This was not a turnaround story
For two years, AI in collections has been mostly promise – conference panels and limited trials, rarely a serious lender reporting movement in a metric that touches losses. Suncoast Credit Union did, and from a position of strength rather than desperation: a disciplined text-and-voicemail program and an experienced collections floor were already in place. If an operation this strong still had room to move its loss curve, the constraint was never skill. It was structural.
Anyone who runs a loan book knows how this works. Most members who fall a little behind will catch up – early-stage delinquency cures at high rates, especially once someone actually reaches them. The trouble starts when no one gets to them in time: the longer an account goes uncontacted, the harder it is to collect, and once a balance rolls past 90 days it is close to charge-off. So the losses on a book aren’t decided in the late stages. They’re decided earlier, by how many accounts you keep from rolling that far. And that is a question of capacity, not collector skill – even a strong team can’t reach every account in time.
The pace is the part most operators underestimate. Suncoast went from decision to live in under four weeks through a quick-start program – no IT lift required – and the loss curve moved within the first month Emma was live. The full result described here took nine weeks, a window in which a change of this scale is more often still being scoped than measured.

THE OUTCOME
The personal-loan result – and the mechanism behind it
Roll rate is the metric that turns into losses. A delinquent member rarely owes a single creditor – a card, a car, a personal loan, perhaps rent – and members tend to pay whoever reaches them first with a clear, respectful request. Suncoast’s human team, like every team, worked accounts over days and weeks. Emma reaches every assigned member within hours of assignment. When dollars are scarce and bills compete, the institution that asks first is the one that gets paid first.
That speed is why this shows up where it matters most. The month Emma went live, the personal-loan balance rolling into 30–59 day delinquency fell 34% – from $11.6M to $7.7M – and held there the next month. With less balance entering the pipeline, the downstream buckets are following: 60–89 day balances fell from $6.9M to $4.9M over the same window. This happened on an already well-run book – Emma did not replace a broken process. She did what no human team has the hours to do: reach every account quickly, every time.

MEMBER RECEPTION
Member reception: what more than 40,000 conversations showed
Across more than 40,000 live conversations, member sentiment reads more like service than collections. This is not a survey or a focus group – it is how members actually behaved on the phone.
![Screenshot 2026-10-20 at 24558 PM | Clutch | The Leading Credit Union Digital Origination Platform 54% cooperative tone with examples like 'Yes' or 'sounds good'—direct to a commitment (informational panel).]](https://withclutch.com/wp-content/uploads/2026/10/Screenshot-2026-10-20-at-24558-PM-1024x150.webp)

WHAT THE DATA SHOWS
Three conclusions for a lending leader:
1). Coverage was the constraint – not collector skill
Suncoast’s collectors were excellent. The ceiling was physics: too many accounts, not enough hours. Emma closed the coverage gap, and the loss curve responded.
2). Speed determines who gets repaid first
Reaching a member within hours instead of weeks isn’t incrementally better – it changes who gets paid first out of a limited paycheck. That is where roll-rate improvement comes from.
3). The gains came on top of a strong program
This was not a struggling shop. A well-run, tech-enabled collections operation still had double-digit room to improve – which means most lenders do too.


THE BOTTOM LINE
The question stopped being “how many collectors do we need.” It became “what should they be doing instead?”
See what Emma can do for your members.