One AI Agent vs. Four Human Collectors: The Math
Vanity conducted a three-month analysis comparing Emma’s outbound performance at Georgia United against the performance of the credit union’s human collections team. The results shifted her thinking.
On promise-to-pay productivity, Emma matched the human collectors. The same proportion of engaged members made a payment commitment. On voicemail reach, the numbers were comparable. But on raw outreach volume, the gap was significant.
“The outreach performance was mind-blowing. One AI collector was able to make twice as many calls compared to the four or five collectors we have.”
– Vanity Tulloch, Director of Lending Operations, Georgia United Credit Union
Equally notable is what happens at the top of the efficiency curve. Emma doesn’t slow down. She doesn’t take breaks, doesn’t call in sick, and doesn’t vary in energy or tone throughout the day. In practice, at some credit unions, Emma is so efficient that she runs out of delinquent accounts to call before the end of the day. Compliance and operational guardrails, rather than capacity, become the binding constraint.
Rethinking the Staffing Model
The productivity data has implications for how credit unions think about hiring and headcount in collections.
Georgia United entered 2025 facing a familiar problem: rising delinquencies, rising charge-offs, and a strong internal case to add a collector and a foreclosure specialist. The foreclosure role was filled, because that work is complex, relationship-intensive, and genuinely requires human judgment. But the additional collector position was reconsidered.
“Because we have this tool, we were able to make a different decision. We hired someone with a technical background, someone who can do process improvement across lending, loan servicing, and collections, and who can also help us build out more AI use cases.”
– Stephanie Walker, COO, Georgia United Credit Union
The decision reflects a broader shift in how AI changes the ROI conversation for staffing. Instead of hiring to add capacity to low-skill, high-volume work, Georgia United invested in someone who could improve the system itself, extending the impact of every human and AI agent already in place.
For credit unions thinking about outsourcing as an alternative, Stephanie is candid about the tradeoffs. Third-party collection agencies are expensive for early-stage delinquencies, and they introduce significant oversight burden. When a member self-cures within 30 days (which Georgia United found was common), the outsourcer’s work generated cost but no incremental value. With AI, that cost structure changes: the credit union pays for outreach they control, at a fraction of the cost, with full audit capability.
The Implementation Is Faster Than You Think
One of the most consistent surprises among the panelists was the speed of deployment.
Stephanie committed to her board in March or April that the credit union would have an AI collections solution by year-end. They were live in less than 90 days.
The technical foundation is simpler than it might appear. The AI agent doesn’t require replacing existing collections software. It works alongside it. Clutch, for example, integrates with core systems to pull real-time account data (so the agent knows whether a member actually has an outstanding balance before calling) and writes back to collection platforms to log notes, promises to pay, and escalations. The AI works with whatever operational infrastructure the credit union already has in place.
Vanity’s advice for anyone going through implementation: give yourself grace, and lean on your team.
“The people doing the job every day are the ones who know how to develop the tool to perform the way your credit union needs it to. Involving your team in that process is also rewarding for them.”
– Vanity Tulloch, Director of Lending Operations, Georgia United Credit Union
Teams that participated in designing how the AI agent would sound, what it would say, and what situations it would handle became invested in its success rather than resistant to its existence.
Start Small. Grow Fast.
Neither credit union started with a fully deployed, end-to-end AI collections operation. Both started with one use case, narrowly defined and low-risk.
Georgia United began with negative share balances: accounts that were overdrawn, where the downside risk of an AI call was minimal. If the member declined to engage, nothing was lost. If the member engaged, the credit union gained a contact it otherwise couldn’t have made.
Center Parc began with early-stage delinquencies, where the goal was simply to be present in the member’s first days of delinquency rather than waiting until a collector had bandwidth.
In both cases, the question quickly shifted from whether to expand to how fast. Tenisha notes that she’s already pushing the Clutch team on a long list of additional use cases. Stephanie says her executives are now asking what else the AI can handle rather than whether they should have started.
Common expansion paths include: later-stage delinquencies, proactive outreach to members at risk before delinquency begins, post-charge-off recovery, and eventually multilingual outreach (Spanish-language support is in active development).
The Case for Acting Now
Tenisha ends with the sentiment that perhaps best captures the opportunity:
“I wish I had Emma a couple of years ago. If you’re passionate about collections, you’re passionate about helping people. It’s great to be able to help 500 people. But what if you could give that same assistance to 50,000?”
– Tenisha Howard, Manager of Collections, Center Parc Credit Union
The economics of AI collections have reached the point where the cost of not adopting is larger than the cost of getting started. Delinquencies are rising across the industry. Staffing collections at scale with humans is expensive and operationally volatile. And the member experience delivered by AI is, by the account of those who’ve deployed it, as good as or better than what most collections teams deliver today.
The path forward is not complicated. Pick a use case. Define it narrowly. Involve your team in building it. Measure the results. Then grow from there.
The credit unions that move first will build institutional knowledge that compounds. The ones that wait will be playing catch-up.
In case you missed it, catch up on Part 1 and Part 2.
About Clutch
Clutch is a consumer loan and deposit account origination platform powering digital and staff-led experiences for 175+ credit unions across the country. Clutch has expanded into AI-powered collections, helping credit unions reduce delinquency, improve member experience, and make their collections teams more effective.