The Phocuswright Conference 2025 marked the moment agentic commerce moved from theoretical to operational for the travel payments ecosystem. Two major protocols — Stripe and OpenAI's Agent Commerce Protocol (ACP) and Visa's Trusted Agent Protocol (TAP) — were publicly detailed by their creators on the main stage, with both already live in limited production. The payments architecture question is no longer "if" AI agents will transact autonomously on behalf of consumers; it is "how" the rails, fraud controls, and merchant-of-record structures must be rebuilt to support that reality. For payments leaders, the window to shape these standards before they harden is narrow.
Financial institutions are simultaneously emerging as the most credible new class of travel seller. Chase Travel is growing at 20% CAGR targeting $12 billion in gross bookings, American Express's travel platform (powered by iSeatz since 2012) participates in a US financial institution travel market estimated at $45–50 billion annually, and Hopper Technology Solutions has made fintech ancillary products — installment plans, price freezes, cancel-for-any-reason coverage — the primary revenue engine of one of travel's most-watched tech companies. Banks and fintechs are not supporting travel distribution; they are becoming its next dominant layer.
Fraud and authentication risk has been thrust to the center of the agentic commerce debate. The core engineering challenge is not just tokenization — it is distinguishing a legitimate AI agent acting on a consumer's behalf from a fraud bot, and ensuring that a consumer's expressed intent at the time of authorization actually matches what an agent ultimately purchases on their behalf. Neither ACP nor TAP fully resolves this; both are designed as open standards that will require iterative refinement as real-world agentic transaction volumes scale. The travel industry, with its high-value, complex, multi-leg transaction structures, is uniquely exposed to the failure modes these protocols are designed to prevent.
On the broader commercial horizon, stablecoins are being positioned by both Stripe and Visa as the settlement layer that makes agentic commerce economically viable at global scale — particularly for multi-currency cross-border payouts in B2B contexts. The convergence of agentic protocols, tokenized credentials, stablecoin settlement, and digital identity (with 500 million standards-based government digital IDs projected by end of 2026) represents a structural shift in travel commerce infrastructure that CFOs and treasury leaders must evaluate now, not in the next planning cycle.
Key Findings
Finding 1: Two Competing Agentic Commerce Standards Are Already Live — and Both Target Travel
In the session “AI Agents, Real Transactions — The New Travel Economy,” Clara Liang of Stripe and Gloria Colgan of Visa detailed their respective protocols in the same room, representing the first public side-by-side comparison of the two dominant frameworks.
ACP (Agent Commerce Protocol), co-developed by Stripe and OpenAI and announced September 2025, went live first with Etsy for physical goods. Its core mechanics: shared payment tokens that allow AI agents to transact without exposing underlying card data, configurable spend limits and time limits, and revocation capabilities for consumers who want to cancel agent authorization. Critically, ACP designates the original seller as the merchant of record — preserving pricing control, inventory management, fulfillment responsibility, and the post-checkout relationship. Stripe layers real-time fraud signals drawn from its network covering approximately 92% of all cards on top of token issuance.
TAP (Trusted Agent Protocol), announced as part of Visa Intelligent Commerce in October 2025, addresses three distinct dimensions: agent authentication (verifying that the entity initiating a transaction is a legitimate AI agent authorized by a real human, not a fraud bot), tokenization (securing credentials end-to-end), and rich buyer intent metadata (recording what the consumer actually wanted so merchants can validate that the agent's purchase matches the original instruction). Both protocols are explicitly designed as open standards intended to support interoperability across hundreds of anticipated AI agent platforms.
Jennifer Watkins of ARC provided the grounding reality: agentic commerce was absent from airline distribution conference agendas in spring 2025 but dominated every panel by November. Airlines' readiness varies sharply; ARC processes approximately 20% of transactions via NDC today, and Watkins estimates "orders" (the next evolutionary step) could have meaningful airline support by 2026.
