Chargeback management software that fights the disputes worth fighting, with every exhibit cited
More chargebacks won, weak ones accepted fast, and no dispute left to expire.
See one case, screen by screen ↓Software that takes a card dispute from the processor's notice to a decision: fight it with a representment response, accept it, or prevent the next one. Chargeback Management adds seven agents that bring every dispute in with its true deadline, price fight or accept, fill the evidence the reason code requires, draft the response and check it as the issuer would — while the chargeback analyst decides and signs.
How to choose AI solutions for back-office and enterprise work →
Updated
A merchant that takes cards gets disputes every day from several processors, each under one of dozens of network reason codes with its own evidence rules and an earlier processor deadline. The analyst opens each one in a processor portal, decides by feel whether it is worth fighting, then hunts for the order, the delivery scan, the sign-in and the service notes across five systems, pastes them into a template letter and uploads it — often in the last two days of the window.
So weak cases are fought and lost, strong ones go out without the one exhibit the reason code needs, and some expire unanswered. A Visa 10.4 fraud claim that would qualify for Compelling Evidence 3.0 is answered as a plain fraud claim, because finding two earlier undisputed orders on the same card, from the same device or IP address, takes longer than anyone has. Outcomes come back weeks later in a processor report, and nobody ties a loss to the missing exhibit or a repeat disputer to the checkout rule that let them through.
Estimated split of the roughly 17 days a response took by hand at the demo company, against the one-day target with the agents.
Seven agents price every dispute, fill the evidence the reason code requires and check each other's work; the analyst decides and signs.
Kestrel & Vane Group sells through four card-accepting businesses and three processors. On Wednesday morning a $612.40 Visa 10.4 fraud claim on a cordless tool kit is waiting for Daniel Okafor. Here is what happens next, screen by screen, in the working solution.
The agents finished the overnight run at 06:40: 61 disputes in from three processors, 44 packs drafted and checked, 12 proposed to accept. Today shows 412 open disputes with $171,860 at stake, 23 due in the next 3 days with 21 packs ready, and a Q3 win rate of 62%, up from 41%. At the top of his list: CB-88213 qualifies for Compelling Evidence 3.0.
Intake agent: “Brought in 61 disputes: Northbeam 44, Halberd 11, Ridgeway 6. Every deadline set to the processor's date.”
CB-88213 is a Visa 10.4 Other Fraud — Card-Absent Environment claim from J. Pell, card ending 4417, on order KVH-4471902. Northbeam wants the response by Sat 17 Oct; the network date is Thu 22 Oct. Priced as a plain fraud claim, the win odds are 31%: 31% × $612.40 − $22 to fight = $167.84. The left side lists a socket for every piece of evidence the rule requires, and the agent fills them as he watches.
“Evidence agent · step 5 of 9 · Matched the device on all three.”
The agent pulls the disputed order, the Corvid Parcel delivery with a photo at the front door on Wed 5 Aug, and a verified sign-in with a code to the account holder's phone six minutes before the order. Then it searches 12 months of the card and finds KVH-4012277, 140 days earlier, and KVH-3768410, 301 days earlier — neither disputed, both from device dv-7f3a-91c2 and IP address 198.51.100.24. Every exhibit opens its source record.
“Highlighted fields are the ones the agents relied on. Read-only copy of the record, taken when the agent read it.”
Nine of nine sockets are filled with eight exhibits, and the Compelling Evidence 3.0 test passes: two earlier orders 120–365 days old, neither disputed nor reported as fraud, the device, IP address and shipping address matching on all three (the rule needs two of four, one of them the device or IP address), same merchant and descriptor. The triage agent re-prices the dispute at 78%: $455.67 expected if he fights, against a $612.40 refund and a $2.00 fee if he accepts.
The evidence agent never invents an exhibit — an empty socket stays empty.
The drafting agent wrote the response with every sentence pointing at an exhibit. The critic scored it 86 against a pass mark of 75, struck a sentence about the three-year warranty and moved the match table to page 1. Daniel asks why the dispute qualifies; the evidence agent answers with the two orders, the matching device and IP address and the rule, each cited.
Critic: “irrelevant to a fraud claim, and it invites a quality argument.”
