Collections and Recoveries: Are We Ready for What Comes Next?

Why enterprise debt collection management needs better data, earlier intervention, stronger agency control, and practical AI decisioning

An EXUS panel discussion in Indonesia recently brought together collections leaders, lenders, and technology specialists to discuss how collections is changing.

The conversation was local, but the challenges were not.

What surfaced in Indonesia will feel familiar to collections and recoveries teams in many markets: poor data quality, difficult customer contactability, outsourced agency control, early intervention, segmentation, AI, and the question of whether organisations are investing in collections capability before pressure arrives. If you work in collections, these are not abstract themes. They are the issues already sitting inside the operating model.

Rather than treating the panel as a one-market discussion, it is worth asking a broader question:

What does this tell us about the challenges facing collections and recoveries teams globally?

Seven themes stand out:

1. The move from reactive to predictive collections
   Are organisations still waiting too long before they act?
2. Data is still the starting point
   Can AI really help if the underlying customer data is incomplete, outdated, or disconnected?
3. Outsourcing collections does not mean outsourcing responsibility
   Are lenders still in control once accounts are passed to third parties?
4. Segmentation still matters, but it needs to evolve
   Are customers still being grouped too broadly when their circumstances and behaviours are different?
5. Contactability is still a major collections problem
   What happens when the strategy is right, but the customer cannot be reached?
6. AI needs to work in the real world of collections
   Can AI support the practical realities of debt collection management rather than simply automate activity?
7. Build or buy? Ask a better question
   Is the issue really whether organisations can build collections technology, or whether they can keep improving it?

Underlying all of this is one bigger point: the best time to invest in collections capability is before the pressure arrives, not after arrears have already started to rise.

The Move from Reactive to Predictive Collections

For many organisations, collections still starts too late. A customer misses a payment, the account rolls into arrears, and only then does the collections process begin. This is the traditional reactive model and, to be fair, it is the model many collections operations have been built around for years.

But the discussion in Indonesia highlighted something important. Technology is now making it easier for collections teams to move from a reactive model to something more proactive, predictive, and customer-specific. That sounds positive, but it needs discipline.

Predictive collections does not simply mean contacting customers earlier and more often. If that is all we do, we risk creating more noise, more complaints, and more customer dissatisfaction. The real opportunity is to use data, analytics, and AI-powered decisioning to understand which customers need support, which customers are likely to self-cure, which customers are becoming harder to reach, and which customers require a more tailored intervention.

In other words, the question is not just: “Can we contact this customer earlier?”

The better question is: “What is the right action for this customer, at this point in time, based on what we know?”

That is where technology can really help. Not by replacing the judgement of collections professionals, but by giving them better insight, better timing, and better decision support.

Data Is Still the Starting Point

One of the most common frustrations in collections is that the data is not good enough. This is not a new issue nor an especially exciting one, but it matters. Many institutions still have customer data sitting across multiple systems - core banking systems, CRM platforms, collections systems, agency systems, and customer communication tools may all hold slightly different versions of the truth. Contact details may be missing, duplicated, out of date, or simply wrong. That becomes a real problem when organisations are trying to build a more intelligent collections operation.

If an SMS bounces, a letter is returned, a phone number no longer works, or an email address is missing, the collections strategy immediately becomes weaker. If those issues are only identified once the customer is already in arrears, the organisation is already on the back foot.

This is where collections teams need to be honest with themselves. It is tempting to start the conversation with AI as it sounds modern, exciting, and transformational. But if the underlying customer data is poor, disconnected, or unreliable, AI will not magically solve the problem. Good collections still starts with good data, and that means auditing contact data, understanding where data quality issues sit, identifying missing fields, validating communication channels, and building processes that keep customer information accurate throughout the customer lifecycle. This may not sound as exciting as AI, but for many lenders, banks, and financial institutions, it is where meaningful debt collection management improvement needs to begin.

Outsourcing Collections Does Not Mean Outsourcing Responsibility

The panel also discussed the use of third-party collections agencies, which is particularly relevant in markets where lenders rely heavily on outsourced activity. There is nothing wrong with outsourcing. In many cases, it is necessary. Agencies can provide scale, reach, specialist capability, and additional capacity when internal teams are under pressure.