For payments architects: The merchant-of-record question is the single most commercially significant design choice in these protocols. ACP's explicit preservation of the original seller as MoR is a deliberate architectural decision — one that protects commission economics for platforms and ensures brands retain post-booking upsell and data rights. Confirm that your token and vault infrastructure can support delegated agent credentials while maintaining MoR designation in settlement flows.
Finding 2: Stablecoins Are Being Actively Positioned as the B2B Settlement Layer for Agentic Travel Commerce
In the same payments panel, both Clara Liang (Stripe) and Gloria Colgan (Visa) described stablecoins not as speculative future technology but as a present-tense enabler for specific, high-priority use cases: real-time cross-border settlement, multi-currency treasury management, and payouts to international operators.
Stripe cited Starlink as a current real-world example of cross-border stablecoin settlement at scale — a non-travel precedent that directly maps to OTA-to-hotel payouts, bulk flight settlement, and wholesale accommodation reconciliation. The panel identified B2B flows as the immediate high-value use case for agentic commerce more broadly: high-volume, repetitive transactions such as OTA-to-hotel payouts or bulk flight bookings are natural starting points precisely because they do not require consumer-facing trust infrastructure to be rebuilt before value can be captured.
The framing from both Visa and Stripe was consistent: stablecoins address the latency and FX conversion inefficiency that makes multi-currency agentic transactions economically unattractive under conventional card rails. For a travel ecosystem where a single OTA platform may process thousands of supplier payouts across dozens of currencies daily, that latency and conversion cost is a material P&L line.
For treasury leaders and CFOs: The stablecoin settlement conversation has graduated from crypto experimentation to mainstream payments infrastructure planning at Visa and Stripe. The relevant near-term question is not whether to hold stablecoins on the balance sheet, but whether your settlement architecture can receive and reconcile stablecoin-denominated B2B payouts from counterparties who adopt these protocols. Engage your banking partners on their stablecoin settlement capabilities now.
Finding 3: Financial Institutions Are Building Vertically Integrated Travel Operating Systems — Not Embedding Third-Party OTAs
The session “Everyone's Trying to Sell Travel — The New Travel Seller Arena” featured Chase Travel CEO Jason Wynn alongside Hopper Technology Solutions President Dakota Smith and iSeatz CEO Kenneth Purcell, and constituted one of the most commercially consequential panels of the conference.
Wynn described Chase Travel as building a vertically integrated travel operating system, executing multiple strategic acquisitions: CX Loyalty (booking engine and fulfillment), Frost (luxury and corporate travel), The Infatuation (dining discovery), a payments and offers platform called Fig, and significant investment in airport lounges. Chase Travel has been growing at approximately 20% CAGR, targeting $12 billion in annual gross bookings. Wynn noted that 65% of new card acquisitions are millennials and Gen Z — a cohort that shows resilience to macro downturns but declining loyalty to traditional travel brands.
Purcell, who has powered the American Express Travel platform through iSeatz since 2012, sized the US financial institution travel marketplace at $45–50 billion in total gross bookings — a fraction of overall travel market volume, but growing rapidly. The key structural insight: financial institutions benefit from pre-existing consumer trust that traditional OTAs have progressively eroded.
Hopper's pivot is equally instructive. Dakota Smith confirmed that the B2B division — Hopper Technology Solutions — now represents over 90% of Hopper's total revenue, with the consumer app serving primarily as an incubation environment for products proven before institutional rollout. The Nubank Brazil case study illustrates the fintech-native advantage: rather than loyalty points (less compelling in a developing market), Nubank offers travelers 12-month installment plans at 0% APR — a powerful proposition in a market where average consumer credit APRs run at 35%. Hopper's agentic AI assistant, ACS Assist, handles approximately 3 million live customer service conversations per year and is being deployed with a major Japanese bank for Japanese-language frontline support.
Expedia's B2B segment was cited by multiple panelists as growing over 25% year-on-year, representing approximately 30% of Expedia's $100+ billion in gross bookings — evidence that OTA supply businesses will continue to benefit as the distribution layer behind financial institution platforms, even as B2C OTAs face structural pressure.
