviewed the consent process before launch. For credit unions that originate loans digitally, consent language can also be incorporated into loan disclosures at origination.
What Happened to the Collectors
If there is one theme that surprised all three panelists, it’s this: the collectors who were most worried about being replaced became the biggest advocates for the AI agent.
The mechanism is straightforward. Before AI, collectors spent significant time on outreach that produced no conversation. They dialed, they left voicemails, they moved on. With Emma handling that workload, the dynamic flipped. Now the escalation queue is populated with members who have already expressed a willingness to engage.
“We went from ‘Is Emma coming to take our jobs?’ to my collectors almost fighting over Emma. When a member escalates, Emma sends us an email. We reach back out and we know: this person is available, this person is willing to have a conversation.”
– Tenisha Howard, Manager of Collections, Center Parc Credit Union
The collectors who had been spending the bulk of their day on unproductive dials are now spending it on meaningful conversations, complex arrangements, and members in genuine hardship. The work became more interesting, more impactful, and by most accounts more satisfying.
Tenisha describes it in terms of the prestige of the role: “I’m not just a person taking calls through the queue. They may be important, they may not. Now I’m a subject matter expert.” She reports that some collectors saw their performance double or triple once Emma was in place.
How Members Have Responded
The assumption that members, especially long-tenured credit union members, would reject an AI caller has not borne out.
Center Parc has been serving members for over 100 years. The panelists expected resistance from older members in particular. Instead, Tenisha reports that Emma performed well across all demographics. Escalations have been driven by situation complexity, not by members objecting to speaking with an AI.
Stephanie adds a nuance worth considering: some members may actually prefer AI for these calls. Collections conversations can be embarrassing. A member who is behind on a payment may be more willing to have an honest conversation with an AI, where the social discomfort is lower, than with a human collector. That dynamic, while not fully measured yet, is emerging as a possible additional benefit.
The C-suite concern that members would refuse to give personal information (such as the last four digits of their Social Security number) to an AI agent also has not materialized. As Stephanie notes: members who are used to receiving collections calls are accustomed to verification steps. They complete them.
Your Members Deserve Better Collections Experiences
So does your team. Meet Emma.
In Part 3, we’ll look at what this means for the economics of your collections department, how two credit unions made the build-vs-outsource decision, and the practical steps to get started with your first AI deployment.
In case you missed it, catch up on Part 1.