One outbound campaign on your own number: accounts current for 6 months or more, first miss only, consent on file, 9 am to 7 pm local, 2 attempts a week, with a person on the same desk for transfers.
AI makes the routine reminder calls. Your team keeps the hard ones.
The AI places pre-due and first-missed calls, verifies the customer and logs the promise. Your team keeps hardship, disputes and payment plans, so customers get speed and care that neither gives alone.
- Can say it is an AI, offers a person
- Verifies before it states an amount
- Hardship goes to a person, same call

Pre-due reminder
Your bill is due Monday. Want a pay link by text?
First missed payment
Right party confirmed, then a pay link or a dated promise.
Return call after your text
The customer reached out, so answer on the first ring.
Hardship or a plan request
I can do $60 a month. The AI hands off in the same call.
Why payment reminder calls will run on AI and people together
Most late bills are timing, not refusal, and 17% of US adults did not pay all bills in full in one month. A people-only desk cannot call that many accounts on time. But a missed bill can also mean job loss or a dispute, and 51% of collection communication complaints were about frequent calls. An unchecked dialer does harm.
- Calls every pre-due and first-missed account on the day it is due
- Same disclosure, verification and read-back on every call
- Answers return calls on the first ring, at 8 pm or on Sunday
- Hears job loss or illness and offers a plan that fits
- Traces a disputed or missing payment through the ledger
- Holds the authority to hold a shutoff or settle a balance
- A customer who forgot gets a short call and a pay link, with no collector's tone
- Hardship moves to your team in the same call, with the transcript and mood on screen
- Your staff spend the day on plans and disputes, not on voicemail and repeat reminders
If the worst case is a repeat call, let the AI finish. If money, terms or a customer in trouble are involved, a person finishes.
Send the pay link by text or email. Call the people who did not act, and answer every return call. Let your own data set the mix.
Track kept promises, 30-day cure and complaints against a held-out control list. Calls handled tells you little about money collected.
How can AI help with payment reminder calls?
AI works best on the short, scripted reminder calls: a heads-up before the due date, a first-missed-payment call, a promise-to-pay confirmation, and the return call from a customer who got your text. It confirms who it is speaking to, states the amount, sends a pay link or records a date, and passes hardship, disputes and anything outside policy to a person with the whole call.
- Text and email first; the call is the second touch for people who did not act
- The AI verifies the person before it says any amount
- One shared call counter per account, across AI and people
- Start with one list, one outcome, and a held-out control group
Most late bills are not refusals. How you reach out decides what happens next.
A missed payment is usually timing, forgetting or a thin month. A reminder that arrives on time helps. A reminder that arrives too often, at the wrong hour, or at the wrong person does damage.
of US adults did not pay all their bills in full the month before the survey; 34% of those with income under $25,000 (Fed SHED, Oct 2024).1
higher chance of on-time payment, and about 2 fewer days late per month, for borrowers sent a pre-due SMS reminder or an incentive (Uganda lender trial, NBER 2011).2
of debt collection complaints about communication tactics were about frequent or repeated calls; 34% were about contact after a request to stop (CFPB, 2024 complaints).3
Where payment reminders show up across the life of a billing account
A reminder program is not one call. It starts when the account opens and runs until the customer is current, on a plan or gone. At each stage the AI handles the routine touch and a person takes the moment that needs judgment.
Read down a column for one stage, or across a row to see what the AI and your team each do from first contact to win-back.
Which payment calls should the AI own, which should it start, and which go straight to a person?
Sort by what a wrong move costs. If the worst case is a repeat call, let the AI finish. If it touches money, terms or a customer in trouble, let the AI start and a person finish. If the law or the person's situation changes the rules, go straight to a person.
One missed bill, from reminder to paid or to a person
A worked example for a utility bill. Each step either moves toward payment or moves the customer to someone who can help. The same record follows the account the whole way.
Text with a pay link
The cheapest touch goes first. The system notes who tapped the link and who paid. Those accounts never get a call.
AI calls the non-payers
The dialer checks quiet hours, the stop list and this week's call count, then places the call in the customer's local time.
