Medra

Commercial flow 07 · Collections

Recovery with criteria, not pressure.

The agent contacts each balance by age and payment history, offers the options you defined and records every commitment. It stops when it should.

Live · Phone call07 / 07
Outgoing call · agentDaniel Ruiz · +57 310 •••• 4471Ended
AgentHi Daniel. I am calling about your outstanding balance of $18,400.
DanielI cannot pay all of it this month.
AgentYou can split it into two installments with no surcharge.
DanielTwo installments works.
As agreed on the call, here is the payment link for your first installment.
Commitment recorded · 2 installments
Act one · the problemPeople don’t pay messages. They pay conversations.

Most collections strategies treat a portfolio as one big number.

Everyone gets the same message, usually some mix of reminder and threat, and debtors learned a long time ago to ignore it.

And nobody really knows why people aren’t paying. A forgotten date? A lost job? A problem with the service? A deliberate choice? Without that answer there’s no offer that makes sense. Call centers, meanwhile, are expensive, hard to scale and almost impossible to audit.

Every debtor, placed by days past due and amount owedRecovery this cycle · 9%The same message, to everyone, at the same time030 days60 days120 days past dueAmount owed →A forgotten date, a lost job, a service dispute and a deliberate choice all got the same reminder-and-threat — and learned to ignore it.Illustrative portfolio
The cost of the status quoCollections
  • One messageThe same reminder-and-threat for the whole portfolio.
  • No whyNobody knows the reason, so no offer makes sense.
  • Call centersExpensive, hard to scale, almost impossible to audit.
Act two · the solutionWhat Medra does

Every debtor is a person with a reason. Every portfolio is a set of cohorts with a strategy.

Debt is segmented by age and amount, WhatsApp and calls work toward the same goal, the reason for non-payment is captured on every contact, and offers stay inside the policy you set.

Overdue portfolio by age of debtRecoveredArrangement or payment, in the thread1–15 days16–30 days31–60 days60+ days82% recovered64% recovered47% — mostly as payment plans26% — escalated or written offAsked not to be contacted — closed immediately, no further touchRecovery with criteria, not pressureIllustrative cycle
Aging buckets, one cycleLimits configured by your finance team

How Medra runs this flow, step by step

  1. Cohort the portfolio.

    Debt is segmented by days past due and amount owed, and each cohort gets its own priority and level of effort.

  2. Sequence the touches.

    WhatsApp and calls work together, and every touch has a job: remind, understand, offer, close.

  3. Understand before you ask.

    The agent finds out why the person isn’t paying and records the reason, so every conversation teaches you something about your portfolio.

  4. Offer something worth saying yes to.

    A concrete benefit for getting current, cheap for you and valuable for them. Payment plans, discounts or new dates, always within the policy you set.

  5. Escalate with care.

    Disputes, conflicts and vulnerable situations go to a person, following your policy.

  6. Learn from every call.

    Calls are listened to and analyzed at scale, and those insights shape the next campaign. Campaign, calls, listening, insights, optimization. And again.

Under the hood

  • Cohort engineAging and amount turned into priorities and playbooks.
  • Voice agents with natural voicesCalls that sound like a person, at the scale of software.
  • Multichannel sequencesWhatsApp and phone working toward the same goal.
  • Non-payment reason classificationThe why behind every overdue account, captured and counted.
  • Policy-based offersFlexible within the limits you define, never outside them.
  • Conversation analytics at scaleEvery call becomes data for the next one.
Act three · the impactWhat changes

Higher recovery, lower cost, and customers who are still customers after they pay.

Figures are illustrative, written to look and feel real. Before publishing they are replaced with a client result or kept labeled as illustrative — and we tell you in the first session what your own baseline says.

  • Recovery rate

    9%→27%

    Baseline, then the same metric with the flow running.

  • Kept payment promises

    35%→68%

    Baseline, then the same metric with the flow running.

  • Cost per dollar recovered

    USD 0.18→USD 0.06

    Baseline, then the same metric with the flow running.

MetricBefore MedraWith Medra
Message strategyOne message for the whole portfolioA strategy per cohort, a conversation per person
Effective contact rate22%61%
Understanding of non-paymentUnknownReason captured in 85% of contacts
The customer after payingBurnedStill a customer

Escalation triggers in this flow

Payment dispute raised·Hardship or vulnerability declared·Legal or regulatory language·Arrangement beyond configured limits·Explicit request to stop contact

What it moves

  • Recovery rate
  • Kept payment promises
  • Cost per dollar recovered

The money back, and the customer still there.

Why Medra is differentThe part nobody else can copy

The Turing Test, passed in production.

Nobody is honest with a dunning robot. At Medra the test runs on every overdue account, on the phone and on WhatsApp, in more than 19 countries, with debtors who have real reasons and no idea they are the judges.

97

out of 100 people can’t tell Medra from a human.

The other 3 catch us on one detail: we reply at 3 am in four seconds.

The moment a debtor realizes it’s a bot, they hang up or promise a date they won’t keep. When it feels human, they tell you why they didn’t pay, when they can and what they can afford this month. That honesty is what separates a recovered account from a lost customer.

How we do it in Collections

There’s no single trick. Six mechanisms work together, each one closing a gap that gives other bots away. These three carry the most weight in this flow.

  1. Natural voice on callsRealistic voices that handle interruptions, pauses and people talking over each other.
  2. Local voiceTone, formality and regional variants of each language adapt to the country and to the person on the other side.
  3. Relationship memoryEvery touch remembers the ones before it, in any channel. The customer never starts from zero.

Medra next to the usual options

 Typical chatbotTypical call centerMedra
Response timeInstant, but scriptedMinutes to hoursSeconds, and it actually converses
ScaleHighLinear with headcountThousands a day, same quality
Feels humanRarelyYes97 out of 100 times

Humans on the loop, not stuck in it

Passing for human doesn’t mean replacing your team.

A dispute, a hardship case, a legal threshold reached: one click and a collector from your team is in, with the whole history and the promise-to-pay record, and the agent picks the case back up when they hand it over.

Every intervention feeds analytics, so you know which flows need tuning and why.

We never hide that we’re AI. We just make sure it never feels like it.

Pick an hour. We’ll bring your numbers to it.

A first session is 30 minutes with the people who would build it — not a sales call. Bring one flow you’d like to fix and we’ll tell you, in the session, whether it’s worth doing.

What happens in the session

  • 30 minutesWith the people who would build it. Not a sales call.
  • One flowBring the flow you would like to fix first, and the numbers behind it.
  • A straight answerWe tell you, in the session, whether it is worth doing.
  • The inviteYou get the confirmation and the video link by email.