Agentic AI for Chargebacks: Findustry AI’s Plan to Automate Dispute Management

Findustry AI founder and CEO Jonathan Razi is pitching a way for merchants to offload the grind of card dispute handling through automation.

Refund reversals have surged, with shoppers more often contesting transactions — even “friendly fraud” cases where buyers keep the product or service and still challenge the charge — creating a costly headache.

A veteran of the payments ecosystem, Razi conceived in 2024 a tool to tame chargeback volume and assembled a compact Chicago-based team, leveraging his CardX network, to build it with AI.

CardX, which enabled surcharge implementation for merchants, changed hands in 2021 when processor Stax Payments acquired the company for an undisclosed sum.

Findustry AI debuted last year and, after Razi raised $2.25 million, introduced a public brand for agentic dispute services in June; he says the startup isn’t seeking additional capital right now.

To date he will publicly cite two clients: Cleveland’s Electronic Merchant Systems, operating as Kurv, and Indianapolis-based facilitator AllPaid. Kurv deploys the chargeback agent internally and resells it to merchants of all sizes, while AllPaid embeds the tool in customer portals to support municipal and county payments.

In an extended interview this month, Razi — also a trained attorney — discussed the new venture alongside broader topics such as agentic commerce, Visa’s latest fraud efforts, and stablecoins.

Editor’s note: This conversation was condensed and clarified.

CardX Sale to Stax: What Happened and What Came Next?

I completed the CardX sale on November 5, 2021. People say you can’t time markets, but that date, looking back, aligned with peak valuations for many growth names, so the timing proved favorable.

Findustry AI Initiative: Building AI Agents to Automate Payments Workflows

At a high level, we create autonomous agents for payments operations, using generative models to execute workflows previously handled by staff.

Traditional dispute management is typically a mix of inbox triage, copy-pasted templates, evidence hunts across multiple systems, and deadline management — whether done in-house, via a BPO, or with a rules-based tool. An agentic approach is different because the software isn’t just generating text; it’s taking goal-driven steps end to end: pulling transaction and order data, requesting and assembling supporting documentation, mapping the case to the right response strategy, drafting a packet, submitting it through the appropriate channel, and then tracking outcomes and follow-ups.

When dispute work becomes an automated workflow rather than a manual scramble, the biggest win is consistency: the same evidence, the same formatting discipline, and the same on-time execution for every case.

Agentic Commerce: How Early Is It and What Will Stick?

Agent-led buying is still nascent. In my view, there are more builders than active users today, and some visions feel overhyped; the mundane, useful pieces will land first. For example, a fully automated “plan my Italy trip and book everything” experience seems distant — many of us still prefer to choose those details ourselves.

When I say “agentic commerce,” I mean software agents that can act on a customer’s behalf to discover products, make decisions within guardrails the customer set, and then execute a purchase with minimal extra input. That shifts the chargeback conversation because the customer experience becomes “I didn’t do that” or “I didn’t mean that” even when the customer did authorize an agent to act, and merchants still have to prove what happened in a way the networks recognize.

In practice, an agent purchase usually has a few steps: the customer grants permission and sets constraints (budget, brands, return preferences, delivery address), the agent authenticates to a wallet or payment credential, it selects a merchant offer, and then it completes checkout by passing billing/shipping details and confirming the order. If the merchant and PSP can capture a clean “consent trail” — what the customer delegated, what the agent chose, and what the customer saw at confirmation — it gets much easier to resolve downstream disputes.

It also creates new dispute scenarios. An agent can buy the “right” thing but the customer doesn’t like the outcome; it can select a similar item with a different SKU; it can trigger edge cases in subscriptions and renewals; it can route to a marketplace seller the customer didn’t notice; or it can create confusion around cancellations, partial shipments, and delivery instructions. Those are all fertile ground for first-party disputes when expectations and disclosures aren’t crisp.

Liability is the tricky part. The cardholder is still the party tied to the account, and the merchant is still the party that must answer the dispute within the card-network framework. The agent operator (whether that’s a wallet provider, an app, or a platform) may have its own terms with the user, but that doesn’t automatically move representment responsibility off the merchant. What changes is that merchants, PSPs, and agent operators can cooperate on better proof of authorization and clearer customer-facing controls so “I never approved this” is less ambiguous.

Merchants can mitigate agent-driven chargebacks by designing checkout and post-purchase flows to make delegation obvious and recoverable: step-up confirmation for higher-risk orders, consistent and descriptive billing descriptors, fast and self-serve refunds, clear cancellation paths, delivery and fulfillment proof, and customer support that can resolve dissatisfaction before it becomes a dispute. On the backend, the strongest posture is disciplined recordkeeping that ties together order details, delivery, communications, and any agent-related authorization signals provided at checkout.

On the standards side, I expect the industry to push toward more structured transaction metadata that travels with the payment — not just a total and a timestamp, but “digital receipt” elements that capture what was purchased, what policies were presented, and what confirmation event occurred. PSPs sit in the middle of that: they can normalize merchant data, pass enhanced fields through their integrations, expose dispute and evidence APIs, and help merchants implement controls (tokenization, authentication options, and logging) that make agent-driven transactions more defensible.

To win disputes in an agent-driven world, the evidence story has to be specific. Merchants will still rely on the fundamentals — product and price details, policy acceptance, proof of fulfillment, and customer communications — but the strongest packets will also include clean traces of authorization and intent, such as authentication results, the confirmation page or receipt presented at checkout, and any available agent-consent or “delegated action” record that shows the purchase was within the customer’s configured guardrails.

Longer term, I think AI will show up on both sides of the equation: more automated purchasing up front and more automated dispute handling behind it. The merchants who do well will be the ones who treat disputes as an operational system — prevention, logging, customer experience, and representment — because the volume and complexity won’t trend down on their own.

Merchant Benefits: Why Use a Chargeback Agent?

Merchants typically see a handful of practical advantages from using this kind of automation:

• Higher dispute win rates.
• Reduced labor costs by shifting repetitive casework away from staff and freeing teams to focus on other priorities.
• Material savings for large enterprises.
• Ability for small merchants to contest disputes they would otherwise write off.
• Automated expertise in reason codes, evidence standards, and defense frameworks.

Chargeback Trends: What’s Happening Now?

Disputes are rising across channels, especially in card-not-present transactions. First-party or “friendly” fraud is common — customers claim nonrecognition when they’re actually dissatisfied — forcing merchants to spend time on responses. Recurring membership billing is a strong fit for this kind of solution as card brands roll out new tools and mandates that add complexity to the lifecycle.

As more purchasing becomes mediated by software, I expect the pressure to increase on merchants to respond with better transaction detail, faster customer resolution loops, and tighter alignment between what the customer intended and what the merchant can prove. The operational bar rises even if the underlying product doesn’t change.

Visa’s VAMP: What Does the New Acquirer Monitoring Mean?

Visa has tightened thresholds, pushing more merchants into higher-risk tiers at lower chargeback ratios than before. The message is clear: let the problem run, and you could face extreme outcomes, including loss of your merchant account.

Regulation Outlook: Federal Versus State Paths

If you had to bet, expect increasing divergence among states, with more jurisdictions crafting their own consumer-protection rules while others take different approaches — a byproduct of how hard national consensus can be.

For nationwide providers, that means products must keep pace with a patchwork of state regimes wherever customers transact, raising compliance costs for virtually every processor and platform.