Agentic Payments Are Stalling in Consumer Commerce and Accelerating in B2B

The first wave of agentic payments did not collapse. It split in two.

That is the real message in Forrester’s latest market reset on agentic payments. Consumer commerce inside answer engines remains weak, uneven, and far less adopted than the hype implied. But underneath that disappointment, the infrastructure race is getting more serious. Protocols are maturing. payment networks are publishing specs. PSPs are repositioning themselves as connectors for AI agents. And the strongest early use cases are shifting away from a chatbot impulse-buy fantasy and toward B2B workflows where payments are repeatable, constrained, and easier to automate safely.

This matters because a lot of recent coverage has treated agentic commerce like a single market. It is not. There is a big difference between asking a consumer to trust an AI assistant with discretionary shopping inside a chat interface and letting software agents execute narrow, pre-approved payments inside procurement, API usage, or recurring operational flows. The first use case depends on behavior change and trust at the point of purchase. The second depends on standards, permissions, and workflow economics.

Those markets do not move at the same speed.

For brands, retailers, payment providers, and operators thinking about AI discovery, that distinction is now the most important one. If you still read every agentic commerce headline as proof that consumer checkout is about to go mainstream, you are reading the wrong layer of the stack. The faster story is infrastructural. The earlier revenue opportunity may be B2B. And the companies that win first may be the ones building the rails, not the ones chasing flashy in-chat conversion demos.

The consumer promise hit reality faster than expected

Forrester’s latest post is unusually useful because it cuts through a lot of narrative fog. The firm says consumer adoption of agentic commerce remains low and relatively stagnant. It specifically points to weak adoption of OpenAI’s Instant Checkout from debut to discontinuation, and says consumer interest in AI agents making purchases is still lukewarm even if curiosity is gradually growing.

That aligns with what the market has been signaling in fragments for weeks. The problem was never that people would never use AI for shopping. The problem was the assumption that they would quickly hand over the final transaction itself.

That assumption always had several weaknesses.

First, checkout is not only a discovery moment. It is a trust moment. A user might happily ask an AI system to compare products, summarize reviews, or shortlist options. That does not mean the same user is ready to let a model complete payment credentials, loyalty logic, shipping preferences, and order confirmation inside the same interface. Discovery can feel reversible. Payment does not.

Second, consumer ecommerce behavior is full of edge cases. Coupon expectations, delivery preferences, returns anxiety, subscriptions, cart-building, family approval, sizing uncertainty, and inventory substitution all complicate a supposedly simple “buy now” command. A clean demo flow can hide how messy the real purchase path actually is.

Third, the economics have not yet justified the leap. If in-chat checkout converts worse than a merchant’s own site, the merchant has no reason to force the experience. Forrester points to OpenAI’s move away from Instant Checkout and toward stronger product discovery and app-based commerce paths. That is not a small product tweak. It is a market correction.

OpenAI’s own commerce documentation still shows the seriousness of its payment architecture. The delegated payment spec explains how payment details can be securely handed to the merchant or its PSP, with single-use constraints, allowances, and merchant-owned settlement. That is real infrastructure work. But the presence of a secure spec does not automatically create consumer demand. It only means the rails can exist if and when the use case earns them.

That distinction is why the market looks less like a failed thesis and more like a premature one.

OpenAI and Google are still building because distribution still matters

The weak consumer adoption story does not mean the major platforms are backing away from commerce. It means they are adjusting where they expect the first serious usage to happen.

Forrester still calls Google’s Universal Commerce Protocol and OpenAI’s Agentic Commerce Protocol the two frontrunners, largely because of their distribution. That is a crucial point. Protocol races are not won only by elegance. They are won by where intent already lives.

Google disclosed that the Gemini app had 750 million monthly active users as of its Q4 2025 earnings materials. OpenAI says ChatGPT now reaches more than 900 million weekly active users. Those numbers matter because protocols attached to massive answer-engine surfaces get an unfair advantage. They are closer to the user intent, closer to the merchant relationship, and closer to the developer ecosystem that has to implement the rails.

That is why the infrastructure layer still matters even when headline consumer adoption is disappointing.

The control point in agentic commerce is not just the checkout button. It is the environment where product discovery, recommendation, identity, payment authorization, and follow-up actions can be stitched together. That is also why Searchless recently argued that Google’s UCP is turning product feeds into AI shopping infrastructure. The deeper story is not one company winning a feature war. It is that the market is building machine-readable standards for how AI systems can discover, trust, and act.