Daniel picks reason code S1 · Evidence complete — Compelling Evidence 3.0, adds a note if he wants one, and approves with his e-signature. The response goes to Northbeam three days before its deadline. Had the critic scored it below 75, he could only have submitted it with reason S3 and a reason in his own words.
“I approve this as Daniel Okafor, Chargeback Analyst — e-signature.”
The triage agent proposes accepting 12 disputes worth $1,485.35: four are Kestrel & Vane's own mistakes — a duplicate capture, a missed cancellation, a forced authorisation and a return never refunded — two look like true fraud, and six cost more to fight than they could recover. Daniel can switch any row to Fight instead, then approves the rest with one reason code inside the free window.
“8% × $459.05 − $22 = $14.72 — below your $15 fight floor.”
On CB-87455 the issuer comes back with photos of a dented door and a $640 repair quote. The triage agent prices every option: escalating to arbitration is worth about $36 after the filing fee, accepting alone costs $1,960.00, and accepting while filing a concealed-damage claim with Trident Freight recovers about $1,568. Daniel chooses; the evidence agent has the claim drafted.
The pattern agent finds 14 repeat disputers this year, $9,820.00 across 33 disputes, and proposes a change for each of the 6 new ones — mostly step-up verification at checkout. It also sees Visa 10.4 up 18% at KV Home Store since the garden promotion, and a descriptor customers do not recognise. Rafael approves or declines each proposal with his signature.
“Nobody is blocked: step-up asks for a code at checkout.”
The outcomes agent matches every recovery to a processor settlement line: $679,400 won in Q3, $671,950 matched, and $7,450 across three items not yet matched, each one explained. It proposes a reserve for open disputes of $74,438.16 — open amount × (1 − weighted win odds) — and journal JV-CB-1014. Helen approves the journal; the agent never posts it.
The same disputes, seen by the five people who carry them — what their week looked like, and what it looks like now, in their own words.
Seven agents price every dispute, fill the evidence the reason code requires and check each other's work; the analyst decides and signs.
Brings every dispute, pre-dispute alert and pre-arbitration in from each processor every 15 minutes, maps the reason code to its playbook, and sets the due date to the processor's date with the network's date beside it.
Prices every dispute from 12 months of your own outcomes for the same reason code and business — win odds × amount − cost to fight — and proposes fight, accept, refund through an alert, or recover elsewhere. It re-prices whenever the evidence changes.
Plans which systems answer each socket, fills the required ones first with cited exhibits, and for every Visa 10.4 searches 12 months of the card for two undisputed orders 120–365 days old and tests the matching elements for Compelling Evidence 3.0.
Writes the response for the reason code, leading with the strongest path and pointing every sentence at an exhibit — one page plus exhibits.
Reads every pack as the issuer's reviewer would, scores it against the reason code, strikes what does not help, and sends weak packs back to the evidence agent with the missing socket named.
Finds repeat disputers, reason codes on the rise and descriptors customers do not recognise, and proposes prevention to the fraud analyst.
Records wins and losses, matches every recovery to a settlement line and the ledger, explains every unmatched item, computes the reserve and win rates, and watches the dispute ratio per account.
Signs every submission and acceptance; the operations manager co-signs above $2,000. The agents propose; a named person decides.
Every screen has the agents in reach. Ask why, what if, or what to do next — they answer with the records behind each claim, and nothing is submitted, accepted or refunded until a person approves it.
Why does it qualify for Compelling Evidence 3.0?
Two earlier orders on the same card fall inside the 120–365-day window: KVH-4012277, 140 days before, and KVH-3768410, 301 days before. Neither was disputed or reported as fraud. All three share device ID dv-7f3a-91c2 and IP address 198.51.100.24 — two matching elements, one of them the device. Same merchant and descriptor. That meets the rule, which shifts liability back to the issuer.
What if we accept instead?
Accepting returns nothing and costs a $2.00 acceptance fee — we are on day 22 of 30. Fighting: 78% × $612.40 − $22 to fight = $455.67 expected. Before the Compelling Evidence 3.0 search it was 31% and $167.84. Fight.
Why accept CB-88299? It is $459.05
Checkout risk scored the order 91 / 100 and it shipped on a manual override to a different state from the billing address. Two of our last three disputes like this were lost; win odds 8%. 8% × $459.05 − $22 = $14.72 — below your $15 fight floor. I passed the override to Rafael Montes.
