By Sagar Shankaran, Founder of CallSphere
The past-due report is the gym process to baseline before buying AI: five numbers to capture, a worked example on a 1,400-member club, and the honest limits.
Key takeaways
Take a 1,400-member club at $52 a month in dues. The draft file runs on the 1st. On the morning of the 3rd the returns come back from the billing processor, and somebody at the front desk — usually the member services associate you hired to give tours and sell memberships — prints the past-due report and starts dialing. Ninety-eight names. Card expired. Account closed. Insufficient funds. "Do not honor." She gets through maybe thirty of them, because the only quiet hours at the desk are between the 5:30 a.m. rush and the 4:30 p.m. rush, and the members she actually reaches at 11 a.m. on a Tuesday are the ones least likely to be a problem in the first place.
The ones she does not reach keep training. Your collections policy says the fob stays live for thirty days, so they scan in Monday, Wednesday and Friday while owing two months of dues. Then they cancel, or they dispute the last good draft, or the account ages past forty-five days and goes to the recovery vendor. Nobody in the club can tell you what share of those dollars came back within thirty days, because nobody has ever written it down.
Declined-draft recovery is the sequence a health club follows after a monthly electronic funds transfer fails — the member is flagged past due, contacted, given a way to fix the card on file, and either recovered or written off — and it is the cleanest place in a gym to prove whether AI paid for itself, because every step already carries a date and a dollar amount.
Every club has the same three-layer workaround. Layer one is the automated dunning email your billing processor sends, which lands in promotions and gets opened by roughly nobody. Layer two is the text blast on the 5th from Mindbody or ABC Ignite or Club Automation, the same message to all ninety-eight people, including the twelve whose cards did not decline at all but whose drafts failed for a bank routing change. Layer three is the alert that pops on the check-in scanner when a past-due member badges in — which the desk clears without reading, because nine people are in line for the 5:45 class and nobody wants the confrontation.
The seasonal pattern makes it worse in exactly the way you would expect. The January surge signs three hundred members in six weeks, a good share of them on cards that were already close to the limit, and those accounts start bouncing in March and April — the same weeks the desk is drowning in the first cancellation wave from the same cohort. So your worst return month and your worst retention month are the same month, and both land on one person.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
Deloitte's State of AI in the Enterprise 2026 found that 84% of organizations investing in AI report a positive return. The interesting part is not the percentage but what those organizations had in common: they took one messy, repetitive process, kept a human reviewing the output, and proved either time saved or errors reduced before they widened the scope to anything else. Nobody in that group started with a club-wide rollout.
For a gym, the declined-draft list is the process that fits that description best. It repeats every single month on a known date. It already produces a report. It is denominated in dollars, not in vague feelings about member experience. And it has a clean before-and-after: the recovery rate in the two cycles before you change anything, against the recovery rate in the three cycles after. If your general manager cannot say what the number was in March, you are not ready to say what it is in September.
flowchart TD
A["Draft file runs on the 1st"] --> B["Returns land on the 3rd"]
B --> C{"Reason code on the return?"}
C -->|Expired card| D["Text link to member's own portal to update"]
C -->|Insufficient funds| E["Retry scheduled to member's stated payday"]
C -->|Account closed| F["Flagged for a human call"]
D --> G["Associate reviews queue at 10am, approves or edits"]
E --> G
F --> G
G --> H["Dollars recovered and hours logged for the cycle"]
H --> B
Take two cycles, not one, and make one of them a shoulder month rather than the January-affected spike. Write the numbers on one sheet of paper and put it in the GM's binder. That sheet is the entire argument you will have in six months.
7:10 a.m. The return file lands. The assistant reads the past-due report out of your club management system and sorts it by reason code instead of alphabetically. Expired cards get a short text pointing to the member's own portal to re-enter the card — never a link it invented, never a card number typed into a chat window. Insufficient-funds accounts get a plain message asking which day works, and the retry is set to that date rather than blindly re-running on the 10th. Account-closed accounts go straight onto a call list for a person, because a closed bank account usually means a bank switch or a member who has already decided to leave, and both need a voice.
