The supply side
of agentic commerce
is not ready.

Most merchants cannot be reliably verified by an AI agent, most PSPs have not adopted the protocols that make agent transactions safe, and the financial liability that accumulates in between sits in places no one in the payments stack has agreed to own.

scroll
73%
of online merchants are not agent-ready
1,200%
YoY growth in AI-sourced retail traffic
1/10
PSPs in production across ACP, UCP and AP2
$500B
US agentic e-commerce by 2030 (Bain, high case)
Method

How we measured the gap

47 interviews across card schemes, PSPs, acquirers, merchants and AI platforms, triangulated against Ballerine's merchant portfolio and published research from analysts, networks, and regulators.

Figure 01
Every side of the transaction, represented.
Conducted under background rules between February and May 2026. Quotes are paraphrased and attributed by role.
11
Card schemes
Visa, Mastercard and one regional network. Product, risk and policy teams.
14
PSPs and acquirers
Global Stripe-tier, mid-market processors, regional EU and APAC acquirers.
13
Merchants
Shopify Plus brands, DTC operators, two enterprise retailers.
9
AI and infra
AI platform eng, product, policy teams plus identity and trust vendors.
Source 1
01

Demand is growing an order of magnitude faster than supply. That mismatch is where the risk accumulates.

The audience for agentic commerce is moving roughly an order of magnitude faster than the infrastructure underneath it. AI-sourced retail traffic grew about 1,200 percent year over year going into 2026. Adobe Analytics measured 4,700 percent year over year growth in AI-driven visits to US retail sites in 2025. On the supply side, close to three quarters of online merchants are still standardizing their product pages for AI agents or have not started at all. Outside Stripe, no payment service provider can credibly claim production maturity across the three protocols that matter. The result is that AI agents are already routing real shoppers to merchants they cannot reliably verify, through payment rails that were not built for them, with the financial liability sitting in places no one has agreed on.

0% YoY growth in AI-sourced retail traffic going into 2026
24%
US consumers comfortable
letting AI complete a purchase
10
Active agent commerce
protocols, none interoperable
$190B
US agentic e-commerce
by 2030, Morgan Stanley low
02 · Insight

The merchant supply side is not ready, and the scale of the gap is larger than published estimates.

Independent assessments place roughly 40 percent of e-commerce businesses still in the process of standardizing product pages for agentic AI, and another 33 percent not started at all. Of the 13 merchants we interviewed, only 2 had completed structured product data work to the level AI platforms expect. 6 had started and stalled. 5 had not begun.

Figure 02: The agent-readiness ladder
Five tiers of merchant readiness. 73% have not reached the bar AI agents need.
73%
Not ready
L1: Not started
33%
L2: Standardizing
40%
L3: Catalog-ready
~20%
L4: Protocol-aware
~6%
L5: Agent-native
~1%
L1
Not started. No structured data, JS-rendered catalog, marketing copy in product fields.
33%
L2
Standardizing. Partial JSON-LD, incomplete attributes, no real-time inventory API.
40%
L3
Catalog-ready. Full JSON-LD, GTIN, real-time price and stock. No agent-specific work yet.
~20%
L4
Protocol-aware. ACP or UCP integration, machine-readable policy, agent-friendly returns.
~6%
L5
Agent-native. Full readiness plus branded agent strategy and post-purchase determinism.
~1%
Source 2
"We do not show up in ChatGPT for the products we sell. We did not even know that was the failure mode until a buyer asked us why."
Shopify Plus operator, DTC apparel, US
Impact
If 73 percent of merchants are not ready when the audience arrives at scale, the addressable market shifts by hundreds of billions of dollars toward the merchants that are. Default merchant status, once set, is sticky.
03 · Insight

AI agents are already recommending unreliable merchants. In production, today.

7 of the 9 AI platform and infrastructure interviewees independently raised merchant-side verification as a problem they did not yet have a real answer for.

"We have spent the last 18 months building trust into the agent. We have not spent any time building trust into the merchant the agent is about to send the user to."
AI platform product lead, US
Figure 03
The new fraud surface, quantified
Measured impact of agent-era attack vectors on financial institutions and merchants.
Rise in bot-initiated transactions (Visa)25%
Same metric, US specifically40%
Organisations seeing more AI-facilitated attacks97%
FIs expecting agent-linked fraud to increase78%
Organisations lacking AI/ML depth to defend71%
$4.5M average annual loss from AI-enabled bad actors per organisation
Source 3
04 · Insight

Outside Stripe, the PSP side is structurally behind.

Stripe co-authored the Agentic Commerce Protocol with OpenAI. Every other major PSP and acquirer is playing catch-up. Of the 14 PSP and acquirer interviewees, 11 described their current posture as "connectors". Only 3 described themselves as setting the agenda on a protocol, rather than implementing one.

Figure 04
The PSP agentic readiness race, May 2026
Production maturity across ACP, UCP and AP2 combined.
StripeCo-author
PayPalProduction
AdyenAP2 partner
WorldpayOpen MCP
Checkout.comACP enterprise
WorldlineUCP, AP2 build
MollieEU delegated
FISBank-led
NexiGoogle Cloud deal
SpreedlyUCP, ACP coming
Source 4
Figure 05
Consumer trust gradient, by authenticator
US consumer trust in AI agents to complete a purchase, by who authenticates the agent.
36%
Bank-backed AI
35%
Network-backed AI
Visa, Mastercard, Amex token.
28%
Independent agent
Source 5
Impact
The PSPs that can hand over a pre-verified, agent-ready merchant portfolio will be in the AI platform partnership conversation. The PSPs that pass raw merchant IDs will not. Portfolio quality is the new payments argument.
05 · Insight

Every major player is verifying the agent. No one is verifying the merchant.

