Most companies today rely on rule-based dunning software to follow up on overdue invoices. These tools send automated reminder emails on a set schedule. The problem is obvious: customers can instantly see that these messages are templated and not written by a human. As a result, they often get ignored.
This is why many finance teams end up hiring teams of credit controllers whose sole job is to send personalized emails, follow up in threads, and escalate in a way that feels human. While this approach improves response rates, it is costly, hard to scale, and still leaves AR teams bogged down in repetitive work.
AI agents change the equation. Just as large language models (LLMs) are transforming customer support and sales outreach, they are now being deployed in AR and collections to run personalized, contextual dunning workflows that feel like they are coming from a human.
1. Personalise collection workflows at scale with contextual outreach and follow-ups
One of the biggest weaknesses of traditional dunning is that everyone gets the same email sequence, regardless of their payment history, relationship with your business, or the context of the overdue invoice.
AI agents can change this by adjusting tone, style, and frequency to match your brand voice and the customer’s behavior. Instead of rigid templates, each message feels like part of an ongoing, human-led dialogue. The principle of threat continuity, where every follow-up feels like a natural progression from the last, ensures customers take outreach seriously just as they would with a real AR manager escalating a collection.
This mirrors many of the same principles that have made AI-powered sales development so effective: contextual sequencing, objection handling, and multi-touch personalization. Just as SDRs win more responses by tailoring outreach to a prospect’s needs, AI agents in collections increase payment rates by tailoring dunning to a customer’s history and behavior. The difference is that now, finance teams can do this at scale without needing to hire dozens of collectors.
2. Beyond reminders: True two-way conversations
AI agents can also engage in two-way dialogue, not just fire off reminders. This is the real breakthrough compared to legacy dunning tools.
For example, when a customer replies with:
“Can you resend the invoice?” → The agent can locate it in your ERP and send it back.
“We need to update our PO number before paying” → The agent can capture the request, suggest next steps, and log it for human approval if needed.
“We’ve already paid” → The agent can check payment records, confirm status, or escalate if there’s a mismatch.
Instead of static sequences, dunning becomes a living conversation, just as if one of your AR reps were writing every email.
3. Capturing and acting on promise-to-pay dates
A huge part of real-world collections is when a customer says: “We’ll pay next week.” A human collector knows to pause outreach until that date, then check whether the payment has arrived, and follow up if it hasn’t.
AI agents can do this automatically:
Detect the promise-to-pay in the customer’s email.
Log the date in the system.
Pause dunning until the date has passed.
Check again if the invoice has been settled.
Resume follow-ups if payment is still outstanding.
This avoids over-messaging customers who have already committed to pay, while ensuring continuity if they fail to deliver.
Final thoughts
AI agents bring a new paradigm to dunning workflows. Instead of static, one-size-fits-all reminder sequences that customers ignore, companies can now run personalized, conversational, and context-aware collection processes at scale.
By capturing promise-to-pay dates, maintaining continuity of escalation, and applying proven tactics from AI-driven sales outreach, AI agents can act like a full AR team, chasing payments, resolving queries, and escalating only when needed.
For CFOs, the benefits are clear: lower DSO, fewer bad debts, reduced headcount pressure, and less reliance on costly collection agencies.