Disclose, then verify
The AI says it is an AI, that the call is recorded, and who it is for. It confirms the person with two details before it mentions an amount.
State the amount, offer two ways to pay
It reads the balance and due date and offers a pay link by text or a dated promise. It reads the date back and logs it.
Hardship or dispute: hand to a person
On "I lost my job" or "I paid already" the AI stops asking, says so, and warm-transfers with a summary, the transcript and the customer's mood.
Reminder, then kept check
A text goes out the morning of the promise. If it is unpaid the next morning, the account goes to the queue with the call history.
Start with one list: first missed payment, 3 to 10 days past due
Pick the calls where the customer most likely forgot and the cost of a mistake is smallest. Run 12 weeks, keep a control group on your current process, read every transfer, and add one thing at a time. If you do not yet text or email reminders, start there; the call is the second touch.
First-missed-payment calls, 3 to 10 days past due, one list
Most of these customers forgot, not refused. The call is short and scripted, disputes are rare, and a kept promise within 30 days is easy to count against a held-out control list.
Set the rules and the list
- Audit consent: keep only numbers the customer gave you, with no stop flag and no hardship flag
- Write the policy table: dates the AI may accept, plans it may offer, what always goes to a person
- Connect the ledger for read-back, the pay-link text, and one attempts counter shared with your staff
- Run 30 test calls with staff: angry, wrong number, "I paid", "I lost my job", "are you a robot"
Small and watched
- Release 100 to 200 accounts a day; hold out a random 30 to 50% of the list on your current process
- Listen to or read every call and every transfer in the first week
- Log each wrong amount, missed hardship cue or odd phrase, and fix the script daily
- Check that every transfer lands with a person inside 30 seconds
Whole bucket, control kept
- Open the full first-miss bucket; keep the control group
- Review all transfers plus a 10% sample of the other calls each week
- Report weekly against control: right-party contact, promises, kept promises, 30-day cure, complaints
- Test call times with your own pickup data, such as weekday evening against midday
Add one thing, then decide
- Add one only: the return-call line as overflow, or the day -3 pre-due list
- Re-check consent and quiet hours for any new state or time zone you now call
- Compare results with the exit conditions below and write down the stop lines
- Decide: widen to 11 to 30 days past due, add a language, or hold
Expand when all of these are true
- 30-day cure rate at or above the control list, and kept-promise rate at or above your human baseline, over 4 weeks
- Complaints, opt-outs and stop requests per 1,000 calls no higher than control, with zero wrong-party disclosures in a 200-call review
- At least 90% of hand-offs answered by a person within 30 seconds, with no customer asked to repeat the story
- Sampled calls get amount, date and disclosure right on 98% or more
Measure cash and cures against a control list, not calls made
Calls handled flatters any program. Count what reaches your bank account and what stays inside the law. No independent benchmark exists for most of these, so your own control list is the baseline.
Planning ranges are targets to test in a pilot, not promises. Measure your own baseline for two weeks before you change anything.
What the numbers look like for a mid-sized biller
Two levers: calls the AI finishes that a person would have taken, and return calls you miss today. Change the inputs to your own. The outcome here is a payment promise that is kept, valued below the full bill.
A worked example. Move the sliders to match yours.
AI at $0.14 a minute with volume credit applied, plus Insights seats for the people who take handed-over calls. Phone line charges are your carrier's and are not included. Illustration, not a quote. Open the full calculator
Where these numbers come from
- Person cost: BLS median for bill and account collectors is $47,030 a year (May 2025) 8. Loaded at an assumed 1.4 times wages, that is about $5,500 a month before software, seats and supervision.
- Missed share: US centers abandoned 8.9% of calls in 2024 7. Return calls spike after a reminder wave, so 12% is a planning figure; read your own abandon rate by hour.
- AI finishes 55% of calls: planning range from the split table, where five of nine call types end with the AI. Set it from your own two-week tag.
- Outcome value $120 is a planning figure, not a benchmark: use your average payment times the share that would not have paid without the call.
