AI + HumanContact center for debt collection

Collections where AI makes routine calls and people take the hard ones

The AI makes the reminder, balance and promise-to-pay calls, on count, on time. Collectors keep hardship, disputes and terms, so more accounts cure and fewer customers file complaints.

  • Confirms who it is talking to
  • Can say it is an AI, offers a person
  • Hardship goes straight to a person
Debt collection: Tuesday 6:12 pm · early stage, 12 days past dueAI agent on the line
AI agentConfirms who it is talking to
OutcomeHardship flagged · warm transfer to Dana with summary · 3 of 7 weekly calls logged

Payment reminder

12 days past due: the AI states the amount and offers a pay link.

Promise-to-pay follow-up

The AI calls the day before and after the date the customer chose.

Hardship call

Hours cut: the AI stops the ask and hands the call to a collector.

Dispute or attorney

The claim is logged, the pressing stops, and a person takes over.

01 · Why AI + Human

Why collections phones will run on AI and people together

Early-stage collections is repeat work: payment reminders, balance questions and promise follow-ups, with daily call quotas and a call cap per person. Delinquency stays high, and 51% of 2024 CFPB call-tactic complaints were about frequent calls. A people-only floor cannot reach every account on time. An AI-only floor cannot hear hardship, a dispute or a lawyer.

AI agentsLow cost per call, instant scale, consistent
  • Calls every account in the window, including Sunday at 8 pm
  • Reads the amount and due date from your system, the same way each time
  • Honors a stop request and logs the call against the cap at once
On its ownIt cannot weigh a job loss, judge a dispute or agree terms outside your plan table.
Your peopleJudgment, authority, care, new solutions
  • Hears hardship and finds a payment plan that fits the month
  • Weighs proof in a dispute and decides what to do next
  • Holds the authority to settle, waive or extend
On their ownA collector cannot call every account on time, and tires of reading balances aloud all day.
The AI clears the routine so collectors can solve the hard accountsOne desk, one record of every call
  • Every reminder and promise follow-up is made, so fewer accounts roll to the next bucket
  • Maria's hours-cut call reaches Dana in the same call, with the story already told
  • One counter shows every call to a person about a debt, so caps are checked, not guessed
Side by side
People only (traditional)
AI only
AI + Human
Reach on early-stage accounts
Limited by hours and call quotas
Every account, every window
AI reaches all; collectors take the replies
Cost per routine call
Highest: a collector reads the balance
Lowest: billed per second
Low for routine, spent on people where it counts
Consistency and call caps
Varies by collector and by day
Same script and cap check every call
Same checks, plus review of every transfer
Hardship, disputes, terms
Strong: empathy and authority
Weak: spots the cue, cannot decide
AI spots the cue, collector decides
Kept promises and cures
Good on the accounts they reach
Good on routine, poor on hard cases
Best when measured against a held-out group
How to think about the split
Split calls by cost of a wrong move

If the worst result is a repeat call, let the AI finish. If it is money, terms or a customer in trouble, let the AI start and a collector finish.

Make the exit to a person instant

Hardship, dispute, attorney and bankruptcy cues should end the ask and bring in a collector in the same call. Test it before launch.

Judge both on kept promises

Do not score calls handled. Score verified contacts, kept promises and cures for AI and collectors on the same accounts and hours.

The short answer

How can lenders and collection agencies use AI on collection calls without raising compliance risk?

Let the AI make and answer the routine early-stage calls: payment reminders, promise-to-pay follow-ups and call-backs. It says it is an AI, confirms the right party before naming any debt, and records the promise. Send hardship, disputes, attorneys and negotiation to a collector with the full call. Start with one inbound call-back line, and measure kept promises, not calls handled.