For payments and fintech leaders: The financial institution travel platform is becoming a genuine product category, not an ancillary loyalty feature. The Chase and Amex models require deep integration between card transaction data, travel booking data, and ancillary financial products (installments, FX, price protection). Fintech infrastructure that can bridge these data flows — cleanly, compliantly, and in real time — is the enabling layer these platforms need most.
Finding 4: Fraud and Agent Authentication Are Unsolved Problems That Will Define Agentic Commerce Viability
The payments panel was direct on the fraud problem without obscuring its difficulty. Visa's TAP explicitly names agent authentication — distinguishing a legitimate AI agent from a fraud bot — as one of its three foundational design problems. The challenge is structurally novel: in conventional card-not-present fraud, the adversary is a human or bot impersonating a human consumer. In agentic commerce, the adversary could be a malicious agent impersonating a legitimate agent, or a legitimate agent that has been compromised or misdirected.
The buyer intent metadata layer in TAP is designed to partially address this: by recording what the consumer explicitly authorized the agent to purchase, merchants gain a basis for challenging transactions where the agent's actual purchase deviates from the recorded intent. But the practical implementation of this at transaction velocity — across hundreds of agent platforms, millions of tokens, and real-time booking flows — is an open engineering problem.
ARC's Jennifer Watkins flagged the airline industry's specific concern: as payments teams navigate new agentic flows, fraud exposure is one of the three headline risks airlines have articulated (alongside disintermediation and distribution channel readiness). Airlines' existing fraud controls were calibrated for human-initiated, card-present and card-not-present transactions — not for agent-delegated credentials that may transact across multiple GDS, NDC, and direct channels simultaneously.
From the fintech innovation sessions: Acai Travel's post-booking AI architecture (a three-agent team — supervisor, front-office, and back-office agents deployed inside TMCs and OTAs) demonstrates that multi-agent architectures are already processing real airline policy, schedule change, and exchange transactions. The fraud surface in these workflows — where an AI back-office agent is processing cancellations and exchanges against live PNRs — is qualitatively different from anything the industry has previously managed.
For fraud and risk leaders: Engage with ACP and TAP working groups now. The authentication and intent-capture specifications are still being iterated, and the travel industry's specific transaction complexity (multi-leg, multi-supplier, multi-currency, high-value, time-sensitive) should be represented in the standards-setting process rather than accommodated after the fact.
Finding 5: Consumer Payment Method Preferences Are Fragmenting by Generation — and Refund Velocity Is the New Battleground
Credit cards remain the dominant payment method — six in ten travelers use them as the primary method for air, hotel, car rental, and in-destination activity spend. However, the generational segmentation is sharp: Baby Boomers remain firmly attached to credit and are unlikely to adopt new methods; Gen X has adopted online payment platforms but lags on mobile wallet adoption; Gen Z and Millennials are the primary drivers of digital wallet and mobile payment growth.
Across all cohorts, ease and security are the top selection factors, followed by purchase protection and rewards. The strategic insight from Schmid: these are table-stakes factors — they will not generate competitive share. The clearest differentiator for newer platforms is fast refunds — a feature valued not just by younger cohorts but by Gen X travelers managing rising costs. Travel brands competing on payment experience should invest in refund velocity as a differentiable product attribute, not a back-office operational metric.
In the high-spend "indulgent explorer" segment — travelers taking an average of 4.7 trips per year, significantly more likely to take international trips — credit card dependence is slightly lower and newer payment method adoption is slightly higher. This segment carries high rewards program membership and redemption rates, making them the natural early adopter base for any new payment instrument that also delivers loyalty value.
For product leaders: Refund velocity is now a consumer-facing product feature, not a settlement operations KPI. The brands winning on payment experience in 2026 will be those that have restructured their refund and cancellation workflows to deliver sub-24-hour resolution — a capability that AI-driven post-booking automation (as described by Acai Travel and Wenrix/DeepFlow) is making increasingly achievable.