But outsourcing collections does not mean stepping away from collections, and this is one of the biggest risks in collections operations. Once accounts are placed with a third party, there can be a tendency to assume that the problem has moved elsewhere where it has not. The lender still owns the customer relationship, carries the reputational risk, and remains accountable for how customers are treated. Ultimately, the lender still needs to know whether the agency is delivering the right outcomes. Outsourcing is not just a capacity decision. It is a management discipline.

Which accounts should be outsourced? At what point in the lifecycle? To which agency? Based on what performance history? With what commission structure? With what customer treatment standards? With what reporting? And with what level of real-time oversight? If these questions are not being answered properly, outsourcing can create as many problems as it solves. Poor agency management can lead to inconsistent customer treatment, weak performance, higher complaints, and reputational damage. In a world where regulators are paying closer attention to customer outcomes, that is not a risk any lender should be comfortable taking.

Segmentation Still Matters, But It Needs to Evolve

Collections segmentation has always been important. Most collections teams already segment by product, days past due, balance, risk level, customer type, secured or unsecured exposure, and previous behaviour. That still makes sense, as a credit card customer is different from a mortgage customer, a micro-business borrower is different from a large corporate borrower, and a customer who has paid late once is different from a customer with a repeated broken-promise history.

But the discussion in Indonesia raised a critical point. AI creates the opportunity to move beyond broad segmentation and towards more personalised treatment. Historically, segmentation has been about grouping customers who look similar. That is useful, but it is also imperfect. Customers may look similar on paper but behave very differently in reality.

AI gives collections teams the opportunity to move closer to what we might call a “segment of one”, where treatment is shaped by the individual customer’s behaviour, circumstances, contactability, likely response, and preferred channel. This is where the future of collections becomes more interesting, but again, it needs control. Personalised treatment does not mean uncontrolled treatment or letting an algorithm decide everything. It means using better intelligence to support better decisions within a governed strategy framework. The objective should be simple: better outcomes for the customer and better outcomes for the organisation.

Contactability Is Still a Major Collections Problem

One of the most practical issues raised during the panel was customer reachability, and this will resonate with collections teams everywhere. You can have the best strategy, segmentation, AI model, and digital journey, but if you cannot reach the customer, the strategy quickly breaks down.

In Indonesia, the discussion focused on the difficulty of contacting customers in rural areas, customers changing SIM cards, limited use of email, and the practical challenge of managing collections across a large and diverse geography, but this is not only an Indonesian issue. In every market, collections teams face contactability challenges. Customers move, phone numbers change, emails are ignored, letters go unopened, digital journeys are not completed. Some customers are avoiding contact, while others may simply not understand what is required of them. This last point matters.

Not every unreachable customer is unwilling to pay. Some may not fully understand the repayment process or know when payment is due. Some may be dealing with seasonal income, unstable work, family pressures, or temporary vulnerability. If all unreachable customers are treated the same way, some of those treatments will almost certainly be wrong. This is where data and customer understanding become so important. The job is not just to find the customer - it is to understand why they have become difficult to reach and what the most appropriate next action should be.

AI Needs to Work in the Real World of Collections

One of the most useful questions during the panel came from a Sharia rural bank. The issue was not theoretical - it was about the reality of collecting from very different types of customers, including rice farmers, chicken farmers, and market merchants, where loan terms, repayment patterns, and customer circumstances can vary significantly.

The challenge was also practical. How do you apply AI or automation when contactability is difficult, SMS is not always effective, customer segments are very different, and in some cases physical agents are still needed? This question matters because it brings the AI conversation back to where collections actually happens.

Not every customer can be managed through the same digital journey, every product be treated in the same way, every repayment issue possess the same cause, nor every collections process be fully automated. In these circumstances, AI should be seen less as a replacement for the collections operation and more as a way of making the operation more intelligent.