- Conversion 35% of recovered return calls: planning figure. Replace it with your measured PTP rate times kept rate.
What the dollar figure leaves out
- Customers who simply forgot get a polite nudge, not a collector's tone
- Your team spends its day on hardship and payment plans, not on voicemail
- Every promise, amount and date sits in one record that audit can replay
The rules that shape a payment reminder call
Whether Reg F applies depends on who is calling. Third-party collectors are covered. A lender or biller collecting its own debt is generally outside the FDCPA but not outside the FTC Act, state law or the TCPA. Not legal advice; have counsel approve scripts and limits.
AI voice needs consent under the TCPA9
The FCC treats AI-generated voices as artificial voices (Feb 2024). A call with one to a cell phone needs the called party's prior express consent. Keep the evidence: wording, date, number. Have counsel confirm the consent basis for numbers customers gave you.
Bank and health exemptions exclude billing10
The FCC's free-to-end-user exemptions for financial and healthcare messages bar billing, accounting and debt-collection content. A payment reminder cannot lean on them. It needs consent.
Stop means stop, in any words, fast10
A called party may revoke consent by any reasonable method; honor it within ten business days at most. Under Reg F, "stop calling" bars calls, not other channels. FCC changes take effect November 25, 2026, so have counsel confirm your opt-out method. Suppress the number across every campaign.
Reg F: 7 calls in 7 days, per debt11
Third-party collectors are presumed compliant at no more than 7 calls in 7 days per debt and none within 7 days after a conversation, even one the consumer started. Voicemail counts as connected; texts and emails do not.
Right party, hours and voicemail12
Before 8 am or after 9 pm local is presumed inconvenient. Generally discuss the account only with the account holder. Voicemail counts as a call under Reg F, and a message that reveals the debt risks third-party disclosure, so use a limited-content script or none.
Check for reassigned numbers13
The FCC Reassigned Numbers Database tells you whether a number was disconnected after you got consent. Safe harbor needs consent from the intended recipient, a query before the call, and a wrongly returned "No". Query before you dial.
Four questions to ask any voice AI vendor, including us
- Can you show consent evidence, and an opt-out that applies to every campaign, per call?
- Is there one attempts counter per account that AI calls and human calls both feed?
- How are card numbers kept out of the AI, the transcript and the recording?
- Do you use my call audio to train any model, and can I switch that off in the contract?
This is general information, not legal advice. Rules differ by state and country and change often. Confirm with your own counsel before you call.
Where payment calls get their facts
A reminder call is only as good as the balance, due date and consent flag behind it. VoiceGenie typically connects to these through the open API, webhooks or Zapier.
Collections and receivables platforms
- LBLatitude by Genesys
- COCollect!
- FIFinvi
- AKAktos
- MAMaxyfi
- CCCGI Credit Studio
Billing, servicing and ledgers
- FIFiserv
- JHJack Henry
- FIFIS
- SASagent
- YAYardi
- NENetSuite
Payment links and processors
- STStripe
- ANAuthorize.net
- PAPaymentus
- SQSquare
CRM, calendars and phone lines
- HUHubSpot
- SASalesforce
- ZOZoho
- TWTwilio
- PLPlivo
- TETelnyx
One platform for reminder calls, your desk and your AI tools
You do not need one vendor for the dialer, one for the AI agent, one for the agent desk and one for AI tools your staff use. Set up the ledger link, the policy table and the consent rules once. Every fix then improves both the AI and your people.
Four tools, four set-ups
own training · own reports
own training · own reports
own training · own reports
own training · own reports
One platform, set up once
- Your phone numbers and carrier
- CRM and calendar connections
- Knowledge and call rules
- Recordings, QA and analytics
Five ways payment reminder programs go wrong
Questions people ask before they start
Can AI make payment reminder calls legally?
Yes, with conditions. The FCC treats AI-generated voices as artificial voices under the TCPA, so calls to cell phones need the customer's prior express consent. You must also honor stop requests in any words, observe local calling hours, and follow Reg F if you are a third-party collector. Have counsel approve your script and limits before launch.
Is a call or a text better for a payment reminder?