Finding 6: The OTA-as-Bank Thesis Is Accelerating — and OTAs Are Taking It Seriously
Agoda CEO Omri Morgenshtern made one of the conference's most direct predictions: “OTAs will increasingly resemble banks.” He grounded this not in aspiration but in operational reality — the enormous fintech infrastructure already operating inside OTAs, covering payment optimization, FX management, and loyalty schemes. He argued that OTAs are the ideal distribution layer for financial products tied to travel expenditure — loans, savings products, FX hedging — and that this convergence will accelerate globally, not just in the US loyalty-driven market.
Morgenshtern specifically identified cross-border payment friction and foreign exchange anxiety as underserved consumer pain points that represent more impactful AI application areas than the conference's dominant topic of AI trip planning. He framed trust as the foundational currency: one bad interaction loses a customer, while solving payment friction earns lifetime loyalty. This is a distinct strategic lens from most AI discussions at the conference, which focused on discovery and booking interfaces rather than financial services layers.
Booking.com's strategy reinforces this: Rob Ransom (Chief Strategy Officer, Booking Holdings) cited payment security and price fairness assurance as core OTA value propositions that persist even as AI reshapes discovery. Booking.com has been consolidating its previously siloed B2B distribution arms — Booking.com, Agoda, and Priceline — into a unified go-to-market organization, a structural move that increases payment volume leverage in supplier negotiations.
For fintech and payments leaders: The OTA-as-bank trajectory creates both partnership and competitive pressure. OTAs with the transaction volume, customer trust, and data depth to offer embedded financial products are natural competitors to bank-affiliated travel platforms — but also natural distribution partners for fintech infrastructure providers who can power the payment, FX, and credit products they want to offer.
Finding 7: The Acai and Wenrix/DeepFlow Models Demonstrate That Post-Booking Is Already the Highest-ROI Payments AI Application
Two sessions from the Innovation track provided concrete operational evidence that AI-driven post-booking servicing — including refunds, exchanges, disruption handling, and ancillary processing — is the most commercially validated AI application in the payments adjacent space today.
Acai Travel (People's Choice Award winner, Phocuswright Innovation Launch 2025): CEO Riccardo Vittoria cited 60% reductions in call center handling time and 20–30 basis-point improvements in CSAT from customers deploying its three-agent AI team (supervisor, front-office, back-office). NDC complexity has increased call center handling times by 20–30% industry-wide, while experienced agents are retiring faster than they can be replaced. Acai has grown 50% per month in its two most recent months, reaching $2M ARR with sub-$10,000 CAC and sales cycles under three months.
Wenrix/DeepFlow (Runner-Up, Travel Innovator of the Year): CEO Amir Balaish framed the problem with precision — the industry has automated roughly 40% of simple in-flight ticket servicing tasks, but the remaining 60% consumes 82% of agent time. DeepFlow is trained on over 50 billion real-world servicing data points and delivers 93% automation rates with one major travel company, and over 90% accuracy on refunds and exchanges for CWT. Integration time is cited as one month. Pricing is per-transaction, complexity-tiered, and pay-as-you-go.
Both platforms handle the financial execution layer — the actual processing of refunds, penalty calculations, exchange fees, and ancillary adjustments against GDS, NDC, and airline policy systems. This is where the payments complexity lives in travel: not in the initial booking transaction, but in the subsequent lifecycle of changes, cancellations, ADM disputes, and disruption remediation.
For payments architects: The post-booking servicing layer is where AI is delivering the clearest near-term financial ROI in travel payments. The companies that build or partner with AI execution layers for this workflow in the next 12–18 months will have a structural cost and speed advantage in customer retention as booking complexity increases under NDC and agentic distribution.
Strategic Implications
The window to shape agentic commerce standards is closing. ACP and TAP are live but still iterating. Travel's specific transaction characteristics — high-value, multi-supplier, multi-currency, time-sensitive, emotionally complex — are not fully represented in protocols developed first for simpler e-commerce use cases like physical goods on Etsy. Travel payments executives who engage now with Stripe's ACP working group and Visa's Intelligent Commerce program will influence specification choices that will affect transaction processing costs, fraud liability, and MoR economics for a decade.