Reviewed November 2026 · 18 sources cited below
  • The AI confirms the right party with two data points before it names a debt
  • Your rules set the caps: 7 calls in 7 days is Reg F's presumption line, not a target
  • Hardship, dispute, attorney and bankruptcy cues go to a person in the same call
  • Judge it on kept promises and cures, not on calls handled
02 · The stakes

Why early-stage calls are worth getting right

Early calls decide whether an account cures or rolls to the next bucket. They are also where complaints begin. A call to the wrong person, or one call too many, costs more than a call that never connected.

5.0%

of US household debt was in some stage of delinquency in Q3 2026, out of $19.1 trillion in total (New York Fed)1

160,200

US bill and account collectors in 2025, median pay $47,030, with daily call quotas; jobs projected to fall 10% by 2035 (BLS)2

51%

of 2024 CFPB communication-tactic complaints were about frequent or repeated calls; 34% about contact after a request to stop3

60%

of 2024 CFPB 'debt not owed' complaints said the debt was not theirs, the mistaken-identity risk any outbound list carries3

03 · The whole journey

The borrower's path from first bill to cured account, with AI and collectors

Most accounts never need a hard call. The AI keeps contact steady from the first payment to the last, and a collector steps in when money, terms or a hard situation is on the line.

Before the due dateSTAGE 1
Repayment setupSTAGE 2
Early past dueSTAGE 3
Promise to paySTAGE 4
Hardship and disputesSTAGE 5
Cured and currentSTAGE 6
Charged-off or lapsedSTAGE 7
The moment
Borrower forgets the date or does not know how to pay; missed first payment.
New borrower is unsure of autopay, due dates or the payoff amount.
Account is 1 to 30 days late; the borrower avoids calls or is unreachable.
A date is promised, then lost; no reminder, so the promise breaks.
Borrower lost income or disputes the debt, and gets pushed instead of heard.
Borrower pays up, then hears nothing and wonders whether the account is clear.
Old account goes quiet after repeated unanswered calls; contact data is stale.
AI agent
Calls or texts a day before, states the amount, sends a pay-by link.
Answers due date and balance questions, confirms contact consent, enrolls in autopay.
Confirms the right party first, then states the amount and records the promise.
Calls the day before and the day after the date, and logs paid or missed.
Stops the ask, logs the reason, and hands the call over with a summary.
Confirms the account is current and answers payoff and receipt questions.
Re-checks the number, honors stop requests, and offers a call-back when allowed.
Your team
Reviews accounts flagged as likely to miss and sets the contact rules.
Handles setup questions about terms, fees or a changed due date.
Takes calls where the borrower objects, asks for terms or wants a person.
Follows up on broken promises and decides whether to extend or escalate.
Listens, offers a plan inside policy, handles disputes and attorney calls.
Reviews accounts that cured then rolled again and looks for the cause.
Negotiates settlements and decides which accounts to place or close.
What to track
On-time first payment rate
Autopay enrollment, consent on file
Right-party contact rate
Kept promise rate
Plans set up, complaints per 1,000
Cure rate, re-delinquency rate
Re-contact rate, settlements agreed

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.

04 · AI or person

Which collection calls should the AI own, start, or never touch?

Sort calls by what a wrong move costs. If the cost is a repeat call, let the AI finish. If it is money, terms or a customer in trouble, let the AI start and a collector finish. If the law or the person's situation changes the rules, go straight to a person.

