Why AI Email Templates Aren't Real AR Automation

Most "AI-powered" accounts receivable platforms don't actually automate collections. They automate email drafting. The AI writes a polite reminder, maybe predicts which invoices will be paid late, and recommends what your team should do next. Then your team still has to do it.

That's not AI accounts receivable automation. That's a suggestion engine with a nice interface.

35% of B2B invoices are still paid late

Despite a decade of email-based AR tools, roughly 55% of all B2B invoiced sales in the US are past due. In Western Europe, the numbers are nearly as bad. Three out of four companies report experiencing late B2B payments.

These numbers haven't improved much since email automation became standard. The tools have gotten smarter, but the invoices are still late. That should tell us something about the limits of email as a collection channel.

An email is easy to archive. Easy to mark as read and forget. Easy to filter into a folder that nobody checks. For the customer who genuinely forgot, a reminder email works fine. But for the customer who's stalling, disputing, or struggling to pay? Another email changes nothing.

What "AI-powered" actually means in most AR software

Open the marketing page for almost any AR automation platform and you'll see the words "AI-powered" or "intelligent automation." But look at what the AI actually does:

• Drafts email templates. The AI generates personalized reminder language based on the customer's payment history and the invoice aging stage. You get a suggested email. You still click send.
• Predicts payment likelihood. Machine learning scores each invoice based on historical patterns. "This invoice has a 73% chance of being paid on time." Useful for prioritization, but it doesn't collect the money.
• Recommends next actions. The system suggests that you should call this customer, or escalate that one, or offer a payment plan to another. The recommendation sits in a dashboard until someone acts on it.
• Categorizes inbound emails. AI reads reply emails and sorts them: "dispute," "promise to pay," "request for invoice copy." Good for triage, but it doesn't resolve any of these.

Every one of these features is text-based. The AI reads text, writes text, and classifies text. It never talks to anyone. It never resolves anything on its own. The moment the situation requires a human conversation, the "AI-powered" platform hands the problem back to your team.

The phone call gap

Here's where the workflow breaks down. Your AR automation platform sent four emails over three weeks. The customer opened two of them. They didn't respond to either. The system's recommendation: "Escalate to phone call."

Now what?

Someone on your team has to find the time to make the call. They have to prepare, look up the invoice, check the history. They have to pick up the phone and have an awkward conversation about money with a customer they probably also sell to. That call takes 5-10 minutes if it goes smoothly, longer if there's a dispute.

Multiply that by 20 overdue invoices. Or 50. Or 200.

Most businesses simply don't make those calls. They focus on the largest invoices and let the smaller ones age. The "AI-powered" AR tool keeps sending emails to customers who have already demonstrated that they don't respond to email. Eventually the invoice hits 90 days and gets written off or sent to collections.

The AR platform worked perfectly. It sent every email on schedule. It predicted which invoices were at risk. It recommended calling. But nobody called, and the money never came in.

What real AI accounts receivable automation looks like

Real automation means the AI doesn't just recommend the phone call. It makes the phone call.

An AI voice agent dials the customer, introduces itself, and has an actual conversation about the overdue invoice. Not a robocall. Not a pre-recorded message. A back-and-forth dialogue where the agent listens to what the customer says and responds accordingly.

If the customer says they never received the invoice, the agent sends it via SMS during the call. If they say they can pay next week, the agent captures that commitment with a specific date. If they want to pay right now, a payment link arrives on their phone before the call ends. If they have a dispute, the agent documents the reason and routes it for resolution.

This is the difference between automation that stops at the inbox and automation that actually closes the loop. The AI doesn't generate a suggestion for your team. It handles the interaction itself, from dial to outcome.

The data: phone calls work, and cutting them costs you

The strongest evidence for voice-based collection comes from one of the Netherlands' largest collection agencies. When this organization reduced its outbound phone contacts by 24% (from 63,000 to 48,000 calls), their amicable resolution rate dropped 18 percentage points, from 57.8% to 40.0%. Cases that used to be resolved with a conversation started going to court instead, which is slower, more expensive, and worse for every party involved.

The correlation is direct. Fewer calls, fewer resolutions. More calls, more resolutions. Email didn't fill the gap. Chat didn't fill the gap. Only conversations resolved cases at scale.

Phone-based collection has consistently shown response rates between 40-60%, compared to 15-25% for email. The challenge has always been cost. An AR clerk making manual calls handles a limited number of accounts per day. At $37,000+ per year per clerk, most SMBs can't justify the staffing.

AI voice removes that constraint. Every overdue invoice gets a call. Every customer gets a conversation. The agent works 24/7, never has a bad day, and captures structured data from every interaction.

What to look for in AR automation

If you're evaluating accounts receivable software, ask a simple question: what happens after the emails stop working?

If the answer is "your team takes over," the platform is an email tool with AI features, not an end-to-end automation solution. Look for these capabilities instead:

• Automated outbound calls that handle the conversation, not just the dialing
• Real-time SMS delivery of invoices and payment links during calls
• Structured outcome capture: promise-to-pay dates, dispute reasons, callback requests
• Tone calibration that matches the invoice aging stage (friendly at 7 days, direct at 60)
• First-party calling under your company's identity, not a third-party collection agency

The gap in AR automation has never been about smarter emails. It's been about the phone call that nobody wants to make. AI voice agents close that gap.

Curious what an AI collection call actually sounds like? Try Dunwise's interactive demo and experience a real AI conversation about an overdue invoice.