Financial institution travel platforms are not a niche. The $45–50 billion US financial institution travel market identified by iSeatz's Kenneth Purcell represents a segment growing faster than traditional B2C OTA volumes. The infrastructure requirements for these platforms — deep integration between card data, booking data, loyalty data, and embedded financial products — create significant demand for payments middleware, identity verification, FX hedging, and installment plan infrastructure that pure-play travel tech providers are not positioned to supply alone.
Stablecoin settlement is moving from pilot to planning horizon. The explicit endorsement of stablecoin-based settlement by both Visa and Stripe — two of the three largest payments infrastructure providers on earth — signals that treasury teams need a position on this in their 2026 planning cycles. The B2B OTA-to-supplier payout use case is the natural first adoption vector: lower regulatory friction than consumer-facing products, clear economic benefit from eliminating multi-currency conversion latency, and a transaction structure that maps well to smart contract settlement logic.
Digital identity will reshape authentication and reduce fraud friction simultaneously. PassiveBolt's Keyshare platform — which uses government-issued digital IDs as hotel access credentials — is operating in four US states with 35,000 travelers enabled, and Gartner projects 500 million standards-based digital identity credentials issued by end of 2026. For payments, this is significant: the same government-issued digital ID that unlocks a hotel room can anchor the authentication layer in an agentic commerce transaction, providing a fraud-resistant signal that no synthetic credential can replicate. The convergence of digital identity, agentic commerce authentication, and payment tokenization is the structural opportunity that payments architects should be modelling now.
Post-booking automation is the near-term P&L lever. While agentic booking remains a 2026+ phenomenon at meaningful scale, AI-driven post-booking servicing — handling refunds, exchanges, disruptions, and ancillaries — is live, validated, and producing measurable ROI today. The 60% handling time reduction documented by Acai Travel and the 93% automation rate documented by Wenrix/DeepFlow translate directly into reduced operating cost in a payment processing environment where NDC complexity is making manual handling more expensive, not less.
Action Items
Immediate (0–90 days)
Map your token infrastructure against ACP and TAP specifications. Both protocols are now publicly documented. Conduct a gap analysis: does your vault, token management, and credential delegation architecture support the spend limits, time limits, and revocation capabilities specified in ACP? Can your fraud scoring systems ingest buyer intent metadata in the TAP format? Assign an owner.
Engage ARC on NDC order readiness. ARC's Jennifer Watkins indicated airline order support — the next step beyond NDC — could arrive in 2026. Orders change the settlement and reconciliation model materially. Understand where your back-office reconciliation systems stand against the order paradigm before it arrives.
Audit refund velocity across your product. Phocuswright research identifies fast refunds as the primary payment product differentiator for all cohorts below Baby Boomer. Run an internal audit: what is your current p50 and p95 refund cycle time? Where are the process bottlenecks? This is a near-term revenue-impacting product gap, not a back-office operational detail.
Near-term (90 days – 12 months)
Pilot AI-driven post-booking servicing. The ROI case for AI execution layers in post-booking (refunds, exchanges, disruptions) is now validated at commercial scale by multiple providers. Issue an RFP or run a proof-of-concept with Acai Travel, Wenrix/DeepFlow, or equivalent providers. Target the 60% of complex edge cases that currently consume 82% of agent time as the initial scope.
Develop a stablecoin settlement position paper for the CFO. Commission an internal analysis of your B2B payout flows (OTA-to-hotel, OTA-to-airline, wholesale reconciliation) and model the economic impact of stablecoin settlement versus conventional card rail settlement at current FX conversion and settlement latency costs. This is the prerequisite for having an informed CFO conversation in the next planning cycle.
Assess financial institution travel platform opportunities. If your organization is a bank, card issuer, or fintech, evaluate whether the iSeatz/Chase/Hopper Technology Solutions model applies to your customer base. The $45–50B US market is growing and structurally advantaged by consumer trust. If you are a payments infrastructure provider, identify which of these platform builders represent pipeline for embedded payment, FX, installment, and loyalty settlement services.
