Maximising Debt Collections Performance with AI

Taking a digital approach to debt collection is starting to become more mainstream across the industry. Both creditors and customers benefit from greater use of digital channels, with customer self-service helping to drive material improvements in debt collections performance.

Of course, it should be no surprise that customers prefer digital channels over traditional means of contact. A study by Swiss multinational UBS, published in January of this year, found that 52 percent of all consumer financial transactions now occur online, making digital the primary channel for modern banking. Just two years ago that figure stood at 33 percent, so the transition is happening fast. Debt collections methods must also adapt to these new consumer expectations.

As in the wider banking environment, consumers do not want to talk to call centre agents about their debts. They prefer to conduct their financial affairs online.

At the same time, the Financial Conduct Authority (FCA) has repeatedly underlined its requirement for financial services companies to treat their customers fairly, ensure vulnerable customers are handled with respect, and provide auditable evidence of compliance with these requirements.

A digital-first debt collections strategy, properly implemented, can meet the priority requirements of both the consumer and the regulator. But while the case for digital collections is now clear, the question remains: what are the most effective debt collections strategies within the digital channel? The challenge now is to figure out how to maximise the debt collections performance of the digital channel.


Automation, and particularly Artificial Intelligence and Machine Learning technologies (henceforth AI), have been offered as vague panaceas to this challenge. But the problem is that this argument often boils down to the reductionist doctrine of ‘replace the call centre staff with robots’. But this misses the point: AI is much more effective as a technology that augments human capability in complex tasks. The chatbot may be an effective, nonintrusive reminder that the consumer now prefers, but it must also seamlessly handoff to a human agent when the consumer has a more complex query or situation.

AI offers the basis for building a whole new approach to collections; one that is fundamentally different from the methods used in the past. The technology must be applied properly to unlock its promise, and in this paper we explore how companies can apply AI to boost their collections performance.


The Challenge

Debt collections professionals are currently working in a business environment governed by two broad strategic drivers:


A near-term economic environment with increasing levels of consumer debt, higher interest rates, and higher rates of default. These factors in turn will put an even greater focus on the commercial performance of debt collections.


A principles-based regulatory regime which mandates customer-centric conduct, and under which all processes are judged in terms of whether they lead to fair and appropriate customer outcomes. These outcomes may be focused on getting customers out of arrears, or on the receipt of appropriate levels of forbearance. In considering how best to reach these outcomes, there is a great emphasis on recognising vulnerability, and in segmenting customer journeys as far as possible, with the end goal of collections.

These macro drivers are compelling debt collections managers to make even more efficient use of their human resources. Economic conditions will require stronger performance from collections operations to maintain even the current levels of return. Throwing more headcount at the problem is an exercise of diminishing returns: the cost increases at a faster rate than the collections performance per head. At the same time, more staff means more training and longer call times to ensure regulatory compliance.

Call centre-only debt collections strategies will not be enough. Firms must boost debt collections performance, while reducing cost and delivering a more carefully monitored debt collections process.

A new solution to this challenge is required, and that solution can leverage AI to meet the dual macro drivers behind this change.

The Solution:

In our March 2017 whitepaper, The Best of Both Worlds, we advocated the development of a broader channel strategy for debt collections which integrated digital self-serve as a central component, while still retaining elements of traditional call and collect methodology. To recap, this paper argued that:

Initiating and maintaining debt collections relationships digitally, rather than by telephone, demonstrates a greater respect for customer privacy and convenience, and fosters a better long-term rapport. This results in a more positive customer experience, as well as a more dependable long-term payment profile.

Offering a non-intrusive, customer-driven point of engagement makes it easier for debtors to engage with creditors earlier in an arrears, or even a prearrears, scenario, helping to avoid the more costly recovery of older debt.

A self-serve digital environment lets customers choose their terms of engagement with the creditor, meaning they can transact at times that suit them and opt for direct human contact if they feel it would benefit them.

At the same time, the interaction data captured by a self-serve system allows creditors to better identify and diagnose vulnerability situations and pinpoint the moments in a customer journey where contact via telephone or other channels would be more appropriate and beneficial.

Similarly, the operation of a platform that automatically records both customer and creditor actions provides a major advantage in terms of auditability, and the evidencing of a compliant customer journey.

We made the case that digital debt collections helps to identify where best to concentrate call centre resources, enabling creditors to achieve a more compliant, commercially efficient operating model that includes the unique benefits of human-to-human interaction.


