AI agents for collections

An AI agent for collections that works the early book the way your best collector would — voice, WhatsApp, SMS, and email — then writes the result into the system you already run.

VoiceWhatsAppSMSEmail
LIVEMaria
00:00
IdentityRight personNegotiatePlanResult
MariaGood morning, am I speaking with Ana?
CustomerYes, that’s me.
MariaYou have a balance of US$2,400. We can settle today or set a date.
CustomerI can’t today. This month is tight.
MariaThen a partial on Friday and the rest in 15 days. Does that work?
CustomerFriday works for the partial.
Promise to payFriday · in 7 days

They trust us

CasheaIntercobrostu KompaUnoMotorsMattildaKalicollectRefermat

Results

9×capacity vs. the team it replaced
+10%recovery lift on live books
−30%cost per action
70%+early work kept in-house

What it does

Notify, negotiate, follow through

Same case across voice, WhatsApp, SMS, and email. Files in. Dispositions out.

Campaigns

Preventive · personal loans

0 worked
AccountAmountDateChannel
4821 · Ana R.US$2,4002026-03-12
Thinking
Voice
3 retries · 09–12
1904 · Luis M.US$8902026-04-02
Thinking
WhatsApp
2 retries · WhatsApp first
7730 · Carla V.US$5,1002026-01-18
Thinking
Voice
4 retries · 18–20
2266 · Diego P.US$3202026-05-20
Thinking
SMS → voice
2 retries · SMS → Voice
WhatsAppConfirmation sent
Confirmation: partial on Friday. I’ll send a reminder the day before.
Ok, thanks.

Promise to pay

Partial · Friday

Next action

WhatsApp reminder · Thursday

Why

PathPilot

Four things a collections leader can see on day one — not a category page.

Multichannel collections

The same case continues on voice, WhatsApp, SMS, and email. Your agents follow up on any channel. The borrower does not start over.

Ana R. · US$2,400

  1. VoiceIdentity and a plan
  2. WhatsAppConfirm the partial
  3. SMSReminder the day before
  4. EmailReceipt after payment

Promise to pay · same file

Learning

Agents learn from your supervisors’ feedback and from what actually recovered. The next contact uses the last result — not a static script.

Next-contact quality

Supervisor marks a weak offer
The book shows what recovered
The next call uses that lesson

0%Next-contact quality

LATAM and Latinos in the US

Built for the hardest books in this population: regional voices, WhatsApp-first habits, and objections generic bots miss.

LATAM · US

MexicoPeruEcuadorCentral AmericaUS Spanish

Technology transfer

We operate the agent first. Then we hand your team the keys, the policy, and the audit — so the capability stays with you.

PathPilot and a lender’s internal AI team working together at the same desk

Testimonials

UnoMotors

We have only used PathPilot on the hardest book, about S/400,000 we were about to write off. With Jessica we started booking around S/30,000 in agreements a month: 7.5% monthly. We cut the collections team by 70%.

Henrry Rojas Collections coordinator, UnoMotors

Read the story

Mattilda

Before, analysis for continuous improvement in collections and service was manual, which made it very hard to find scalable actions with immediate impact on results. With PathPilot’s Quality Agent, we now know exactly where to focus to improve.

Christelle Falcon C. Mattilda

Intercobros

With PathPilot, scale no longer depends on hiring, training, and supervising more people. We increased management capacity and reduced operating cost, without growing headcount at the same pace.

Fernando Bastidas Intercobros

tu Kompa

Reminders and early overdue are now all handled by AI. That makes our operation more efficient and faster.

Felipe, founder of tu Kompa

Kalicollect

With PathPilot virtual agents we can monitor the operation, evolve it, and take on new books with confidence. As we say: no fear of success.

Adriana Chacón, Kalicollect Solutions

Cost, efficiency, and early-stage

Reduce cost to collect

Lower cost per collection action — about −30% on live books — without sending more of the early book to an agency.

How to cut collections cost

Improve collections efficiency

More accounts, same team. 9× capacity versus the floor the agent replaced. Collectors take the cases that need a person.

How to improve efficiency

AI for early-stage collections

Preventive and first-miss work stays with you. 70%+ of early work kept in-house. Agencies take late and complex files.

Early-stage collections

For lenders and fintechs

Consumer and SME credit — voice and WhatsApp — not invoice DSO software. Banks, fintechs, and agencies.

AI collections for lenders

Best voice agent for collections

Proprietary contact analysis picks the hour. Regional accent. Your phone numbers and call center.

Voice agent for collections

Questions from collections leaders

What is an AI agent for collections?

An AI agent for collections contacts accounts across voice, WhatsApp, SMS, and email, negotiates payment plans inside your policy, logs promises to pay, and follows up. It is not a reminder blast or an IVR. PathPilot’s Collections Agent is trained on your scripts and writes results into your CRM or loan system.

How is this different from a generic voice bot?

A generic bot is a prompt and a minute rate. PathPilot configures and fine-tunes the agent with you, analyzes every contact to improve the next one, prices successful outcomes, and can transfer operation to your internal AI team.

Can we use it before sending accounts to an agency?

Yes. The Collections Agent is built for preventive and early-stage collections so you keep the customer relationship. Agencies still make sense for late or already-lost accounts.

Will our own AI team be able to run it later?

That is the point of technology transfer. We operate and improve the agent first, then train your team on policies, offers, objections, and flows, and connect to the models and infrastructure you already approve.

How does PathPilot price the Collections Agent?

You pay for the result of the engagement — a successful collections action — not for minutes or tokens. Non-effective contact and no-connect use a lower rate. Channels share the same prices.

What does the agent write back to our systems?

Dispositions, promises to pay, next actions, and the conversation record. Files in and files out, or APIs and webhooks into the CRM or core you already run. We do not replace your servicing stack.

How do we reduce cost to collect?

Pay for successful collections actions instead of idle minutes, keep early work in-house, and send only late or complex files to an agency. Live books show about −30% cost per action. Guide: https://www.getpathpilot.com/blog/reduce-collections-cost/.

How do we improve collections efficiency?

Work more accounts per collector, contact earlier, and stop restarting the case on every channel. Operators see about 9× capacity versus the team the agent replaced. Guide: https://www.getpathpilot.com/blog/improve-collections-efficiency/.

What is AI for early-stage collections?

An agent that works preventive and first-miss accounts on the lender’s channels so the file does not have to go to an agency for its first serious attempt. PathPilot is built so 70%+ of early work stays in-house. Guide: https://www.getpathpilot.com/blog/ai-early-stage-collections/.

Is this AI collections for lenders and fintechs?

Yes. PathPilot is for banks, fintechs, and collections agencies — consumer and SME credit — not invoice or accounts-receivable software.

Keep reading

Run the early book with an agent you can take over

We train it. You keep the policy, the audit trail, and the option to run it yourselves.