This is the full editorial research companion to Nuvei’s interactive How Agents Pay experience, covering agentic commerce foundations, consumer survey data and Nuvei Agentic payment infrastructure.
Full research and findings
Overview and the evolution of agentic commerce
The proof
AI agents have become a discovery layer between consumers and merchants. The next step is the agent that pays on the consumer's behalf. With Visa, Nuvei completed the first agentic payment authorized across multiple issuing partners in Europe, with purchase, authorization, and payment completing inside the agent. That merchant-first capability has a name: Nuvei Agentic.
Agentic commerce foundations
Agentic payments, agentic discovery: what do these terms mean in practice? This section defines them, explains how agents change the relationship between buyer, merchant, and payment, and lays out the stages in the shift from assistive AI to delegated spending.
The three stages of online commerce
Commerce has reinvented itself twice in twenty-five years, and each time the payment system had to be rebuilt to match. Agentic is the third stage, and it is arriving faster than the infrastructure built for the first two can absorb. McKinsey projects agentic commerce could drive $3 trillion to $5 trillion in global consumer transactions by 2030, with up to $1 trillion in the US retail market alone.
At each stage, it was not only the interface for purchasing that changed. Both consumer behavior and the payment layer had to be re-engineered before the new model could scale. The companies that built the rails early set the terms for everyone who arrived later. Agentic is the stage still being defined, and the first live proofs are already here.
Defining the age of agentic : commerce vs discovery
"Agentic commerce," "AI-driven purchasing," and "LLM shopping discovery" are terms frequently used interchangeably across the industry, but they describe different things. Agentic commerce is the medium through which consumers start making purchases. And it has two phases: the AI-led discovery phase, happening today and the agentic payment phase, which is still being built. One is about where decisions are made; the other is about how the money moves. Let's break them down.
Agentic payments are the next step. The consumer sets a mandate in advance, say a limit of two hundred dollars on a defined task. The agent then initiates and completes the transaction itself, inside its own environment, with no redirect to a merchant checkout. The point of initiating the payment moves from the human, the way it was done in the eCommerce world, to the machine.
That shift - from a person initiating the payment in the AI-discovery and eCommerce stage to one where an agent makes the payment on the user's behalf - is where the current infrastructure has the biggest bottleneck.
For decades, the payments infrastructure relied on a human at every stage of the purchase. A transaction with no human at the checkout has to carry trust in two directions at once: trust in the consumer, validated as it is today via KYC, and trust in the agent acting for them, validated via new protocols, such as Know Your Agent.
As Mentxu Trivino, VP of Payments Partnership at Nuvei puts it :
The question is no longer whether a payment can be processed. It is who, or what, is authorizing it, what that actor is allowed to do, and who is liable when something goes wrong.
This next era of agentic commerce requires a different infrastructure and different rules of governance, making sure the agent initiating the payment is verified, trustworthy, and carries the credible user intent. This is exactly where Nuvei is building.
First-party and third-party agents
There is a second distinction that decides how merchants meet this shift. A first-party agent is one a merchant runs itself, inside its own site, app, and customer relationship. A third-party agent is a public assistant like ChatGPT or Gemini that acts across many merchants at once. The two raise different trust questions, and the merchants Nuvei works with are clear that they will meet the first-party version first. That ordering shapes everything Nuvei is building.
But before we get into that, let's see what consumers have to say about how they use AI when making purchases.
Consumer survey: what people will let AI do
Research: What the consumer path to agentic commerce looks like.
A Nuvei survey of US consumers on how they shop with AI today, what they will let an agent do, and what they need before they will let it pay.
Delegated AI purchasing will run on trusted payment rails .
Consumers are already using AI as a shopping layer.
Consumers are ready to delegate purchasing tasks, but not purchasing authority. When asked about their ideal AI shopping experience:
The hesitation is about governance, not capability.
Concern concentrates on money moving and the ability to undo it.
Consumers would be more comfortable with supervision they can see.
Trust is expected to be shared across all parties in a payment.
The Nuvei data shows that consumers want limits, approvals, reversibility, and clear accountability before they will trust AI to move money on their behalf. Next, Nuvei Agentic: what it takes to deliver them in the payment layer.
Nuvei Agentic: bringing payments into AI purchasing
When an AI agent is ready to pay, it calls Nuvei Agentic: one integration that carries proof the consumer authorized the purchase, confirms the agent is who it claims to be, and clears the payment across networks. Built on the infrastructure Nuvei already runs for thousands of merchants.
With Visa, Nuvei completed the first agentic payment authorized across multiple issuing partners in Europe. A merchant's AI agent bought a product on a shopper's behalf and paid inside the agent, with no hand-off to a separate checkout. It ran on Arvato Systems' platform for the fashion brand Kings and Priests, settled on live Visa rails, and stayed inside shopper-set guardrails for spend and category. The pilot keeps purchase, authorization, and payment inside the agent, and establishes the proof point for Nuvei Agentic: a protocol- and network-agnostic layer any AI agent can call to pay.
Nuvei completes first-party in-agent payment with Visa; unveils merchant-led agentic payments strategy.
