Voice
Samantha
0:00/1:34
Samantha is Upstream Group’s AI agent built for late-stage debt collection, reaching out to customers with settlement options and securing promises to pay at scale. While the technology performed well, low repayment rates showed the market wasn’t ready — offering important lessons and paving the way for a new, refined approach.
Year :
2025
Industry :
Debt collection
Client :
Upstream Group
Project Duration :
8 weeks
Problem :
The Upstream Group engaged us to explore how AI could support their late-stage debt collection process. Their teams were dealing with prescribed debt — cases that had been on the books for a long time, with low repayment rates. Traditionally, this meant large amounts of manual outbound calling, offering settlements and discounts in the hope of securing some form of payment. It was costly, resource-heavy work, with limited returns.
Solution :
We introduced Samantha, an AI agent designed to manage this difficult outreach at scale. Her role was to:
Call out to the entire late-stage debt book.
Engage customers in conversation to secure promises to pay.
Offer settlement options, discounts, and payment terms tailored to the situation.
Aim to convert more of these challenging accounts into repayments without increasing operational costs.
Challenge :
On the voice side, Samantha performed strongly. She managed to connect with customers, hold natural conversations, and secure a significant number of promises to pay. The challenge, however, was the market itself. Debt collection at this stage is notoriously difficult, and while promises were made, very few translated into actual payments. This highlighted a truth that goes beyond the technology: sometimes the industry conditions simply aren’t ready for AI at scale, even when the agent does her job well.
Evolution :
While the first version of Samantha didn’t achieve the outcomes we hoped for, the project was far from a failure. It proved that:
AI can handle the mechanics of late-stage debt collection reliably.
The main barriers to success lay in customer behaviour and industry readiness, not the agent’s performance.
Being transparent about learnings is as important as celebrating wins.
We remain on strong terms with the client and are already exploring a new approach that better fits both the technology and the realities of the debt collection market. Samantha represents not just an AI agent, but also a lesson in iteration — showing that innovation sometimes means adjusting strategy, not just scaling success.
More Agents
Voice
Samantha
0:00/1:34
Samantha is Upstream Group’s AI agent built for late-stage debt collection, reaching out to customers with settlement options and securing promises to pay at scale. While the technology performed well, low repayment rates showed the market wasn’t ready — offering important lessons and paving the way for a new, refined approach.
Year :
2025
Industry :
Debt collection
Client :
Upstream Group
Project Duration :
8 weeks
Problem :
The Upstream Group engaged us to explore how AI could support their late-stage debt collection process. Their teams were dealing with prescribed debt — cases that had been on the books for a long time, with low repayment rates. Traditionally, this meant large amounts of manual outbound calling, offering settlements and discounts in the hope of securing some form of payment. It was costly, resource-heavy work, with limited returns.
Solution :
We introduced Samantha, an AI agent designed to manage this difficult outreach at scale. Her role was to:
Call out to the entire late-stage debt book.
Engage customers in conversation to secure promises to pay.
Offer settlement options, discounts, and payment terms tailored to the situation.
Aim to convert more of these challenging accounts into repayments without increasing operational costs.
Challenge :
On the voice side, Samantha performed strongly. She managed to connect with customers, hold natural conversations, and secure a significant number of promises to pay. The challenge, however, was the market itself. Debt collection at this stage is notoriously difficult, and while promises were made, very few translated into actual payments. This highlighted a truth that goes beyond the technology: sometimes the industry conditions simply aren’t ready for AI at scale, even when the agent does her job well.
Evolution :
While the first version of Samantha didn’t achieve the outcomes we hoped for, the project was far from a failure. It proved that:
AI can handle the mechanics of late-stage debt collection reliably.
The main barriers to success lay in customer behaviour and industry readiness, not the agent’s performance.
Being transparent about learnings is as important as celebrating wins.
We remain on strong terms with the client and are already exploring a new approach that better fits both the technology and the realities of the debt collection market. Samantha represents not just an AI agent, but also a lesson in iteration — showing that innovation sometimes means adjusting strategy, not just scaling success.
More Agents
Voice
Samantha
0:00/1:34
Samantha is Upstream Group’s AI agent built for late-stage debt collection, reaching out to customers with settlement options and securing promises to pay at scale. While the technology performed well, low repayment rates showed the market wasn’t ready — offering important lessons and paving the way for a new, refined approach.
Year :
2025
Industry :
Debt collection
Client :
Upstream Group
Project Duration :
8 weeks
Problem :
The Upstream Group engaged us to explore how AI could support their late-stage debt collection process. Their teams were dealing with prescribed debt — cases that had been on the books for a long time, with low repayment rates. Traditionally, this meant large amounts of manual outbound calling, offering settlements and discounts in the hope of securing some form of payment. It was costly, resource-heavy work, with limited returns.
Solution :
We introduced Samantha, an AI agent designed to manage this difficult outreach at scale. Her role was to:
Call out to the entire late-stage debt book.
Engage customers in conversation to secure promises to pay.
Offer settlement options, discounts, and payment terms tailored to the situation.
Aim to convert more of these challenging accounts into repayments without increasing operational costs.
Challenge :
On the voice side, Samantha performed strongly. She managed to connect with customers, hold natural conversations, and secure a significant number of promises to pay. The challenge, however, was the market itself. Debt collection at this stage is notoriously difficult, and while promises were made, very few translated into actual payments. This highlighted a truth that goes beyond the technology: sometimes the industry conditions simply aren’t ready for AI at scale, even when the agent does her job well.
Evolution :
While the first version of Samantha didn’t achieve the outcomes we hoped for, the project was far from a failure. It proved that:
AI can handle the mechanics of late-stage debt collection reliably.
The main barriers to success lay in customer behaviour and industry readiness, not the agent’s performance.
Being transparent about learnings is as important as celebrating wins.
We remain on strong terms with the client and are already exploring a new approach that better fits both the technology and the realities of the debt collection market. Samantha represents not just an AI agent, but also a lesson in iteration — showing that innovation sometimes means adjusting strategy, not just scaling success.