Automated Collection Calls: Which AR Platforms Actually Phone Late Customers

Automated collection calls are outbound phone calls placed by software rather than a collector, used to chase an overdue invoice once email has stopped working. As of August 2026 only a small number of receivables platforms actually place them: Sidetrade, through its Aimie agent, and AccountsReceivable.ai. Quadient AR, Versapay, Billtrust, HighRadius, Gaviti, Tesorio and FIS GETPAID all stop at email and portal messaging and hand the phone work back to your team. That single distinction decides more AR evaluations than any feature grid, because the invoices costing you the most are the ones email has already failed on.

Almost every vendor in this category will tell you they automate collections. Read the demo carefully and you will usually find that what gets automated is the reminder, the escalation flag and the aging dashboard. The call itself, the part where someone reaches a human at the customer and gets a commitment to pay, is still a person on your team with a phone and a list. Here is who really dials, when a call is worth making, and what the US rules require if software is doing the talking.

Which AR automation platforms actually make collection phone calls?

Very few. The category grew up around email sequencing and payment portals, and calling is a genuinely different capability to build. This is the current state of the major platforms.

We build a competing product, so treat our own row as our position rather than a neutral verdict. We are naming Sidetrade first anyway, because a lot of comparison pages in this category still claim no competitor calls, and that has not been true since Sidetrade launched autonomous calling. If a roundup tells you Sidetrade is email-only, it is copying an older source, and you should discount the rest of it. For a fuller side-by-side of that platform against the mid-market option people most often weigh it against, see our Sidetrade vs YayPay comparison.

What is an automated collection call?

An automated collection call is an outbound voice call triggered by your receivables system when an invoice hits a defined condition, such as 15 days past due with no reply to three emails. Software places the call, speaks with whoever answers, states which invoice is outstanding and for how much, asks for a payment commitment or a reason for the delay, records the outcome against the account, and escalates to a human when the situation calls for it.

That is meaningfully different from two older things it gets confused with. It is not a robocall playing a recorded script at whoever picks up, and it is not a power dialer that queues numbers for a human collector to work. A modern AI collection agent holds a real conversation: it can answer "which invoice is this about", accept "we paid that on the 14th", and route a disputed line to your team instead of arguing about it.

Why does email-only dunning stall at 60 days?

Because the reminder and the decision live in different places. Your dunning sequence lands in an accounts payable inbox, which is usually a shared mailbox processed in batches by someone with no authority to release a payment. A well-written reminder works on the invoices that were going to get paid anyway, where the only problem was that it fell off a list. It does very little on the invoices that are late for a reason.

By 60 days past due, an invoice has typically absorbed three to five automated emails. A sixth is not new information. What has usually happened is one of four things: the invoice never reached the right approver, there is an unraised dispute about a line item, the customer has a cash problem and is prioritizing other suppliers, or a purchase order or remittance detail is missing and nobody flagged it. Every one of those is discovered in about ninety seconds on a phone call and almost never surfaces by email, because responding to an email costs the recipient effort and ignoring it costs them nothing.

That asymmetry is the whole argument for escalation. A call changes the cost balance: it puts a person on the spot, and it produces either a commitment or a reason. Our guide to reducing DSO covers where escalation fits in the wider cadence, and the collections call script covers what to actually say when a human makes the call.

When does a phone call beat another reminder email?

Call when the email channel has demonstrably failed, when the amount justifies the effort, or when you need information rather than payment. In practice that means an invoice past 30 to 45 days with no response to the sequence, any invoice large enough to move your month, and any account where the aging pattern changed suddenly. Do not call on day two of a five-day slip.

The sequencing that works for most mid-market teams looks like this. Days 1 to 14 past due, automated email reminders escalating in firmness. Around day 15 to 21, add SMS, which has a far higher open rate and reaches a person rather than a shared mailbox. From roughly day 30, escalate to a call. That is the point where automation stops being a nudge and starts being collections.

Two caveats worth stating plainly. Calling early and often on small balances annoys good customers and costs you goodwill you will want later. And calling is not a substitute for a credit process: if an account keeps reaching day 60, the fix is a credit policy and a limit, not more calls.

Is it legal to use AI to make collection calls in the United States?

Yes for business-to-business collections in most circumstances, but the rules are real and they tightened. This is a summary, not legal advice, and you should have counsel review your specific setup before switching on outbound calling.