Now, we are deploying AI technology as a powerful new tool to deploy as part of a digital debt collections strategy. AI can be used to further automate and tailor the digital side of the recovery process. Rather than treat all customers the same, different debt collections strategies are selected to best fit the characteristics of a given customer. The machine learns to optimise its choice of debt collections strategy in an iterative process of automated case review and debt collections performance monitoring. Here’s how:

Step One – Chatbots as a contact calibration strategy

One of the key benefits of a digital debt collections strategy is the ability to use automation to flag points in the customer journey where human interaction can be deployed to best effect.

Although, in a great deal of ordinary debt collections situations, an AI agent can guide a customer through the journey, there are other situations where human intuition and context-based decision making is required from both an efficiency and a compliance standpoint.

Chatbots can play a role in moving contact centres towards greater automation because they combine machine learning with expertise gained from live debt collections agents. Replacement of certain agent functions with chatbot resources offers the following benefits:

  • Customers gain all-hours access to an interactive resource which can use natural language to talk through their options.
  • Customers can self-select preference for a more conversational collections approach, informing creditors as to their likely contact preferences from an early stage in the collections journey.
  • Customers begin their interactions from a position of comfort, as they can discuss their issues without the perceived pressure or judgement of a live human agent.
  • Creditors can automatically harvest information about customer behaviour and circumstances from conversational cues. This means they can flag issues such as vulnerability triggers early on, and highlight the occasions when customers would benefit from interaction with a call centre agent.
  • Where human interaction is identified as necessary, chatbots provide a logical segue into web chat channels, or a natural handoff point to phone contact.

In summary, chatbots represent an excellent frontline resource for connecting digital and traditional channels, and a way to use AI powered automation to effectively concentrate human resource to boost debt collection performance.

Step Two – Champion/Challenger: Machine learning as an iterative design tool

The next question is how best to assemble the debt collections journey that customers are taken through, given the large number of possible paths. What is the optimal path that leverages automation and call centre staff to best advantage?

AI also has a role to play here, based upon the classic champion/challenger model for optimising business process:

  1. To begin with, in addition to the company’s existing customer journey, a number of variant flows are designed by staff. These flows contain alternative refinements in terms of what channel is offered at what stage of the journey, or which options (for example, ‘pay now’, ‘promise to pay’, ‘signpost financial advice’) are presented at which points in the journey.
  2. A digital debt collections platform deploys the “champion” journey and one or more variant journeys, with a machine learning engine diverting customer interactions down each pathway.
  3. Over a period of time, the debt collections system will learn which of the various customer journey variants consistently delivers the best outcome. If a ‘challenger’ is found to outperform the incumbent ‘champion’, then it would take the champion’s place, and the process would begin afresh with fresh challenger models.
  4. Through successive repetitions of this process, the customer journey evolves via a series of iterations to produce a greater and greater proportion of optimal outcomes for customers.
  5. Further refinement is possible as the machine learning engine is trained to correlate characteristics of customer situations with different debt collections pathways. The champion/challenger evolves to a customer-category set of optimal collections strategies.

Based on information detected about the customer’s circumstances (through historical data and chatbot interactions), an AI-based system can filter customers into journey variants that are proven to be more suitable for their detected needs. Every individual customer is led through the collections journey that is statistically most likely to address their specific situation, with prompted options for call centre intervention at points along the way.


The economic and regulatory environments present increasing business challenges for the debt collections industry, and leveraging the digital channel for debt collections is critical to meeting these challenges. Within digital, AI has a critical role to play in maximising debt collections performance. But it is not a ‘silver bullet’, indeed the right way to look at AI is to see that it will augment, not replace, existing debt collections strategies. However the opportunity to leverage AI is clear, and firms should be investigating and experimenting now. Intelligent Environments is deploying machine learning and champion/challenger approaches as solutions to our clients’ debt collections challenges. We are excited by the potential for this technology to deliver significantly improved debt collections outcomes for our clients.

About the Author

Clayton Locke,

Clayton Locke,

Chief Technology Officer

Intelligent Environments

Clayton is responsible for the ongoing development of our software products and delivery excellence of our service operations. He leads our Software Architecture and Service Operations teams, closing the DevOps loop to ensure Intelligent Environments delivers high-performance technology solutions to our clients.

Clayton has over 30 years’ experience in the software development and consulting, where he has led technical teams design and deliver solutions for the financial services and telecommunications sectors.