AI-Powered Early-Stage Collections: How to Anticipate Risk Before It Becomes a Loss
The ability to anticipate risk loses value if there is no effective contact execution. That is why world-class organizations such as Pentafon are already incorporating AI models that monitor network traffic in real time and route interactions through the highest-performing carriers based on geographic location, improving contact rates and supporting more precise and timely collections strategies.
In an environment of increased credit origination, more digital channels, and customers who are less willing to answer traditional phone calls, companies need to anticipate risk before it turns into a loss.
The difference lies in identifying risk earlier, making better contact, and executing consistently.
The growth of credit in Mexico has increased the operational challenge for banks, fintechs, retailers, and telecommunications companies.
The challenge is no longer just about originating more credit. It is about managing risk more effectively from the very first signs of deterioration.
For credit cards, Banco de México reported that as of June 2025, the comparable portfolio of revolving cardholders consisted of 11.2 million cards, with an outstanding credit balance of 357.7 billion pesos. During the same period, the delinquency rate stood at 3.3%.
These figures confirm an operational reality. Collections can no longer begin only once delinquency appears. They must start earlier, supported by data, segmentation, and strong contact capabilities.
Losses are not created at the end of the process. They begin to build during early-stage delinquency.
In this context, generative AI, advanced analytics, and omnichannel strategies make it possible to identify customers with a higher likelihood of deterioration, prioritize accounts, determine the best channel, and execute early interventions with greater consistency.
Modern collections must operate as a system. It must integrate technology, specialized talent, data governance, cybersecurity, and operational execution.
It is not about contacting more customers. It is about contacting them better and sooner.
The Real Shift in Collections Is Strategic
The most important change in collections is not technological. It is strategic.
For years, many operations relied on general rules: days past due, balance amount, overdue balance, and mass outreach campaigns. That model is still useful, but it is no longer sufficient for high-volume portfolios.
Today, the difference lies in identifying which customer requires immediate intervention, which customer can self-manage, which channel has the highest probability of generating contact, and which message is most likely to drive a meaningful response.
McKinsey has pointed out that leading institutions use advanced analytics and machine learning to transform collections models and generate value quickly. The logic is clear: better data leads to better decisions.
In practice, this means moving from volume-based collections to probability-based collections.
The operation stops asking only, “How many accounts were contacted?” It starts asking, “Which accounts should have been contacted, when, through which channel, and with what treatment strategy?”
Three Capabilities That Define Outcomes
AI applied to early-stage collections must solve three operational challenges.
First, dynamic risk prediction.
Organizations must identify signs of deterioration before default occurs. Behavioral analysis, payment history, exposure levels, responses to previous campaigns, and pattern changes allow for more effective prioritization.
Second, omnichannel orchestration.
Collections can no longer operate as an isolated channel. Voice, WhatsApp, SMS, email, IVR, transactional bots, self-service, and specialized agents must operate under a unified contact strategy.
Third, personalization at scale.
Every customer requires a different approach. AI makes it possible to define messages, schedules, self-service paths, and repayment options with greater precision. This improves the customer experience and prevents unnecessary friction.
Collections stop operating through persistence and start operating through intelligence.
Generative AI: Efficiency Without Losing Control
Generative AI does not replace operational discipline. It makes it scalable.
Its value lies in processing larger volumes of information, automating repetitive decisions, assisting agents, generating contact workflows, and improving interaction quality.
Antonio Fajer, CEO of Pentafon, has stated it directly: “AI comes to complement the services that agents provide and make our operations more efficient so we can meet the increasingly demanding service expectations of our clients’ customers.”
This vision is especially relevant to early-stage collections.
Pentafon has already identified use cases where technology enables services that were previously not cost-effective, such as welcome calls, informational pre-collections outreach, early-stage collections, virtual agent assistants, training simulators, and interaction automation.
The opportunity lies in combining AI with operations—not replacing one with the other.
PwC estimates that broad AI adoption in banking can improve the efficiency ratio by up to 15 percentage points. The impact does not come from a single tool. It comes from redesigning entire processes around data, automation, and governance.