Two federal regimes matter. The first is the Fair Debt Collection Practices Act, which governs third-party collectors pursuing consumer debts. A business collecting debts owed to itself is generally not a "debt collector" under the FDCPA, and commercial trade debt between two businesses generally falls outside its scope entirely. That is why B2B receivables sits on much simpler ground than consumer collections.

The second is the Telephone Consumer Protection Act, and this is where AI calling changed. On February 8, 2024 the FCC issued a unanimous declaratory ruling confirming that the TCPA's restrictions on "artificial or prerecorded voice" cover AI technologies that generate human-sounding voices, including real-time conversational agents. The FCC was explicit that the statute does not allow a carve-out for technology that claims to be the equivalent of a live agent. The ruling took effect immediately. The practical consequence is that an AI voice call is treated as an artificial voice call, and prior consent is required where the TCPA applies.

Scope is what saves most B2B programs. The TCPA's artificial-voice restrictions attach to calls placed to residential lines and to wireless numbers. Calls to a business landline are on much safer footing than calls to a mobile, and a great many AP contacts are mobile numbers, which is exactly the trap. One further wrinkle: in February 2026 the Fifth Circuit held in Bradford v. Sovereign Pest Control of Texas that the TCPA's text requires only prior express consent rather than prior express written consent for artificial-voice calls. That applies in Texas, Louisiana and Mississippi. In the other 47 states courts continue to apply the FCC's written-consent rule, so do not plan a national program around the Fifth Circuit position.

On top of federal law, state rules vary. Several states require all-party consent to record a call, and a growing number require disclosure when a caller is an AI rather than a person. The workable compliance posture for B2B AR is straightforward: capture consent to contact in your credit application and terms, keep a documented record of it, disclose at the start of the call that it is an automated system, honor opt-outs immediately, and keep an easy path to a human. Ask any vendor how each of those is implemented in their product, not whether they are "TCPA compliant", which is a claim rather than a control.

What does an AI collection call actually sound like?

A good one is short, specific and unremarkable. It identifies the company calling, names the invoice number, date and amount, asks whether there is anything preventing payment, and then does one of three things: takes a commitment date, captures a stated reason, or routes the account to a person. It does not negotiate, threaten or improvise around a dispute.

The technology is not unique to receivables. The same class of voice agent is now common in outbound sales, where AI systems that qualify leads and book meetings by phone have made the underlying capability fairly ordinary. What is specific to collections is the restraint required. A sales agent is rewarded for persistence. A collections agent talking to a customer you want to keep for another decade is not, and an agent that argues with a legitimate dispute will cost you more than the invoice.

The failure mode to test for in a demo is what happens when the call goes off-script. Say "we already paid that", say "I am not the right person", say "we are disputing the freight charge", and say nothing at all. A weak agent loops or hangs up. A good one logs the outcome accurately and hands off.

How do you evaluate a vendor that claims it makes calls?

Ask for specifics, because "AI-powered collections" is used to describe email sequencing far more often than it is used to describe calling. Six questions separate a real capability from a roadmap.

That last one matters more than teams expect. Chasing a customer who paid you three days ago is the fastest way to damage an account, and it happens constantly when the integration is a nightly export rather than a live sync. Our guide to AR automation ERP connectors covers how to tell those apart before you sign.

Do you need calling, or just a better email sequence?

Be honest about where your cash is stuck before you buy for this. Pull your aging and look at the balance sitting past 45 days as a share of total receivables. If it is small and most of your late invoices clear within a couple of weeks of the first reminder, your problem is cadence, not channel, and a well-configured email sequence with SMS will get you most of the way. If a meaningful share of your balance is aging past 60 days and your team says the honest reason is that nobody has time to work the list, no amount of additional email will fix that, because email is what already failed.

The other case for calling is capacity rather than effectiveness. Plenty of finance teams know exactly which twenty accounts need a call this week and simply never make them, because collections competes with close, reporting and everything else. Software that makes those calls consistently beats a person who makes them occasionally, which is usually the real comparison.

If that describes your situation, our collections automation software page covers how the escalation from email to SMS to a live AI call works against your ledger, and the AI collections agent page covers what the agent does once an account is assigned to it. If you are still building a shortlist, how to choose accounts receivable software walks through scoping the buy before you sit through demos.