Start with a text or email carrying a pay link, then call people who did not act. A 2019 McKinsey survey of delinquent card customers found most prefer email, then text, and that matching the preferred channel raised payments. A smaller group responds better to phone. Use the call as a second touch and for return calls.
How many times can you call about a late payment?
Under Reg F, a third-party collector is presumed compliant at no more than 7 calls in 7 days per debt and none within 7 days after a conversation. Lenders and billers collecting their own debts are generally outside Reg F, but state law and unfair-practices rules apply. Keep one counter per account across AI and staff and set your cap lower than the limit.
What should a payment reminder call say?
Open with who is calling and that it is an AI, that the call is recorded, and who it is for. Confirm the person with two details before stating the amount and due date. Then offer two ways to pay: a pay link by text or a dated promise. Read the date back and say a person is available any time.
What happens when a customer says they cannot pay?
The AI stops the ask, says it is sorry, and transfers to a person while the customer is still on the line. The person sees a one-line summary, the transcript, the customer's mood and what the AI already did, so the customer does not repeat the story. Hardship and payment plans outside your policy table belong to your team.
How much do AI payment reminder calls cost?
VoiceGenie AI is $0.14 per minute, billed per second, with the phone line extra and no minimum commitment. A 2.5-minute reminder call is about $0.35. Human seats for the calls that need a person are $29, $59 or $99 per person per month on annual billing. Your carrier bill and pay-link fees are separate.
How long does it take to start a pilot?
Plan on two weeks to set up and test, then a controlled start on one list. The slow part is rarely the AI. It is the consent audit, the policy table of what the AI may offer, and a read of the ledger and pay-link text. Enterprise customers can ask about done-for-you setup. Treat the timeline as a plan, not a guarantee.
Will customers hate getting a reminder from an AI?
Not if it is short, honest and easy to leave. A 2026 Gartner survey of 3,566 customers found 87% say a way to reach a human is essential when AI is used. The CFPB warns about bot loops with no exit. Say it is an AI up front, offer a person at any point, and measure complaints against a control list.
Related guides
Hear it handle your calls before you change anything.
Talk to our AI agent in your browser, or start a trial and point one line at it. Your team keeps every call that needs them.
No credit card · No minimum commitment · Keep your carrierSources
- Federal Reserve Board, Economic Well-Being of U.S. Households in 2024 (2025).
- NBER (Cadena and Schoar), Remembering to Pay? Reminders vs. Financial Incentives for Loan Payments (NBER WP 17020) (2011).
- CFPB, Fair Debt Collection Practices Act Annual Report 2025 (2025).
- BRG, Flipping the model: digital and AI-enabled collections in consumer banking (2026).
- FusionCX, KPIs for first-party collections (2025).
- LexisNexis Risk Solutions, 2025 State of Collections Study, Part Two (with Celent) (2025).
- ContactBabel, US Contact Center Decision-Makers' Guide 2025 (17th edition) (2025).
- U.S. Bureau of Labor Statistics, Bill and Account Collectors, Occupational Outlook Handbook (2025).
- Federal Communications Commission, Declaratory Ruling on AI-generated voices under the TCPA (FCC 24-17) (2024).
- Cornell Legal Information Institute (eCFR text), 47 CFR 64.1200 Delivery restrictions (TCPA rules) (2026).
- CFPB, 12 CFR 1006.14 Harassing, oppressive, or abusive conduct (Reg F) (2026).
- CFPB, 12 CFR 1006.6 Communications in connection with debt collection (Reg F) (2026).
- Federal Communications Commission, Reassigned Numbers Database (2023).
- McKinsey & Company, The customer mandate to digitize collections strategies (2019).
- CFPB, Debt collection rule FAQs (2026).
- Federal Trade Commission, Think your company's not covered by the FDCPA? You may want to think again (2015).
- CFPB, Chatbots in consumer finance (Issue Spotlight) (2023).
- Gartner, Gartner Survey Finds 87% of Customers Say Companies Using GenAI for Customer Service Must Provide Access to a Human Agent (2026).