From Hype to Execution: AI in Collections at NPL Europe 2026

At this year’s NPL Europe 2026, one thing became immediately clear: the conversation around AI in collections has shifted decisively — from ambition to execution.

While previous years focused on potential, innovation, and future possibilities, this year’s discussions were grounded in a more practical question:

Where is AI already delivering measurable impact — and what is still holding it back?

As part of the panel on Technology & Innovation in Servicing, I had the opportunity to contribute to this discussion alongside industry peers. What emerged was a much more mature, experience-based view of how AI is being adopted across collections and recoveries.

Where AI is actually creating value today

There is growing consensus across the industry that AI is no longer experimental in collections. However, its impact is not evenly distributed across the lifecycle.

The most consistent and measurable gains today are seen in:

• segmentation and prioritisation
• strategy selection
• targeted borrower engagement

These are areas where better decisions translate directly into:

• higher cash collection
• lower cost-to-collect
• improved operational efficiency

Rather than replacing existing processes, AI is enhancing the decision layer — making collections more focused, timely, and effective.

The reality check: data still defines success

Despite the progress, implementation challenges remain very real.

Across multiple discussions and exchanges, two constraints were repeatedly highlighted:

• data quality and availability
• legacy system architecture

Many organisations still operate in environments where data is fragmented, partially unstructured, or difficult to access in real time. In such cases, even the most advanced models cannot deliver meaningful results. AI, in practice, is not limited by algorithms — it is limited by data readiness and system integration.

Rethinking ROI: more than just recovery rates

One of the key themes of the panel was how institutions evaluate the success of AI initiatives.

Traditionally, collections performance has been measured through:

• recovery rates
• cash collected
• cost efficiency

While these remain essential, they no longer capture the full value of AI.

Organisations are increasingly recognising additional dimensions, such as:

• process scalability
• consistency of decision-making
• reduction of manual intervention
• customer experience improvements

This broader perspective reflects a shift: AI is not just improving performance — it is changing how operations are structured and scaled.

From reactive to event-driven operations

Perhaps the most fundamental transformation observed is the move away from agent-driven, reactive workflows toward automated, event-driven processes.

Three years ago, many collections operations relied heavily on:

• manual case selection
• static strategies
• repetitive agent tasks

Today, we increasingly see:

• automated prioritisation of accounts
• dynamic strategy assignment
• trigger-based communication workflows

This does not eliminate the need for human involvement — but it significantly reduces low-value manual work and improves consistency across operations.

The rise of hybrid engagement models

Another strong trend is the evolution of borrower interaction models. Digital channels — including portals, messaging, and conversational AI — are taking over a growing share of communication, particularly in early-stage collections. However, the industry is not moving toward full automation.

Instead, the dominant model is becoming hybrid:

• digital channels handle scale and standardisation
• human agents handle complexity and sensitive cases

This shift is also redefining the role of collections agents, who are increasingly expected to:

• interpret insights
• make informed decisions
• manage complex customer situations

rather than simply execute predefined workflows.

Regulation as a design driver, not a barrier

A recurring topic throughout the event was the role of regulation in AI adoption. Contrary to common perception, regulation is not slowing innovation — it is shaping it.

Requirements around:

• model explainability
• auditability
• fairness and transparency

are pushing organisations to design AI solutions that are not only effective, but also robust and controllable. In the European context especially, governance frameworks are becoming a core component of any AI initiative.

Lessons from the field: what works — and what doesn’t

One of the most valuable aspects of the discussion was the sharing of real-world experiences — including where things did not work as expected.

Common pitfalls include:

• overestimating data quality
• applying models too broadly without proper validation
• misunderstanding customer channel preferences

Successful implementations, on the other hand, tend to share a common approach: Start with clearly defined use cases, ensure data reliability, and scale only after measurable impact is proven. This pragmatic approach is increasingly replacing large-scale, top-down transformation attempts.

Where the industry is heading next

Looking ahead, several trends are becoming more visible:

• Deeper integration of AI into core decision processes, rather than standalone tools
• Increased use of real-time data to support dynamic strategies
• Greater focus on customer-centric engagement models
• Continued evolution of regulatory expectations, particularly around AI governance

At the same time, areas such as litigation optimisation, portfolio valuation, and advanced recovery modelling are expected to see further development, although they are currently less mature than early-stage collections use cases.

Conclusion: a more mature phase of AI adoption

NPL Europe 2026 highlighted a clear shift in the industry. The conversation is no longer about whether AI will transform collections —but about how to implement it effectively, responsibly, and at scale.

The key takeaway is simple: AI delivers value where it is grounded in data, integrated into workflows, and aligned with real operational needs. For organisations that approach it in this way, the opportunity is not just incremental improvement — but a fundamental upgrade in how collections and recoveries are managed.