Picture two delinquent accounts on the same portfolio: One belongs to a borrower who opens every text within the hour, clicks through to a payment portal, and sets up a plan without ever talking to a person.
The other belongs to someone who hasn't touched the app since he downloaded it, doesn't answer numbers he doesn't recognize because he's in the middle of a family emergency, and is fully willing to pay, but only once he can talk it through with someone on the phone.
Maybe it's a language barrier that makes a text thread useless. Maybe it's a rural area with spotty data. Maybe it's just someone who has never trusted an automated system with money and isn't about to start now.
Most AI-driven collections tools are excellent at the first account and quietly bad at the second. Not because the technology can't handle a phone call, but because almost nobody built it to notice him in the first place.
We call this the orphaned borrower. Not lost, not unwilling, just invisible to a system that was tuned on the borrowers who were always going to be easy: comfortable online, quick to self-serve, generating the fast wins that make a pilot program look good in a QBR.
Everyone else gets whatever's left over in the process, a lower contact priority, a generic script, a slower path to a human, because nothing in the stack was built to flag that this account needed a different approach from day one.
The scale of this isn't a hunch. TransUnion consumer research found nearly 80 percent of consumers say phone contact is important when dealing with a business, and specifically for personal matters, the kind debt falls squarely into, 64 percent said they'd prefer to handle it by phone. And a widely cited McKinsey study of delinquent credit card customers found that when issuers matched traditional-preferring customers with phone and letter contact instead of defaulting them into digital outreach, results improved by 17 percent, direct evidence that channel-matching, not channel-availability, is what drives the outcome.
The problem isn't that AI can't hand off to a human. It's that most systems can't hand off well because they weren't built to.
Here's the part that gets missed in most AI vendor conversations: Bolting a chatbot or scoring engine onto an existing call center isn't the same as integrating them. When the two run as separate systems, connected by a nightly sync or nothing at all, the AI has no reliable way to say "this one needs a person, now" before the borrower gives up. And when an account does finally reach an agent, that agent is starting from zero. No visibility into what's already been tried, what was already offered, what the borrower already said. The borrower ends up living through both failures at once: slow to reach a human, and then treated like a stranger when they finally do.
That gap is exactly what we designed AiRE to close. The same model that scores who's likely to convert through a text or a portal is also reading for the opposite signal - hardship language, repeated non-response, an account history too complex for a script, and routing those borrowers to a live agent before they go quiet. When the handoff happens, the agent isn't picking up cold. They're picking up mid-conversation, with everything the borrower has already been through sitting right in front of them.
What changes when the handoff actually works
We've seen roll rate improvements in the 21 to 39 percent range on programs built this way, and the biggest gains consistently show up in the population everyone else is quietly failing: the borrowers who were being run through a generic cadence regardless of whether a bot or a person was ever the right first move for them. None of that lift comes from doing less with human agents. It comes from making sure the right accounts reach them sooner, with the groundwork already laid.
The industry has spent a lot of energy on how much of collections AI should take over. We think the better question is who your current system is quietly failing right now, and whether it's even built to notice.
If you want a working answer to that question for your own operation, get in touch with us here for an assessment of your current stack and how much of a lift you can expect to see with AiRE.
Scott Carter is Chief Strategy Officer at Bounce AI, bringing extensive financial services leadership across some of the industry's most respected institutions and emerging fintech companies, including Capital One, American Express, Barclaycard, TransUnion, Affirm, and Upstart. His experience spanning enterprise-scale organizations and high-growth fintechs gives him a distinctive perspective on how recovery operations must evolve to meet the demands of modern lending.