The Only Calls Left Are the Hard Ones

Pull up what your collectors actually worked last week and sort it by difficulty.

Hardship disclosures. Bankruptcy notices. Attorney letters. Deceased account holders. Disputes. Consumers already angry at three prior attempts from two prior agencies. Accounts where the right answer depends on the consumer's situation, the client's authority limits, and a state rule, all resolved live, in about eight minutes.

Then look at what a straightforward account looks like in that same week. There are fewer of them than there were two years ago, and there will be fewer again next year.

Three separate forces are reshaping the ARM queue right now. They have nothing to do with each other, and they all push the same direction.

Force one: the easy contacts are leaving the human queue

Routine contacts are being automated across the industry. Payment reminders. Balance inquiries. The consumer who was always going to pay and needed a path to do it. That's sensible, the cost savings are real, and agencies that hold out will end up competing on price against agencies that didn't.

The part that gets left out of the business case is that those calls were doing a second job. They were where a new collector learned to talk to someone who owes money, at low stakes, before meeting anything genuinely hard.

Remove them and the skill floor for every remaining conversation goes up while the supply of low-stakes practice goes down.

This observation is not ours and it is not new. Simulation vendors have been making it since at least 2023, and the trade press has covered it steadily. What has changed is that it stopped being a forecast. The queue described above is what a shift actually looks like now, and the other two forces below arrived on top of it.

Force two: the paper coming in is older

The Federal Reserve Bank of New York's Q2 2026 household debt report puts total US household debt at $18.8 trillion, with 4.7% of balances in some stage of delinquency. Card balances 90 or more days late reached 12.8%, up from 7.6% in 2022.

What the New York Fed attributes that climb to changes the meaning of it. The flow of new delinquencies has held roughly steady for two years. The stock rose largely because charged-off debts are now reported for longer. The growth in the pool is aged paper.

You already know what aged paper does to a shift. Contact data that failed for somebody else first. Consumers who have had this conversation twice. Thinner economics per account, which means less room to absorb a call that goes badly. More situations where the answer is a hardship program or nothing.

Force three: the people who are good at hard calls are the ones you have the fewest of

It takes a long time to build a collector who is reliably good at empathy, compliance, objection handling, and reading ability to pay, all of it live and in the same conversation. Operators commonly describe it in years. At one national agency working bank and card-issuer paper, roughly one in six new hires is still on the floor at twelve months.

So the population capable of handling the queue described above is your most tenured people, and your most tenured people are your scarcest resource and the one you're least able to expand by hiring.

Why hard calls punish you harder than they used to

On a difficult call the odds of resolution drop and the risk profile changes with them.

Right party contact rates on outbound attempts sit in the single digits, and consumers who reach RPC typically engage once. There's no second attempt at the hardship conversation. If the collector pushes for payment before acknowledging what the consumer just disclosed, or lands the mini-Miranda out of the order that client requires, or offers terms the client never authorized, the account doesn't get resolved next week instead. It goes back in the queue and eventually to the next agency in the chain.

And unlike a routine call that simply fails, a mishandled hard call generates the things collections leaders actually lose sleep over. Complaints. FDCPA exposure, which consumers can pursue directly with no regulator involved. A finding in a client's vendor oversight review.

The upside shrank and the downside grew at the same time.

How do you train collectors for hardship and bankruptcy calls?

If the average call is harder and the average collector is newer, there are only three places to intervene, and they map to three moments.

Before the call. A collector should meet the interrupter, the hardship disclosure, and the third-party answer somewhere other than a live account. That used to happen naturally, through volume, on calls that no longer exist. It now has to be deliberate. Practice against your scripts, your client's authority levels and your scorecard, not a generic library, which is why most agencies who tried simulation once didn't try it twice.

During the call. Collectors touching several placements in a shift are holding the same required disclosures in different orders with different settlement authorities. When a certified collector still doesn't know whether this client allows a twelve-month plan, the options are hold, guess, or apply the last placement's rule. All three cost you the call, and only one of them is visible to you afterward.

After the call. A common QA standard is around eight calls per collector per month. Against ten to twenty live conversations a day over roughly twenty working days, that's fewer than one in twenty. The hard calls are mostly in the nineteen nobody reviewed, which means the conversations that carry the most risk and the most recovery are the ones you have the least information about.

The question worth asking any vendor

When a QA finding says a collector missed something, what happens next?

Most stacks stop at the finding. Analytics surfaces it. Guidance tools try to prevent a repeat using a prompt somebody wrote months ago. Simulation tools drill scenarios with no connection to what your QA team actually scored. The finding, the answer the collector should have given, and the practice that fixes it live in three systems, maintained by three teams, against three copies of a policy that drift the moment a client revises a script.

If a change to a client's settlement authority doesn't update what your collectors can look up, what they practice against, and what they're scored on, you don't have a loop. You have three tools and a maintenance burden.

Where Posh fits

Posh was built for the three moments above, and is the only platform that runs all three on one knowledge base.

Posh Simulator runs the certification step. Collectors clear the interrupter, the hardship disclosure and the third-party answer before they take a live account, scored against your scorecard and your client's authority levels. The scenarios generate from your own scripts, so a placement changing its settlement terms changes the drill rather than creating an authoring project for your team.

Posh Knowledge Assistant answers the collector mid-call, from that client's manual, with the document and section cited and the lookup logged. The hold doesn't happen. Bankruptcy notices, attorney letters, cease-and-desist requests and state-specific requirements come back the same way, which matters more as those calls become a larger share of the queue.

Posh CoachQA scores every conversation against the criteria your QA team already uses, including disclosure sequence rather than only whether a required phrase appeared. Interpreted calls get scored like any other, so bilingual volume stops being a hole in your coverage. Then the miss becomes the scenario that collector runs in Simulator before the next shift.

That's the loop. One knowledge base underneath it, so a client's script revision moves the answer, the drill and the scoring criteria together instead of three teams chasing the same change.

Does AI in collections create FDCPA risk?

Agencies frequently stall on AI in collections, and the caution is well founded given direct consumer litigation risk.

The distinction that matters is whether a given tool speaks to the consumer. Automated outreach does, and it deserves the scrutiny it gets. The three capabilities above don't. Practice talks to your collector. Retrieval answers your collector. Scoring reads a transcript of a conversation a human already had.

Evaluate any tool on which side of that line it sits, rather than on whether the word AI appears in the deck. It's usually the difference between a six-month compliance review and a decision reached in the first meeting.

Household debt and delinquency figures come from the Federal Reserve Bank of New York, Quarterly Report on Household Debt and Credit, Q2 2026. The change from 7.6% to 12.8% spans Q3 2022 to Q1 2026. Contact rate, ramp and QA coverage figures reflect the operating experience of a national third-party collection agency collecting for bank and card-issuer clients, shared with permission, and are illustrative rather than industry benchmarks.

Start with one hard call. Send us one client script and one scenario your collectors struggle with. We'll show you how that requirement becomes a practice drill, a source-backed live answer, and a documented QA criterion. 45 minutes, nothing to prepare beyond the script.