AI handles it end to endAI starts, a person finishes with the contextStraight to a person
Call type
Volume
Who takes it
Why
Early-stage payment reminder (outbound)
"Your $386 payment from the 3rd is open. Can you pay today?" 1 to 30 days past due, right party confirmed.
high
AI handles it
Short, scripted, one outcome. The AI states the amount, offers a pay-by link or a date inside policy, and records the answer.
Promise-to-pay reminder and kept check
"You said Friday. Want the pay link now?" Calls the day before and the day after the promised date.
high
AI handles it
The customer already agreed. Low conflict, high value: a kept promise stops the account rolling to the next bucket.
Return calls: balance, due date, how to pay
"Someone called me about my account." Verify, read the amount, pay by link or choose a date.
high
AI handles it
The customer reached out, so the right party is likely. Voicemail on this line is a lost contact. Answer on ring one, any hour.
Wrong number and "stop calling me"
"This isn't Maria's phone." "Take me off your list." Apologize, give no detail, flag the number, honor the request.
medium
AI handles it
Speed and exactness beat judgment. The AI suppresses the number or channel at once; a person reviews the log.
Lower payment or longer plan than policy
"I can do $150 a month, not $386." The offer sits outside your pre-approved plan table.
medium
AI, then a person
The AI records the offer and the reason. A collector with authority decides. The AI never improvises terms.
Hardship: job loss, illness, disaster
"My hours got cut." "I'm off work after surgery." The AI stops the ask, says sorry, and brings in a collector.
medium
AI, then a person
The caller needs options and empathy, and forbearance rules are yours. The AI spots it and hands off in the same call.
Dispute: "not my debt" or wrong amount
"I paid this in May." "I never opened this account." The AI logs the claim and time, stops pressing for payment.
medium
AI, then a person
Disputes need proof and a decision. The AI captures the facts so the collector starts at the evidence, not at hello.
Settlement and large or charged-off balances
"Would you take half to close it?" Or any balance above your review threshold.
low
Straight to a person
It trades money for authority and risk. Collectors with a settlement matrix lead, using the AI's summary.
Attorney, bankruptcy, deceased, distress
"Call my lawyer." "I filed Chapter 7." "My husband passed." "I can't go on."
low
Straight to a person
Contact rules change or a person must act now. Reg F routes represented consumers through their attorney.
TakeawayThe AI finishes the first four rows end to end and starts the next three. A person takes the last two straight away. Tag two weeks of your own calls with these nine labels before you set any target.
05 · One call, start to finish

One early-stage call, step by step

A lender calls a customer 12 days past due. This is what happens between the dialer and the collector, and what each side knows at each step.

Before dialing

Pre-dial checks

Consent on file, no stop request, local time 6:12 pm, 2 of 7 weekly calls used, number checked against the reassigned numbers database.

0:00

Opens with who, what and recording

"This is an AI assistant calling for Maria Alvarez from Lakeshore Auto Finance. This call is recorded." No debt is named yet.

0:20

Confirms the right party

Date of birth and the last four of the loan must both match. If someone else answers, the AI shares no detail and ends the call politely.

0:45

States the debt, offers approved options

Gives the required debt-collection statement, reads back the amount and due date, and offers pay-today by link or a date inside policy.

1:40

Hardship heard, person brought in

Maria says her hours were cut. The AI stops the ask and transfers to Dana with a one-line summary, the transcript, Maria's mood and what was said.

After the call

Record, counter and follow-up

AI Assist drafts the disposition and call-back time for Dana to accept. The call is logged against the weekly cap and QA-scored.

06 · Start small

Start with the collections call-back line, on overflow and after hours.

Pick the calls where the customer already reached out. They are the likeliest right-party contacts you will get, and they add no outbound dialing risk. Run 12 weeks, read every transcript, and add one thing at a time.

Start with this

Call-backs to your collections line, on overflow and after hours

The caller picked up the phone, so right-party contact is likely and no outbound consent or calling-hour rule is in play. Voicemail on this line is a lost contact. It builds verification, hand-off and review habits before any outbound list.

How calls reach the AI

Your number and phone system stay as they are. After four rings, when the queue passes 60 seconds, or after 6 pm, your PBX or carrier forwards the call to a VoiceGenie number.