So yes, consumer checkout hype has cooled. But the strategic significance of these protocols has not. If anything, the reduced hype makes it easier to see what they are really for: not a magic button for retail disruption, but a coordination layer for machine-mediated commerce.

The cleanest early use case is not shopping, it is machine-to-machine payment

The most important line in Forrester’s update is not about OpenAI at all. It is the shift toward agent-to-agent payments.

Stripe and Tempo recently launched the Machine Payments Protocol, with Visa as a design partner. The core idea is straightforward. AI agents and services need a standard way to coordinate payments programmatically without turning every transaction into a custom integration mess. Stripe describes MPP as a specification for agents and services to coordinate payments programmatically. Visa has already published technical specifications and SDK support tied to that direction.

This matters because machine payments are a much cleaner starting point than consumer checkout.

A consumer purchase inside chat is messy, emotional, brand-sensitive, and full of exceptions. A machine payment for API access, network usage, procurement triggers, or repeatable back-office tasks is the opposite. It can be policy-bound. It can be capped. It can be logged. It can be audited. It can run within explicit permissions. And crucially, it often lives in environments where the buyer values speed, automation, and repeatability more than browsing pleasure.

That is why Forrester says this likely takes off in B2B payments first.

The logic is strong. B2B workflows already tolerate system-driven approvals, procurement rules, recurring billing, and delegated authority. Enterprises are also rapidly normalizing the idea that agents can perform real work inside finance, sales, engineering, and operations. OpenAI’s own enterprise update says AI now drives more than 40 percent of its revenue mix and is moving from experimentation into real deployment across business workflows. Once agents are trusted to research vendors, score leads, file support actions, or orchestrate engineering tasks, the next natural question is which narrow financial actions they can also complete.

That is a much more believable adoption curve than the idea that millions of consumers suddenly want to buy everything through a chat window.

Why the market is splitting into behavior problems and infrastructure problems

The easiest way to understand the current state of agentic payments is to separate two kinds of friction.

Consumer friction is mostly behavioral

Consumers need to trust the interface, understand what the agent is doing, feel confident in the merchant, and believe the AI-driven path is actually easier than the alternatives. That is a high bar. It requires product design, branding, user education, and habit change.

This is why so many consumer stories are still discovery-first rather than transaction-first. Users are comfortable asking, comparing, and exploring with AI. They are much less comfortable surrendering the final act of payment unless the category is simple, the merchant is trusted, and the purchase feels reversible.

B2B friction is mostly infrastructural

In B2B, the challenge is different. The question is not whether a finance or procurement team conceptually understands software doing work. It is whether the workflow is controlled enough to automate safely. That means identity, permissions, tokenization, PSP support, logging, spending limits, dispute handling, and standards compatibility.

Those are hard problems, but they are engineering and governance problems. Markets solve those more predictably than they solve mass behavior change.

This is why the narrative split matters so much. If you think agentic payments is one market, the current picture looks confusing. If you treat consumer and B2B as separate adoption curves sharing some rails, the picture becomes much cleaner.

Consumer commerce says, “Interesting, but not trusted enough yet.”

B2B infrastructure says, “This is annoying to implement, but economically rational once the standards settle.”

The second usually wins first.

Merchant PSPs are quietly repositioning for the connector era

Forrester also notes that merchant payment providers are trying to become connectors in this emerging ecosystem. That may sound minor, but it is one of the clearest strategic signals in the whole piece.

When a market fragments into multiple protocols, answer engines, trusted-agent standards, and merchant environments, someone has to translate between them. PSPs are well positioned to play that role because they already sit at the intersection of merchants, tokenization, compliance, fraud systems, and settlement logic.

That means the payment provider opportunity is larger than “support AI checkout.”

The bigger opportunity is to become the compatibility layer between merchants and many forms of agentic demand.

If merchants eventually need support for Google UCP, OpenAI ACP, Visa-linked machine payment standards, AI-specific payment tokens, and identity or trust protocols, the PSP that makes those rails interoperable becomes more valuable. In that world, the winners are not only the engines that originate intent. They are also the connectors that make intent executable.

This is why the connector strategy matters for Searchless readers. It suggests that visibility in the answer-engine era will not be determined by content alone. It will increasingly be determined by which systems can expose trusted, machine-usable actions. That is true in publishing, in SaaS, and now in payments.

What this means for retailers, brands, and operators right now

Most operators should stop asking one lazy question, “When will agentic checkout take off?” and start asking three better ones.

1. Which parts of the journey are discovery, and which parts are execution?

If AI is already shaping product discovery, category comparison, and shortlist formation, then you already need machine-readable product, inventory, pricing, and policy data even if direct in-chat payment remains weak. Waiting for checkout adoption misses the point. Discovery influence arrives first.

