Order matters in collections. Before an agent asks a customer in arrears for payment, it should look for signs that the customer is in financial difficulty, and if it finds them, the approach should change. A request for money that ignores visible hardship can push someone further into it. The check comes first, every time, and its result decides what the agent says next.
Why must the check come before the ask?
Most collections processes are built around the ask: a reminder, a call, a request for a payment or a plan. Difficulty is something the process responds to if the customer raises it. That puts the burden on the customer, often at the moment they are least able to carry it.
An agent can reverse this. It has access to information a busy collections officer may not have time to read: transaction patterns, past contact notes, recent changes to income, other products the customer holds with the bank. It can look at all of it before it makes contact. If the signs point to difficulty, the first message should be an offer of help, not a demand.
In the UK, the FCA's Consumer Duty expects firms to deliver good outcomes for retail customers, including those in vulnerable circumstances. Other markets frame it differently, but the direction is common. This is an explanation of supervisory direction, not legal advice; confirm how it applies to you with counsel.
What signs can an agent look for?
The question of whether an AI agent can spot a customer in financial difficulty has a qualified yes: it can spot signals, which a person then interprets. Useful signals fall into a few groups.
| Signal group | Examples an agent can see | What it may indicate |
|---|---|---|
| Income | Regular salary credits stopped or reduced | Job loss, reduced hours, illness |
| Essential spending | Missed rent, utility or local tax payments | Prioritising some bills over others under strain |
| Borrowing behaviour | Persistent use of overdraft for essentials, new high-cost credit | Using credit to cover living costs |
| Arrears pattern | First missed payment after years of regular payment | A sudden change in circumstances |
| What the customer has said | Mentions of bereavement, illness, separation or job loss in past contacts | A life event affecting ability to pay |
| Account control | Unusual transfers to another person, changes in who operates the account | Possible financial abuse |
The last row needs particular care. Signs of financial abuse, where someone else is controlling or taking a customer's money, change the situation entirely. Contacting the customer through usual channels may alert the abuser. These cases should go straight to a specialist person.
What should change when difficulty is detected?
Detection is only useful if it changes what happens. The agent's behaviour should shift in clear, predictable ways.
- Pause automated asks. No standard payment reminders while the case is reviewed.
- Change the first contact. Lead with acknowledgement and options, not with the arrears amount. Offer to talk to someone.
- Offer appropriate support. Explain what forbearance options exist and how to access free, independent debt advice where available.
- Hand to a person. A trained collections officer reviews the case and decides on any arrangement.
- Record the signal and the response. File what the agent saw, what it changed and who took over.
The playbook on how to identify customers in financial difficulty with an AI agent sets out these steps in operational detail.
What stays with a person?
Agents can detect and adapt. They should not decide how a customer in difficulty is treated. Agreeing a reduced payment, freezing interest, deciding whether to refer the debt, or concluding that a customer is or is not vulnerable are judgements with lasting effects on a person under strain. A trained officer makes them.
The agent's role is to make sure the officer sees the case early, with the evidence, before any harm from a poorly judged ask.
How should the collections system be set up?
The check-first principle has to be enforced by the platform, not left to the agent's prompt. When you connect an AI agent to a collections system, the outbound contact capability should require a current difficulty check result before any message is sent. If the check has not run, or is out of date, the message is held.
The collections contact strategy matrix should carry difficulty and vulnerability flags as inputs, so that frequency, channel and content all change when a flag is set. A contact rule that ignores the flags is a gap.
What are the trade-offs?
Signals produce false positives. A salary that stops may mean a new job paid on a different date. Treating every signal as difficulty would slow recovery and irritate customers who are fine. The answer is not to ignore signals, but to let them change the tone and route of the first contact rather than to stop all contact. A gentle, open message costs little if the customer is fine, and matters a great deal if they are not.
Where Pryme fits
In Pryme, the minimum chain can require a current difficulty check before any collections contact, and hold the contact if the check is missing or flags a concern. Difficulty and abuse signals lower the agent's depth for that customer to drafting, and the case goes to a named officer. See AI agents in banking: what they may and may not do.
Questions people ask
Can an agent ask customers directly if they are struggling?
Yes, and a well-worded, open question is often the best first contact. The agent should make it easy to say yes and explain what help is available.
Does spotting difficulty mean the debt is written off?
No. It means the approach changes: more time, a sustainable arrangement, or support. Decisions about the debt itself stay with a trained officer.
How recent should the difficulty check be?
Run it at the moment of each contact. A customer's circumstances can change between a campaign being planned and a message being sent.