Strategic (12+ months)
Participate in digital identity standards alignment. The convergence of PassiveBolt-style digital ID authentication, ACP/TAP agent authentication, and payment credential tokenization is the structural architecture that will define agentic commerce security for the next decade. Engage with relevant standards bodies (W3C Verifiable Credentials, ISO 18013-5 for mobile driving licenses, European Digital Identity Wallet consortium) to ensure your authentication requirements are represented.
Model the OTA-as-bank competitive scenario. Agoda's Morgenshtern named this trajectory explicitly. If major OTAs with hundreds of millions of customers begin offering embedded financial products — travel loans, FX hedging, savings products tied to travel spend — what is the competitive response for your organization? Scenario-plan this now, before the product launches make the question urgent.
Sessions to Watch
1. "AI Agents, Real Transactions — The New Travel Economy" (Clara Liang/Stripe, Gloria Colgan/Visa, Jennifer Watkins/ARC)
The definitive agentic commerce payments briefing of the conference. Watch for the side-by-side comparison of ACP and TAP, the stablecoin settlement discussion, and ARC's airline readiness assessment. Essential for anyone building or planning payments infrastructure for agentic workflows.
2. "Everyone's Trying to Sell Travel — The New Travel Seller Arena" (Jason Wynn/Chase Travel, Dakota Smith/Hopper Technology Solutions, Kenneth Purcell/iSeatz)
The financial institution travel platform session. Chase's vertical integration strategy, Hopper's B2B pivot and Nubank installment case study, and the $45–50B market sizing are all here. Required viewing for any bank, card issuer, or fintech evaluating travel as a product category.
3. "Payments to Priorities — The U.S. Traveler Mindset in 2025" (Alicia Schmid/Phocuswright)
The consumer data layer. Generational payment method segmentation, the fast-refund differentiator insight, and the indulgent explorer payment profile are actionable research findings that should inform product roadmaps and segment-level payment experience investments.
4. "Acai Travel — WINNER People's Choice Award — Phocuswright Innovation: Launch 2025" (Riccardo Vittoria/Acai Travel)
The best-documented case study for AI post-booking servicing ROI at the conference. The 60% handling time reduction, NDC complexity data, and competitive moat analysis are directly applicable for anyone evaluating AI-driven payment operations automation.
5. "Wenrix — RUNNER-UP Travel Innovator of the Year" (Amir Balaish/Wenrix)
The technical counterpart to Acai. DeepFlow's 50 billion training data points, 93% automation benchmark, and the "execution gap" framework provide the operational architecture language for scoping an AI servicing automation initiative.
6. "How Agoda's CEO Omri Morgenshtern is Pushing Boundaries Far Beyond Its Home Turf" (Omri Morgenshtern/Agoda)
Morgenshtern's OTA-as-bank thesis and his identification of cross-border payment friction and FX anxiety as the highest-impact underserved consumer pain points make this session essential for fintech leaders thinking about OTA partnership or competition strategy.
7. "Street Talk: Finance, Investment, M&A, IPOs in Travel" (Mark Mahaney/Evercore, Lloyd Walmsley/Mizuho, Emma Taylor/Barclays)
Emma Taylor's framing of fintech, ancillary services, and back-end AI ROI as the investment categories with strongest investor appetite — and the structural distinction between AI's impact on consumer-facing booking versus back-end payment operations ROI — is the analyst perspective payments leaders need for internal investment prioritization.
8. "Passivebolt — Global Startup Pitch Scaleup Winner" (Kabir Maiga/PassiveBolt)
The digital identity session. Gartner's 500 million digital credential projection by end of 2026, the convergence of government-issued ID and hotel access, and the authentication framework are directly relevant to anyone building the identity layer for agentic commerce transaction security.
*This brief was produced from session summaries captured at The Phocuswright Conference 2025. All data points, projections, and executive statements reflect conference-floor reporting and should be verified against primary sources before use in investment or strategic planning decisions.*