Weeks 1-2

Baseline and rules

  • Log 30 days of the line by hour: answered, abandoned, voicemail, call-back attempts
  • Tag top call reasons and write the do-not-answer list: dispute, hardship, attorney, bankruptcy, deceased
  • Write the policy table the AI may offer: pay in full, pay-by dates, plan sizes; counsel signs the script
  • Decide how AI call logs feed your 7-in-7 counter and stop lists
Weeks 3-4

Staff test

  • Staff place 60 scripted calls; 25 are traps: wrong person, spouse answers, 'are you a robot', dispute, bankruptcy
  • Read every transcript; fix wrong amounts, long pauses and over-explaining
  • Test hand-off in hours and the next-morning queue after hours; test the pay link end to end
  • Test 'person' and 'stop calling' in the first sentence; both must work every time
Weeks 5-8

Live on overflow and after hours

  • Turn on forwarding for no answer in four rings, queue over 60 seconds, and after 6 pm
  • Review every call in week 5, then a 20% sample plus every transfer, dispute and stop request
  • Hold a 30-minute weekly review with the collections and compliance leads
  • Post five numbers weekly: answered, verified, kept promises, transfers, errors
Weeks 9-12

Decide and add one thing

  • Compare results with the baseline and the exit conditions below
  • If met, add one: a 'make a payment' IVR option, or reminders to one list of customers who already made a promise
  • Confirm consent evidence and the reassigned-numbers check before any outbound list goes live
  • If not met, change the script or stop; no other process depended on the AI

Expand when all of these are true

  • Zero debt details spoken before verification and zero missed dispute, stop, attorney or bankruptcy cues, in a full audit
  • Verified callers pay or promise at a rate at least equal to your staffed line in the same hours
  • Abandoned and voicemail calls in covered hours cut at least in half against the baseline weeks
  • At least 9 in 10 sampled hand-offs reach a collector with the right summary; no caller repeats the story
07 · What to measure

What to measure, and how to define it

Pick one denominator for each rate and keep it. A right-party rate of 60% of connects can be 15% of attempts on the same day. Compare AI and collectors on the same accounts and hours, and keep a held-out group.

Metric
Typical today
Planning range with AI and people
Right-party contact (RPC) rate4
Verified account holders reached, divided by dial attempts. Also report per connect. Verified means two data points matched.
Planning range: 15 to 30% of attempts, early stage
+10 to 15% in pilots (BRG client results)
Promise-to-pay (PTP) rate
Promises divided by right-party contacts, not dials. Count only a dated amount the customer stated.
Planning range: 30 to 50% of right-party contacts
Match your staffed line, same segment
Kept-promise rate
Promises paid by the promised date plus grace days, divided by promises due. Read it at day 0, 3 and 7.
Planning range: 50 to 70% kept
65 to 80% with a reminder
Roll rate and cure rate1
Roll: accounts moving to the next past-due bucket. Cure: accounts back to current. Compare with a held-out group.
Cards, flow into 90+ days: 6.87% annualized, Q3 2026
Set against a control group, not a number
Cost per dollar collected5
All cost of the treated segment (AI minutes, seats, telephony, collector time) divided by dollars collected.
Rises as accounts age; no public US benchmark
Below staffed line, same segment
Cap and stop-request breaches3
Calls to one person about one debt above your cap (Reg F presumption line: 7 in 7 days), plus any contact after a stop request.
51% of 2024 CFPB contact complaints: call frequency
Zero above cap; zero after a stop request
Wrong-party disclosures3
Calls where debt detail reached anyone but the consumer. Audit every call with a wrong-party or no-answer cue.
56% of 2024 CFPB third-party complaints: debt discussed
Zero
Abandon rate on the call-back line6
Callers who hang up before reaching a person or the AI, divided by calls offered, by hour. Voicemail counts as a miss.
8.9% across US centers in 2024
Under 3% in covered hours

Planning ranges are targets to test in a pilot, not promises. Measure your own baseline for two weeks before you change anything.

08 · The return

What a call-back line is worth

Two numbers drive the case: the calls you miss today and how many of them become kept payments. Labor saved is real but second. Replace the planning values with your own line's numbers.

