Payments is changing faster this year than at any point in a decade. Visa and Mastercard are running live transactions placed by AI agents. A $38 billion interchange settlement is close to final and will give merchants new reasons to renegotiate and new ways to surcharge. Chargebacks keep climbing, and Visa has tightened the dispute ratio that gets a merchant flagged. For ISOs, agents and independent processors, the core business has not changed: replace the merchants who leave and keep the rest. What has changed is how fast merchants expect answers, and how much AI is now on the other side of every transaction.
Where payments stands in October 2026
Agentic commerce has moved past announcements. Mastercard said in September 2025 that all U.S. Mastercard cardholders would be enabled for its Agent Pay program by the 2025 holiday season, and in 2026 it introduced Verifiable Intent, an open standard developed with Google that proves a shopper actually authorized what an agent buys. In April 2026 Visa launched Intelligent Commerce Connect, a single integration that lets businesses accept agent-initiated payments with tokenization, spend controls and authentication built in, now piloting with partners including AWS and Payabli. Visa has also announced a partnership with OpenAI on AI commerce. McKinsey projects agentic commerce could reach $900 billion to $1 trillion in U.S. retail revenue by 2030. For processors, that brings new questions about tokenized credentials, agent authentication, and who is liable when an agent buys the wrong thing.
The interchange settlement will reset conversations. In June 2026 a federal judge gave preliminary approval to a revised Visa and Mastercard settlement with merchants, valued at about $38 billion in savings. Under the reported terms, the average effective credit interchange rate drops 10 basis points for five years, standard consumer credit rates are capped at 125 basis points, merchants can choose which categories of credit cards to accept, and surcharging rules loosen. Large merchant groups have objected and final approval is pending. Either way, every merchant in a portfolio is about to hear about it, and many will call their processor to ask what it means for them.
Disputes keep growing. Mastercard's research with Datos Insights projects global chargebacks will grow 37% from 2025 to 2029, to 359 million a year and $46.1 billion in value. Merchants told Mastercard and Javelin that each chargeback costs them about $128 all in, $46 in fees and $82 in internal handling, against an average U.S. dispute of $110. Mastercard estimates that fraud, both third-party and first-party "friendly" fraud, accounts for about 45% of chargeback volume. The networks are tightening, too. Since April 1, 2026, Visa's Acquirer Monitoring Program treats a U.S. merchant as "excessive" at a combined fraud and dispute ratio of 1.5%, down from 2.2%, and flags acquirers at 0.5% and 0.7%. Every ISO now has to manage dispute ratios across the whole portfolio, not one merchant at a time.
Attrition never stopped. The Strawhecker Group has estimated small and mid-sized merchant attrition between about 21% and 24% a year. Pricing is the most-cited reason, and merchants shop when interchange resets in spring and fall. An ISO has to replace a fifth of its book every year just to stand still.
A merchant who gets voicemail during an interchange reset is a merchant who takes the competitor's rate review.
Where AI actually helps an ISO or processor
Answering merchants fast. Statement questions, deposit timing, equipment problems, rate questions and now settlement questions all arrive by phone and email. An agent that answers from the merchant's own records, around the clock, and escalates anything unusual to a person is the cheapest retention tool available.
Onboarding follow-up. Applications stall on missing documents and unanswered questions. Agents that chase the paperwork, confirm activation and check in after the first statements shorten the time to first transaction and catch problems before the merchant leaves.
Rate reviews before the competitor calls. Outreach to existing merchants timed to interchange resets and the settlement changes, with a rate review in the merchant's own numbers, turns the riskiest months of the year into retention months.
Dispute evidence packs. Chargeback responses are deadline-driven and evidence-heavy. Agents that assemble evidence from the transaction record by reason code and track every deadline raise win rates and keep dispute ratios under program thresholds.
Finding new merchants. Verified merchant lists by vertical, first touches drafted from your playbook and approved by your agents, and every call and email logged, so new-merchant acquisition keeps pace with attrition.
Where AI creates new risk
PCI DSS 4.0 treats AI as a system component. Every one of PCI DSS 4.0's future-dated requirements has been mandatory since March 31, 2025, including multi-factor authentication for all access to the cardholder data environment, targeted risk analyses, automated log review and payment-page script controls. An AI tool that reads merchant records or transaction data is in scope. A vendor that cannot answer the PCI questions of where it runs, who can reach it, what it logs and how access is authenticated becomes a problem at your next assessment.
Merchant data is sensitive data. Statements, bank details, owner Social Security numbers and transaction histories all pass through an ISO's systems. Pasting any of it into a consumer chatbot that trains on what it is given by default is a data-handling failure, whether or not anyone ever finds out. Financial-sector breaches averaged $6.29 million in IBM's 2026 report.
A practical AI checklist for ISOs and processors
1. Put AI tools into your PCI scope review. Know which tools touch merchant or cardholder data, where they run and how access is controlled and logged.
2. Answer merchants from their own records. Grounded answers about their account, never guesses about fees or funding.
3. Plan outreach around resets and the settlement. Contact merchants before they start shopping.
4. Treat disputes as a portfolio metric. Track deadlines and ratios across every merchant, not one case at a time.
5. Keep humans on rates and contracts. AI drafts the proposal and the follow-up. A person approves pricing and signs.
How Acclivity Labs helps
Acclivity Labs builds and runs AI agents for ISOs, merchant services providers and other small and mid-sized businesses. It is one private platform where the agents share one CRM, one approval model and one audit trail:
Merchant acquisition. Verified merchant lists by vertical, first touches drafted from your playbook and approved by your team, calls placed and logged, and a pipeline in one CRM with residual tracking.
Onboarding and retention. Application follow-up, activation confirmation, rate-review outreach timed to interchange resets, and statement and equipment questions answered from your records by a receptionist agent that hands anything unusual to your staff.
Disputes and risk. Chargeback evidence packs assembled from the transaction record, deadlines tracked, and portfolio risk flags before a merchant crosses a program threshold.
Proposals, agreements and finance. Rate proposals in the merchant's own numbers, merchant agreements signed and audited through eSign, residual reconciliation, agent payouts and monthly reporting.
Marketing and web. Acclivity Digital Marketing, our affiliated agency, handles websites, SEO and AI-answer visibility, so merchants searching for a processor in your verticals, and the AI assistants they ask, find you. Content is drafted for your approval before anything posts.
Security and compliance. Continuous PCI DSS 4.0 evidence, including MFA, log review, targeted risk analyses, payment-page controls and penetration testing, with managed IT from our affiliated partner.
Built for your assessor. Our agents run on our own Nvidia DGX hardware using open-weight models. Merchant data stays in your own tenant under your retention policy and is never used to train a model. A hosted frontier model is used only where a task needs it, and only under commercial no-training terms. Every agent action is logged and access is authenticated, so you can answer your assessor's questions about it.
The processors that grow through the next few years will be the ones that answer every merchant before the competitor does, and can show their assessor exactly what their AI touches.