Security Is Part of the Model, Not an Add-On
Collections operations handle sensitive information. They manage personal, financial, and contact data, as well as payment histories and customer behavior information.
For this reason, any AI-powered collections model must incorporate security by design.
Antonio Fajer has warned that with artificial intelligence, data protection risks are amplified because AI is fueled by information and has the ability to manage millions of data points, filter them, establish relationships, and make decisions.
The conclusion is clear. Intelligent collections cannot exist without data governance.
Companies must ensure that their internal operations or service providers maintain robust certifications. According to Fajer, PCI DSS Level 1 Version 4, together with ISO 27001, have become “the only way to guarantee company information and customers’ sensitive data.”
The most effective way to scale AI-driven collections is by combining data, security, technology, and execution.
The Right Partner Determines the Outcome
Designing an advanced strategy is only half the challenge.
The other half is executing it every day with consistency, traceability, and control.
In collections, differentiation is not found solely in the analytical model. It lies in the ability to integrate it with channels, dialing systems, bots, agents, QA, compliance, reporting, security, and continuous improvement.
This requires four capabilities.
1. Data-Driven Operations
Every decision must be supported by information.
Operations must know who to contact, through which channel, at what time, with what message, and for what purpose. Static rules must evolve into dynamic models that learn from actual portfolio behavior.
2. Information Governance and Security
Traceability is no longer optional.
Financial institutions and credit-intensive organizations require evidence, access controls, data protection, monitoring, and compliance. AI must operate within a secure, auditable framework aligned with regulatory requirements.
3. Omnichannel Execution at Scale
Strategy only creates value when executed at scale.
A modern operation must coordinate millions of monthly interactions without losing consistency. It must integrate both human and digital channels. It must measure contact rates, right-party contact, payment promises, conversions, recoveries, abandonment, cost, and customer experience impact.
4. Specialized Talent
Collections remains a discipline centered on human interaction.
Technology prioritizes, automates, and assists. But critical moments still require judgment, empathy, negotiation skills, and operational expertise. Specialized talent remains essential for turning contact into recovery.
Expected Impact of Advanced Collections Models
When these capabilities are integrated into a unified model, collections undergoes a structural transformation.
In advanced collections models, the integration of analytics, automation, and AI can impact key metrics such as recovery rates, operating costs, abandonment rates, effective contact rates, and customer experience.
The outcome does not come from a single tool.
It comes from a complete operating model: data, technology, processes, security, and execution aligned with business objectives.
Pentafon’s Role
Pentafon integrates operations, technology, talent, and security to transform collections, customer service, and customer experience processes.
The company has invested in international certifications for security, quality, and availability, as well as new artificial intelligence models and technology modernization initiatives for customer engagement.
Its approach enables the operation of early-stage collections models, informational pre-collections outreach, contact automation, agent assistants, speech analytics, transactional IVR, bots, and omnichannel self-service journeys.
The goal is not to add AI as another layer. The goal is to redesign the operation so that AI delivers measurable impact.
The objective is clear: prioritize better, engage earlier, reduce friction, protect data, and improve recovery outcomes.
Conclusion
Early-stage delinquency defines a significant portion of financial performance.
Poor prioritization, delayed outreach, or inconsistent execution reduces recoveries and increases costs.
Today, technology makes it possible to process more accounts, make better decisions, and scale strategies without increasing costs at the same rate.
But technology alone does not solve the problem.
Results depend on four variables: reliable data, security, omnichannel orchestration, and operational discipline.
That is where the difference between recovering a portfolio and selling it early is determined.
That is also where the value of a strategic partner is defined.
Collections does not begin with delinquency. It begins with the ability to anticipate risk.
Sources
Banco de México. Basic Credit Card Indicators. Data as of June 2025.
McKinsey & Company. The Analytics-Enabled Collections Model.
PwC. How AI Is Reshaping Banking.
Pentafon. Interview with Antonio Fajer, CEO of Pentafon, on artificial intelligence, early-stage collections, cybersecurity, and certifications.