10:00 a.m. The associate opens the queue and reads it. She approves most of it, rewrites two messages, and kills the one going to the member whose husband died in April — which she knows and the software does not. Nothing sends unreviewed in the first two cycles. Every contact writes a note back into the member record, so when that member calls the desk on Saturday the next person can see the whole history instead of asking them to explain it again.
The assistant stops at a hard line: no fee waivers, no cancellations, no freezes, nothing said about contract terms. Those decisions belong to the GM.
Assumptions, all of them illustrative — put your own numbers in the same boxes: 1,400 members, average dues $52, so a monthly draft of $72,800. Suppose 7% of drafts return, which is 98 accounts and $5,096 at risk. Suppose your current thirty-day recovery is 55% of those dollars, and the associate spends 22 hours a month chasing them at $19 an hour fully loaded.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
| Line | Before | After (illustrative) |
|---|---|---|
| Dollars bounced on the 3rd | $5,096 | $5,096 |
| Recovered within 30 days | 55% = $2,803 | 72% = $3,669 |
| Members lost from the unrecovered group | 19 | 11 |
| Remaining agreement months on those members (avg 8) | $7,904 of future dues | $4,576 |
| Desk hours on the chase | 22 hrs = $418 | 7 hrs = $133 |
| Tool cost | $0 | $400 |
That is $866 more collected this month, $285 of labor back, and roughly $3,300 of future dues that did not walk out the door — against $400. The number that settles the argument in a room with your accountant is not "AI is working." It is dollars recovered within thirty days per 100 returned drafts, measured for two cycles before and three cycles after. One line. Either it moved or it did not.
A past-due balance is sometimes a payment problem and sometimes a life event, and the difference is invisible in a reason code. Deaths, deployments, a layoff at the plant, a member three weeks post-surgery who thought the medical freeze paperwork went through — every one of those needs a person who can hear the pause on the line. Automating a polite chase into that situation is how a club ends up in a complaint letter to the state attorney general's office, and health-club contracts are one of the few consumer categories most state AGs actually read the mail about.
Two other hard stops. The last contact before write-off should be a human call, because that is your last chance to save the member rather than the balance. And any account where the member says "I already cancelled" comes off the automated list immediately and goes to the GM with the paperwork pulled.
You are collecting your own debt, which is a different legal posture from a third-party collection agency, but the texting rules still apply: you need the member's consent for that number, every message honors STOP immediately, and you keep contact inside normal hours. Most club agreements already collect texting consent at join — check that yours does, and have your attorney read the message templates once before the first cycle.
Your processor retries files. It does not know that this member has scanned in eleven times this month, or that she has a training package with four sessions left, or that the last three declines were all the same expired card. The recovery message that works is the one that knows those things, and those live in your club software, not in the billing file.
Five cycles. Two before, three after. Anything shorter and you are reading noise, because one bad month for the local plant will move your return rate more than any tool will.
CallSphere builds AI voice and chat agents that answer club phone lines and web chat, book intro sessions and tours, and capture leads around the clock. If your baseline shows that the member services desk is losing hours to the phone during the same window it is supposed to be working the past-due report, that is the overlap worth looking at — the agent takes the ringing line so the person can do the work that needs judgment.
Written by
Sagar Shankaran· Founder, CallSphere
LinkedInSagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
Map one messy process and prove it: spray ticket records, the five baseline numbers to capture before you start, and the error rate that ends the debate.
A cloned voice can get a prepaid training package refunded and a 24-hour fob turned back on. The callback rule that closes both, and what one incident costs.
Deloitte found 84% of AI investors report positive returns. For P&C carriers the provable process is FNOL intake - here are the four numbers to baseline first.
How a security guard company proves AI paid for itself: map the open-shift callout, take a 30-day baseline, and track overtime as a share of billed hours.
Pull twelve months of factor deductions, convert to chargeback dollars per $100,000 shipped, and you have a baseline that settles the AI argument in 90 days.
How a benefits agency proves AI paid for itself: reconcile every carrier commission statement, capture five baseline numbers, and count recovered dollars.
© 2026 CallSphere Inc. All rights reserved.
Made within San Francisco
Watch how CallSphere handles real customer calls, schedules appointments, and processes payments — live.
Try Live DemoBook a DemoCalculate Your ROI