Technology can help organisations understand which actions are working, which actions are cost-effective, and which treatments are most appropriate for different customer groups. It can also help guide the use of agents, digital reminders, or field activity based on what is most likely to produce the right outcome. AI should not simply be used to automate activity. It should be used to understand what works.

Where field activity is still required, AI can help determine which customers should be prioritised, what information agents need before contact, and what the next best action should be after that interaction. Where digital reminders are effective, AI can help identify the best timing, channel, and message. Where customers are not responding, AI can help identify whether the issue is avoidance, affordability, lack of understanding, poor contact data, or something else entirely.

All this, however, still needs to sit inside a controlled framework. Organisations cannot let the system operate without rules, compliance boundaries, and local market understanding. That is particularly important in collections, where the wrong action can quickly damage customer trust, increase complaints, and create regulatory risk. So the better question is not “will AI replace collections agents? but rather “how can AI help collections teams make better decisions in the real-world conditions they are operating in?” That is a far more useful conversation.

Build or Buy? Ask a Better Question

The panel also touched on whether organisations should build their own collections technology or buy a specialist solution. This question comes up often, and the answer depends on the organisation. Some fintechs and highly technology-led lenders may have the capability, resource, and appetite to build their own platforms. For them, technology may be part of their core business model. But for many banks, lenders, and financial institutions, building is not as simple as it first appears.

Building the first version is only the beginning. What happens when regulation changes, new channels need to be added, AI governance requirements become more demanding, the business wants new reporting, new segmentation, new agency management capability, or new customer journeys, and/or the people who built the system move on?

The real question should not be “can we build it?” but instead “can we continue to develop, maintain, govern, and improve it at the pace the business requires?” Collections technology is not static. It needs to evolve constantly because customer behaviour, regulation, operating models, and economic conditions are constantly changing. For many organisations, buying a specialist debt collection management platform may allow them to focus on what they are really there to do: lend responsibly, support customers, manage risk, and recover effectively.

The Best Time to Invest Is Before the Pressure Arrives

This was probably the strongest message from the panel: do not wait until the crisis hits. It is very easy for collections investment to be delayed when conditions are stable, volumes are manageable, performance is acceptable, the business has other priorities or when transformation programmes can wait. But when arrears begin to rise, customers come under pressure, operational capacity is stretched, and regulators start asking more questions, it is too late to begin building the foundations.

Collections capability cannot be created overnight - organisations need the data, workflows, digital channels, agency controls, reporting, segmentation, governance and, increasingly, the AI and decisioning capability to make sense of what is happening quickly enough to act. The organisations that wait until the problem is visible will always be behind those that invested when things still looked calm, and this is where collections leaders need to challenge their own organisations. If consumer pressure is building, operating costs are rising, customers are becoming harder to reach, and regulatory expectations are increasing, then investment in collections should not be a difficult argument to make - it should be obvious.

So, What Should Collections Leaders Do Now?

The Indonesia panel discussion was a useful reminder that while markets differ, core collections challenges are remarkably consistent. There are five practical actions collections and recoveries leaders should be thinking about now:

• First, review the quality of customer and contact data. If the data is poor, fix that before expecting AI or automation to deliver meaningful results.
• Second, assess whether the collections model is still too reactive. If intervention only starts once the customer has already fallen into difficulty, the organisation may be missing the best opportunity to prevent the problem.
• Third, review agency management and oversight. Outsourcing may provide capacity, but it does not remove accountability.
• Fourth, look again at segmentation and treatment strategies. Are they still broad and static, or are they becoming more dynamic, behaviour-led, and customer-specific?
• Fifth, think carefully about how AI can support the real-world collections operation. The best use of AI may not be to automate everything, but to help teams make better decisions about customers, channels, agents, timing, and treatment.

Collections is changing; that much is clear. The organisations that succeed will not be those that simply adopt the newest technology or talk the most about AI. They will be the organisations that combine better data, better decisioning, better customer understanding, and better operational control. That is what will separate collections teams that are simply busy from collections teams that are genuinely effective, and in the environment we are moving into, that difference will matter.

Talk to an EXUS expert to explore how EXUS EFS can help strengthen your debt collection management strategy and support more proactive portfolio management.