2. Which workflows are repeatable enough for delegated or machine payment?

For B2B operators, the right first use cases are not glamorous. They are narrow, repeated, and easy to govern. Think API usage, supplier interactions, recurring procurement events, approved spend bands, or internal agent workflows where the financial action is bounded. That is where trust can be built economically.

3. Which partner in your stack is becoming the control point?

In some categories, the answer-engine platform will be the control point. In others, the PSP or merchant platform will matter more. In others, a retailer-owned assistant or app layer will keep control. The strategic mistake is assuming the interface you see is the only leverage point that matters.

For brands, the practical implication is simple. You should build for machine legibility before you build for machine conversion.

That means clearer product data, cleaner policy signals, structured merchant identity, easier handoffs, and content that makes your offers interpretable to recommendation systems. It also means being realistic. Most brands do not need an “AI checkout strategy” yet. They need an AI readiness strategy for commerce surfaces that increasingly shape who gets recommended in the first place.

If you want to understand whether your brand is legible to these systems before the transaction layer matures, the more useful move is to run an AI visibility audit, not to chase every flashy commerce pilot.

The first winners may be boring on purpose

The market tends to reward dramatic stories. “Consumers buy everything through chat” is dramatic. “Standards emerge slowly and B2B machine payments adopt first in narrow workflows” is less exciting.

It is also much more plausible.

Many of the most valuable infrastructure markets look boring right before they matter. APIs, cloud primitives, fraud layers, identity rails, and procurement systems rarely win headlines because they do not feel revolutionary in isolation. But once they become standard, they shape the economics for everyone above them.

Agentic payments increasingly looks like that kind of market.

The near-term winners may not be the brands with the coolest shopping demos. They may be the companies that make delegated payment safe, machine payment auditable, merchant connectivity easier, and multi-protocol support less painful. They may be the systems that reduce friction for other systems.

That changes how Searchless readers should interpret the category. The important question is no longer whether agentic commerce is real. It clearly is. The better question is where the first durable economic behavior shows up.

Right now, the strongest answer is not consumer checkout. It is B2B automation.

The broader Searchless takeaway

This is also a useful reminder about the post-search economy more broadly.

Markets rarely change in one clean jump from old behavior to new behavior. They usually split. One layer moves fast because the incentives are strong and the friction is mostly technical. Another layer moves slower because the friction is human, legal, emotional, or organizational.

Agentic payments is following that pattern almost perfectly.

Consumer shopping through AI will keep improving. Some categories will work sooner than others. Trusted merchants, low-friction products, persistent carts, loyalty-linked flows, and better handoff design will all help. But the more immediate commercial shift is underneath the surface, where protocols, PSP connectors, machine-payment standards, and enterprise agent workflows are quietly making payments executable by software.

That is why today’s cleanest strategic reading is this: the first phase of agentic payments did not produce a consumer breakthrough. It produced a market map.

And that map says B2B will likely get there first.

Big picture for operators

If you lead growth, commerce, payments, or digital strategy, the next 12 months should not be spent waiting for consumer AI checkout to become normal. They should be spent preparing for a world where AI systems influence product discovery now, payment rails become more programmable, and machine-readable trust becomes a competitive advantage.

The companies that separate hype from actual adoption curves will make better bets. The ones that lump everything under “agentic commerce” will keep misreading where the opportunity is.

The split is visible now.

That is what makes this a real inflection point.

Sources

• Forrester: Agentic Payments In B2C Commerce, Where We Are Now
• OpenAI Developers: Delegated Payment Spec
• OpenAI: Scaling AI for everyone
• OpenAI: The next phase of enterprise AI
• Alphabet Q4 2025 earnings release PDF
• Stripe and Tempo launch Machine Payments Protocol
• Visa technical support for Machine Payments Protocol

FAQ

Are agentic payments failing?

No. The category is not failing. The consumer in-chat checkout story is proving slower and weaker than expected, while the infrastructure and B2B automation side is gaining momentum.

Why is B2B likely to adopt first?

Because B2B workflows are more repeatable, easier to permission, easier to audit, and more economically rational to automate than discretionary consumer purchases inside a chat interface.

What should brands do now?

Prioritize machine legibility first. Improve product and policy data, evaluate AI-driven discovery surfaces, and prepare payment and commerce partners for a more programmable future instead of over-investing in consumer demo theater.

Explore what that means for your brand’s discoverability at audit.searchless.ai.

If you want the broader measurement layer behind that shift, see AI visibility.