Did any agent act without a person?
Only under the alert refund rule Ingrid Solberg set: 31 refunds this week, each under $25 on an unshipped order, each logged with the rule version. Every submission, acceptance, pre-arbitration answer and journal carries a person's reason code and e-signature.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds the agents work within are settings. Every change is versioned and on the audit trail — like Ingrid Solberg raising the critic pass mark from 70 to 75.
| Setting | Default | Choose from |
|---|---|---|
| Fight floor | $15 | Propose to fight only when win odds × amount − cost to fight is above it |
| Cost to fight one dispute | $22 | Analyst minutes and processor fee, used in every expected value |
| Critic pass mark | 75 / 100 | A pack below it goes only with an override and reason S3 |
| Co-sign above | $2,000 | Submissions above it need the payments operations manager |
| Submit lead time | 2 days | Packs are ready this many days before the processor's due date |
| Alert refund rule | $25 | Pre-dispute alerts refunded automatically up to it when the order has not shipped |
| Escalate pre-arbitration only when | $100 | Expected value after the filing fee is above it |
| Keep device and IP data for disputes | 18 months | Then deleted; used for nothing else |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate ($1.55M = $5.92M disputed a year at the fictional group, fought share 48% → 74% and win rate 41% → 62%; the split of days is estimated too) · “typical” = published figures (Mastercard 2025 global chargebacks outlook — 324 million chargebacks a year by 2028, 45% reported by merchants as fraud; LexisNexis Risk Solutions True Cost of Fraud 2025 — $4.61 per $1 of fraud, US). Visa Compelling Evidence 3.0 and the network dispute-ratio threshold are paraphrased. People, companies, processors, carriers and cardholders named on this page are fictional — characters and sample data in the working solution.
Software that takes a card dispute from the processor's notice to a decision: fight it with a representment response, accept it, or prevent the next one. Chargeback Management adds seven agents that bring every dispute in with its true deadline, price fight or accept, fill the evidence the reason code requires, draft the response and check it as the issuer would — while the chargeback analyst decides and signs.
The triage agent prices every dispute from 12 months of your own outcomes for the same reason code and business: win odds × amount − the cost to fight. When that is below your fight floor it proposes accepting, with the reason and the arithmetic on the row. It only proposes; the analyst approves, or switches any row to fight instead.
On every Visa 10.4 fraud claim the evidence agent searches 12 months of the card for two earlier undisputed orders 120–365 days old and tests whether at least two of four elements match, one of them the IP address or device, with the same merchant and descriptor. When the test passes, the response is built on it — in the demo, win odds rose from 31% to 78%.
No. Agents never submit, accept, refund, contact a cardholder, change a deadline, post a journal or block a customer. Every submission and acceptance carries the analyst's reason code and e-signature, and the payments operations manager co-signs above $2,000. The one exception is a pre-dispute alert refund under a rule the manager sets, logged each time.
Yes. The intake agent reads disputes, alerts and pre-arbitrations from each processor every 15 minutes and sets each due date to the processor's own date, with the network's date beside it. Reason codes from Visa, Mastercard, American Express and Discover each have a playbook of the evidence they require, which your team can edit.
The outcomes agent matches every win to a processor settlement line and the ledger and explains anything unmatched. It proposes the month-end reserve for open disputes — open amount × (1 − weighted win odds), by business — and the journal, which the finance controller approves; the agent never posts it.
Yes. The fight floor, the cost to fight, the critic pass mark, the co-sign limit, the submit lead time, the alert refund rule, the pre-arbitration threshold and how long device and IP data are kept are all settings, versioned and on the audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your dispute policy, processor connections and 12 months of dispute history for the baseline — and it goes live once every quality gate has passed. We will walk you through it on your own past disputes first.
Policy changes reach every article in a day, and the bot never answers from a stale page.
1-min demoWarranty Claims ManagementInvalid claims stopped before payment, and supplier defects charged back with proof.
1-min demoReturns ManagementFraud held before the refund; every item sent to the lane that recovers the most.
1-min demoWe'll run Chargeback Management on a sample of your own past disputes and the systems their evidence lives in.