A worked example. Move the sliders to match yours.

6 minaverage call length
$5,600loaded monthly cost per person
55 a daycalls one person handles
$200 · 30%value of one kept promise to pay · share of recovered calls that become one
Estimated monthly gain
$74,764a month
Today, per call
$4.07
With VoiceGenie
$1.68
1,080more calls answered a month
324extra kept promise to pays a month
$64,800value of those outcomes
2,700calls your team still takes, with context

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

  • Cost per person: BLS median pay for bill and account collectors is $47,030 2. Adding about 43% for benefits and taxes (estimate) gives roughly $5,600 a month, before seats, software and management.
  • Calls per person per day: 55 assumes a 6-minute call, 80% occupancy and about 7 productive hours. A planning figure; use your own.
  • Missed 18%: US centers abandoned 8.9% of calls in 2024 6. Call-back lines with voicemail after hours run higher. Planning range 10 to 25%.
  • Outcome value $200: a typical early-stage payment is about $340 (planning), cut because some customers would have paid later anyway. This is cash, not margin.
  • AI finishes 55%: planning range 40 to 60% on call-back lines (balance, pay link, wrong number, in-policy plan). No independent public benchmark exists.

What the dollar figure leaves out

  • Collectors spend the day on hardship and disputes, not on reading balances aloud
  • Every call is recorded, transcribed and QA-scored, so audits start from all calls, not a sample
  • Customers can settle up at 8 pm on a Sunday instead of waiting for Monday
09 · Rules to get right

The rules to get right, and what to ask any vendor

Reg F binds third-party collectors. Lenders collecting their own debts sit outside the FDCPA but not outside the FTC Act, state law or the TCPA, so most follow the same standard. This is not legal advice; have counsel approve the script.

US

7 calls in 7 days, and a 7-day pause7

Reg F presumes a collector complies if it places no more than 7 calls about a debt in 7 days, and none within 7 days after a phone conversation about it. Exceeding it is a presumed violation. Count per person, per debt, across AI, collectors and vendors.

US

Voicemails count; texts need an opt-out8

A call that reaches voicemail counts toward the cap even if no message is left. A limited-content message may carry only your business name, a call-back number, a request to reply and a person's name. Texts need a simple opt-out.

US

Right party first; hours; attorneys9

Do not discuss the debt with anyone but the consumer, spouse, attorney or creditor. Before 8 am or after 9 pm local time is presumed inconvenient. If you know the consumer has an attorney, you generally contact the attorney.

US

Debt-collector statement and validation notice10

The first communication must say it is an attempt to collect a debt; later ones must say they come from a debt collector. Validation information goes out in the first communication or within 5 days, opening a 30-day dispute period.

US

AI voice calls need consent; stop means stop11

The FCC ruled in February 2024 that AI-generated voices are artificial voices under the TCPA, so autodialed or prerecorded calls to mobile numbers need prior express consent. Honor a reasonable revocation within a reasonable time, up to 10 business days. Confirm current text with counsel.

US states

Some states and countries set tighter caps12

Massachusetts allows two calls or texts per debt in 7 days to a debtor's personal phone and covers creditors collecting their own debts. Australia's ACCC guidance says no more than 3 calls a week or 10 a month. Set caps by the consumer's location.

Four questions to ask any voice AI vendor, including us

  1. Does the AI say it is an AI and that the call is recorded in its first sentence, and offer a person at any time?
  2. How do the AI's call logs reach our call counter, stop lists and state caps, and how fast?
  3. Is our call audio used to train any model, and where is it stored and for how long?
  4. What does the AI do on a dispute, attorney, bankruptcy or hardship cue, and can we see it in the transcript?
10 · Where it plugs in

The systems a collections call touches

Typically connected through the open API, post-call webhooks or Zapier, not as native integrations. The AI reads from and writes to your system of record. It does not replace it.

