At the Portfolio Debt Collection Trends 2026 conference, a session examined how behavioral science and automation are changing debt collection practice. Speakers presented large-scale experiments and live deployments showing that simpler choices and lower initial instalments can increase repayments, AI-mediated communication can reduce perceived stigma, and modern agents already consolidate data, prioritise cases and execute repeatable administrative tasks. Human judgement, however, remains essential for critical decisions.
What behavioural science reveals
Dr. Sebastian Clajus, behavioural lead at PAIR Finance, argued that missed payments have diverse causes: misunderstanding, lack of funds in a particular month, shame or procrastination. In many such cases, increasing pressure can worsen cooperation. Clajus therefore recommends simplifying the payment journey through clearer communication, fewer options, better timing and removing unnecessary steps.
He cited a German–Swiss experiment that examined nearly 289,000 cases. Debtors could set the size of their first instalment using a slider; the only system difference was whether the slider defaulted to 20% or 80% of the total debt. With a 20% default, Germany saw 40% more accepted payment solutions and Switzerland 30% more, and the within-13-day repayment rate doubled. The result demonstrates that higher initial expectations do not necessarily yield more recoveries.
AI communication and reduced stigma
A psychological study conducted across 11 European countries with 3,514 participants tested interactions with human versus AI assistants while keeping debt amounts, instalment options and messaging identical. The AI reduced the experienced risk of stigmatization by 36% without decreasing trust, and the effect was stronger among older age groups.
However, the research also showed where human communication still outperforms machines: empathy, fairness and reciprocity were rated higher for human agents. Accordingly, AI appears most useful for initial contact, reminders and setting up instalment plans, while disputed claims, vulnerable clients and renegotiations are better handled by people.
Consolidating data and prioritising tasks in practice
Murvai Tamás, founder of AI Path Pro, demonstrated practical capabilities. Many organisations still operate with multiple ERP systems, internal apps, bank data and Excel spreadsheets that staff must manually combine. In his example, he fed nine different data sources to an AI: the system produced a consolidated workbook with formulas, an executive dashboard and a concrete task list, while also checking data quality (spotting inconsistent date formats, mis-stored numbers and mismatched currency notations).
The output identified long-overdue receivables, cases where payment promises had been broken, missing documents, and discrepancies between bank records and internal ledgers. Modern systems therefore do more than summarise: they set priorities and prepare the next actions for each case.
Autonomous agents and recurring processes
Murvai gave an example where an AI agent located invoices received by e-mail during a year, filtered duplicates, renamed files consistently and forwarded them to the accounting department. Once the process proved reliable, it was turned into a monthly recurring task and the agent continued to run it without further human intervention.
The same logic applies to debt collection: a system can periodically collect available data, verify its quality, re-rank cases and prepare next steps. This goes well beyond asking a chatbot to draft a letter; it is executing end-to-end process steps.
When traditional automation suffices
In the session’s closing discussion, Hörömpöli-Tóth Levente (Portfolio Csoport) spoke with Sebastian Clajus and Murvai Tamás about the limits of automation. Murvai warned against first choosing a new technology and then looking for a use case. He recommended starting from the tasks that cause the most pain for staff or are least liked.
He offered an example: an HR colleague spent two to three days a month cutting payslips out of a 100+ page PDF, uploading and sending them. The company considered an AI agent, but a phone call revealed the accounting system could automate the task in 20 minutes. The case illustrates that conventional automation can be simpler, cheaper and more controllable than an AI implementation.
Scale, boundaries and the role of humans
PAIR Finance handles roughly 400,000 cases per month with fewer than 300 staff; Dr. Sebastian Clajus said more than 70% of cases can already be closed without human intervention. Much of this relies on simple automation and decision trees, but critical decisions with legal consequences still require human deliberation.
Speakers emphasised safety and control: the risk of autonomous agents grows if they are given overly broad access and autonomy. New processes should be tested in controlled environments and only widely automated once their behaviour is predictable. Data protection, access rights and GDPR issues must be resolved prior to deployment.
Opportunities in monitoring and personalisation
One of the largest opportunities in corporate debt collection is continuous client monitoring and early-warning detection: the volume of available data about a client can be too large for manual, continuous oversight, but technology can flag changes that indicate rising risk.
On the retail side, having more relevant information about a debtor’s situation allows better tailoring of communication and proposed solutions. Automation can take over routine tasks, process large datasets and, in some cases, reduce the social burden on the client. Human involvement remains strongest where disputes, vulnerability or significant consequences demand judgement.