Across all four interviewee groups, 38 of 47 (81 percent) raised the merchant trust question without being prompted.

Figure 06
The two-sided trust gap
Both sides must be verified for agent-mediated commerce to scale safely.
Side A / Being built

Is the agent who it says it is?

Cryptographic mandates, delegated payment tokens, agent-scoped credentials, bot vs agent discrimination at the edge.

Who: Visa, Mastercard, Amex, FIDO Alliance, Stripe, Google, Akamai, DataDome, Microblink, Vouched, Signifyd, Sift
Side B / Mostly open

Is the merchant actually safe to engage with?

Verified identity, accurate product claims, machine-readable refund terms, licensing, geographic restrictions, continuous behavioural monitoring.

Who: Ballerine, Feedonomics and MetaRouter on catalog. Most slots empty.
Source 6

What is needed, concretely, is a machine-readable merchant trust layer that an AI agent or a PSP can query before recommending, routing to or transacting with a merchant. The category is small today. Most of our interviewees expect it to consolidate through 2026 and 2027 and to become a required input to any agent-mediated transaction by 2028.

06

The Agentic Commerce Ecosystem

The companies building the payment and safety infrastructure that agentic commerce requires at scale.

07

Where the gaps are

Structural deficiencies ranked by urgency. G1 blocks everything else.

G1
Two-sided trust layer. Agent-side identity is being built. Merchant-side verification is mostly open.
Binding
G2
Returns, refunds, disputes. No standard for agent-initiated returns. No agreed liability allocation.
2026–2027
G3
Merchant attribution. Agents return a transaction. Everything before it is invisible. Retail media goes dark.
2026
G4
Multi-merchant carts. ACP handles single items today. Cross-merchant baskets pencilled for 2027.
2027
G5
Subscriptions and recurring. No native semantics for agent renewal, plan switching or cancellation.
2026–2027
G6
Promotions and loyalty. Personalized offers do not flow cleanly into AI channels. Merchants are blind here.
2027
G7
Governance and regulation. EU AI Act treats autonomous transactional systems as high-risk. US has no equivalent yet.
2027–2028
08 · Insight

The payments ecosystem will be redesigned before 2029.

Bain forecasts US agentic e-commerce at 300 to 500 billion dollars by 2030. Morgan Stanley puts it at 190 to 385 billion. McKinsey's broader unlocked-value estimate is 3 to 5 trillion dollars globally.

Figure 09
The timeline to 2029
Events that compress the strategic window from "we have time" to "we are out of time".
Now
Most PSPs in announcement mode, not production. 73% of merchants unready.
Q3 2026
Ninth Circuit ruling in Amazon v. Perplexity.Sets precedent on platform vs user authorization.
Q4 2026
Holiday 2026.First season with protocols at scale.
2026–27
First major US state passes agentic commerce consumer protection legislation.
2027
Returns and disputes standards consolidate. Multi-item carts ship as standard.
2027–28
Merchant trust layer becomes required input. Category consolidates.
2028
Federal US agentic commerce framework debated. First Fortune 500 B2B agent-to-agent deployments.
2029
Deadline.10–25% of US e-commerce through agent-mediated channels. Payments ecosystem redesigned. Winners locked in.
Source 7
Bottom line
87 percent of CTOs and heads of payments at financial institutions believe trust will be the most significant barrier to agentic payments adoption. Only 24 percent of US consumers are comfortable using AI to complete a purchase today.
About Ballerine
Ballerine is the AI-native merchant risk platform built for payment service providers, acquirers, marketplaces and financial institutions. The platform handles merchant onboarding, KYB, fraud detection and continuous monitoring across the full merchant lifecycle, with up to 80 percent fewer false positives, 3x faster onboarding and 70 percent reduction in manual review work. Ballerine is now extending its core verification capability into a machine-readable merchant trust layer purpose-built for the agentic commerce era.
Media contact  /  press@ballerine.com  /  ballerine.com
Study cutoff  /  May 16, 2026
Underlying data tables and analyst briefings available on request.
Sources
  1. Ballerine field interviews, Feb to May 2026. Triangulated against Bain, Morgan Stanley, McKinsey, Forrester, Visa, Mastercard, Adobe, Liminal, F-Secure, Aurascape, Riskified and Digital Applied published research, plus FY2026 SEC filings.
  2. Digital Applied agent readiness assessment, Feb 2026, triangulated with Ballerine merchant interviews and portfolio observations.
  3. Visa H2 2025 to H1 2026 bot transaction telemetry; Darwinium 2026 AI threat report; Mastercard FY2026 DEF 14A risk disclosures.
  4. acpready.com directory, March 2026, cross-referenced with Forrester April 2026 PSP analysis and Ballerine PSP interviews.
  5. Visa B2AI Report 2026, n=2,000 US consumers.
  6. Ballerine interview synthesis across 47 sources. 38 of 47 (81%) raised merchant-side verification unprompted.
  7. Synthesis of Bain, Morgan Stanley, McKinsey, Forrester, Visa and Mastercard forecasts with Ballerine interview judgements.