Collections systems of record

  • LBLatitude by Genesys
  • COCollect!
  • FDFICO Debt Manager
  • CCCGI Credit Studio
  • FIFinvi
  • MAMaxyfi

Loan servicing and core

  • FIFiserv
  • JHJack Henry
  • FIFIS
  • IMICE Mortgage Servicing (MSP)
  • SASagent

Dialers, carriers and number hygiene

  • Your existing dialer or ACD
  • TWTwilio
  • PLPlivo
  • TETelnyx
  • VOVonage
  • FRFCC Reassigned Numbers Database

Payments and CRM

  • STStripe
  • ANAuthorize.net
  • PAPaymentus
  • SASalesforce
  • HUHubSpot
  • ZOZoho
Keep your numbers and carrierTwilio, Plivo, Telnyx, Vonage or any SIP trunk or PBX. Put the AI behind one menu option, on overflow or after hours.
CRM and calendarsHubSpot, Salesforce, Zoho and Google connect directly.
Everything elseThe systems above connect through the open API, post-call webhooks or Zapier.
11 · One platform

One platform for the dialer, the AI agent and the collector desk

You do not need one vendor for the dialer, another for the AI agent, another for the collector desk and a fourth for the tools that help collectors. Set it up once. Each input, such as a connection to your collections system, a policy article, a corrected answer or a QA finding, improves the AI and your collectors together.

The usual way

Four tools, four set-ups

Dialerown contract · own integration
own training · own reports
Voice AI agentown contract · own integration
own training · own reports
Agent deskown contract · own integration
own training · own reports
Agent assist and QAown contract · own integration
own training · own reports
With VoiceGenie

One platform, set up once

AI agents
AI-native dialer
Human Agent Desk
AI Assist
Shared by the AI and your people
  • Your phone numbers and carrier
  • CRM and calendar connections
  • Knowledge and call rules
  • Recordings, QA and analytics
One connection to your system

Connect your collections system of record once. The AI and the collector see the same balance, due date and notes.

One policy and script base

Plan tables, disclosures and do-not-answer rules are written once. The AI answers from them and AI Assist shows them to collectors.

One recording and QA view

Every AI and collector call is recorded, scored and shown together. One QA finding fixes the script and coaches the person.

12 · What goes wrong

Five mistakes that sink collections AI projects

Starting with a cold outbound list of old numbers
Start with calls customers place to you. Before any outbound list, confirm consent evidence and check numbers against the FCC reassigned numbers database.
Counting calls in the dialer only
Voicemails count and a conversation starts a 7-day pause. Keep one counter per person per debt that AI, collectors and outside vendors all write to.
Scripting the AI to push through hardship or a dispute
The CFPB has warned about bots that trap people with no way to a human. Make hardship, dispute, attorney and bankruptcy cues stop the ask and bring in a person.
Judging success by calls handled or containment
Containment counts hang-ups as wins. Track verified contacts, kept promises, cures and complaints against a held-out group.
Letting the AI say more than the policy table
In 2024, 91% of CFPB false-statement complaints involved an incorrect amount. The AI reads amounts and dates from the system and offers only terms you pre-approved.
13 · Questions

Questions people ask before they start

Is it legal to use AI on debt collection calls?

Yes, but the same rules apply as for human calls, plus more. Reg F's call caps, hours, disclosures and right-party rules still bind third-party collectors, and the TCPA treats AI voices as artificial voices that need consent for mobile numbers. Lenders collecting their own debts still face the FTC Act and state law. Have counsel approve the script.

Does the FDCPA apply if we collect our own loans?

Generally no, because the FDCPA covers third-party collectors. But the FTC says the FTC Act still bars deceptive or unfair practices, including revealing a debt to others, and a creditor collecting under another name can be treated as a collector. State laws, such as Massachusetts's, can reach creditors. Most lenders follow Reg F anyway.

What should an AI say at the start of a collection call?

State who is calling, that it is an AI, and that the call is recorded. Ask for the person, then confirm identity with two data points before naming the debt. After that, give the required statement that this is an attempt to collect a debt. If someone else answers, share no detail and end politely. Offer a person at any time.

How much does AI collections calling cost?

VoiceGenie charges $0.14 per minute for AI calls, billed per second, and the phone line is extra. A six-minute call is about $0.84 in AI time. Human seats are $29, $59 or $99 per person per month on annual billing. There is no minimum commitment and there is free trial credit, so a one-line pilot is cheap to test.

How long does a pilot take, and where should it start?

Plan 12 weeks. Weeks 1 and 2 set the baseline and rules, weeks 3 and 4 are a staff test, weeks 5 to 8 run live, and weeks 9 to 12 decide. Start with your call-back line on overflow and after hours, because the caller already reached out. Add one thing at a time, such as a payment option or one reminder list.

Will borrowers hang up on an AI collector?

Some will, and some will ask for a person at once, so make that request work every time. The CFPB has warned about chatbots that leave people stuck with no route to a human. Keep calls short, state the amount, offer a way to pay now, and track hang-ups and requests for a person every week.

How does the hand-off to a collector work?

When the caller raises hardship, a dispute or a request outside policy, the AI transfers the call to the right collector queue. The collector sees a one-line summary, the full transcript, the caller's mood and what the AI already did, so the caller does not repeat the story. AI Assist then drafts the disposition and call-back time for one-click acceptance.

What is a good right-party contact rate?

It depends on the denominator and the list. Some teams count right-party contacts per dial, others per connected call, and the same day can read 15% or 60%. Fix one definition, confirm identity with two data points, and compare with your own staffed baseline on the same accounts and hours. Public US benchmarks are thin, so treat ranges as planning figures.

Try it on one call type

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 carrier

Sources

  1. Federal Reserve Bank of New York, Household Debt and Credit Report, Q3 2026 (press release) (2026).
  2. US Bureau of Labor Statistics, Bill and Account Collectors, Occupational Outlook Handbook (2025).
  3. Consumer Financial Protection Bureau, Fair Debt Collection Practices Act Annual Report 2025 (2025).
  4. Berkeley Research Group, Flipping the model: digital and AI-enabled collections in consumer banking (2026).
  5. LexisNexis Risk Solutions, 2025 State of Collections Study (with Celent) (2025).
  6. ContactBabel, US Contact Center Decision-Makers' Guide 2025 (17th edition) (2025).
  7. Consumer Financial Protection Bureau, 12 CFR 1006.14 Harassing, oppressive, or abusive conduct (Regulation F) (2026).
  8. Consumer Financial Protection Bureau, Debt collection rule FAQs (2026).
  9. Consumer Financial Protection Bureau, 12 CFR 1006.6 Communications in connection with debt collection (Regulation F) (2026).
  10. Consumer Financial Protection Bureau, 12 CFR 1006.18 Additional prohibitions on harassment or abuse, false or misleading representations, and unfair practices (2026).
  11. Federal Communications Commission, Declaratory Ruling on AI-generated voices under the TCPA (FCC 24-17), Commissioner statement (2024).
  12. Massachusetts Attorney General, 940 CMR 7.00 Debt collection regulations (2017).
  13. Consumer Financial Protection Bureau, 12 CFR 1006.34 Notice for validation of debts (Regulation F) (2026).
  14. Legal Information Institute, Cornell Law School (eCFR text), 47 CFR 64.1200 Delivery restrictions (TCPA rules) (2026).
  15. Federal Communications Commission, Reassigned Numbers Database (2023).
  16. Federal Trade Commission, Think your company's not covered by the FDCPA? You may want to think again (2015).
  17. Consumer Financial Protection Bureau, Chatbots in consumer finance (Issue Spotlight) (2023).
  18. Australian Competition and Consumer Commission, What debt collectors can and can't do (2026).