First live in-agent purchase authorized across multiple issuers on Visa rails. Nuvei Agentic delivers what merchants asked for: first-party agent capabilities today, third-party agents when the market demands.
First-party first
This is not a guess at where the market is going. At Nuvei's Global Customer Advisory Board, the priority was clear: start with first-party agents, the AI experiences merchants run themselves, on their own sites and apps, inside their own customer relationships. Make that work first, then extend the same controls to public, third-party agents as the market develops. The pilot demonstrated exactly that. The agent ran on the merchant's own platform, and the brand kept the customer relationship. First-party first means merchants learn this channel on ground they own, and when the public-agent moment arrives they are extending something that already works.
The two building blocks
A single transaction was never the hard part. The hard part is making one integration handle whatever protocol an agent uses, carry verifiable proof that the consumer authorized the purchase, and clear across any network, with fraud models built for how agents behave rather than how people do. Two pieces make that work.
BLOCK 1: PROTOCOL COMPATIBILITY LAYER
A merchant integrates once and accepts whichever standard an agent uses, whether the Agentic Commerce Protocol, AP2, or the Model Context Protocol. Payments route across both Visa's and Mastercard's agentic frameworks, with Nuvei certifying against Visa Intelligent Commerce and Mastercard Agent Pay, rather than betting on one.
Know Your Agent adds identity and governance. It registers and credentials agents, validates the consumer's mandate before any money moves, scores the agent's track record, and keeps every action auditable. It answers the three questions agentic payments turn on: who is this agent, did the consumer authorize this, and can we prove it later.
Open by design
Nuvei Agentic is a neutral layer, not a closed system. A merchant integrates once and keeps both the customer relationship and the merchant-of-record. As the agents and the rails evolve, the integration keeps working. The companies that shape this infrastructure now will set the terms for the next decade, and Nuvei intends to be the open layer merchants control.
Already running at scale
The pieces an agent payment depends on already run at scale today: a Level 1 PCI-certified environment, more than 720 payment methods across 200-plus markets, and real-time fraud and risk tooling. Because these capabilities reach merchants through the platforms they already use, they can switch them on without re-engineering their payments stack, and without giving up the customer relationship or the merchant-of-record.
Availability
Nuvei is targeting initial availability in the second half of 2026, covering protocol compatibility, the Know Your Agent registry and agent risk scoring, network certifications, and a developer sandbox for the teams building these experiences. The pilot is the first proof point, and Nuvei intends to be honest at each step about what is in pilot and what is live.
Complete consumer survey response data
Each table reproduces the published total-sample response percentages behind the report charts. Certain questions allow multiple responses, so percentages need not total 100%.
I find AI platforms useful for researching products and comparing prices .
58% find AI useful for researching products and comparing prices.
| Survey response | Respondents |
|---|---|
| Agree | 58% |
| Neutral | 23% |
| Disagree | 19% |
I am using AI more while shopping than I was three months ago.
39% say they are using AI while shopping or researching purchases more than they were three months ago.
| Survey response | Respondents |
|---|---|
| Using more | 39% |
| Not more | 37% |
| Neutral | 24% |
How comfortable are you with AI handling parts of a purchase ?
Consumers want to stay in control: 71% want to review purchases before checkout, while only around a quarter are comfortable storing payment details or letting AI complete a purchase automatically.
| Survey response | Respondents |
|---|---|
| Want to review before checkout | 71% |
| Comfortable letting AI complete low cost or routine purchases | 31% |
| Comfortable letting AI complete a purchase automatically | 28% |
| Comfortable storing payment details on an AI platform | 27% |
What is your ideal AI shopping experience ?
Most consumers want to keep approval in their own hands; only 7% are open to any form of unsupervised AI buying.
| Survey response | Respondents |
|---|---|
| AI helps me research, I buy | 40% |
| I do not want AI involved | 34% |
| AI helps but I approve each purchase | 19% |
| AI makes routine purchases within limits I set | 6% |
| AI fully automates shopping | 1% |
What is the maximum you would let AI spend without your approval ?
56% would never let AI spend without approval, and among those who would, almost all cap it under $100.
| Survey response | Respondents |
|---|---|
| Never without approval | 56% |
| $11 to $50 | 15% |
| $51 to $100 | 13% |
| Under $10 | 11% |
| $101 to $250 | 4% |
| Over $251 | 1% |
What are your biggest concerns about letting AI buy on your behalf?
Fraud and unauthorized charges (76%) is the single biggest concern, ahead of buying the wrong item, exposing payment information, or losing control over spending.
| Survey response | Respondents |
|---|---|
| Fraud or unauthorized charges | 76% |
| Buying the wrong item | 33% |
| Exposing payment information | 22% |
| Losing control over spending | 22% |
| How personal data is used or shared | 16% |
| Hidden fees or unexpected charges | 13% |
| Difficulty disputing AI purchases | 7% |
| Wrong size or version | 7% |
What is your overall view of AI-powered commerce ?
Opinion is still forming: the most common response is uncertainty (28%), followed by the view that AI will increase fraud (23%).
| Survey response | Respondents |
|---|---|
| Unsure | 28% |
| Will increase fraud and scams | 23% |
| Will make shopping more efficient | 15% |
| Will reduce control over spending | 12% |
| Will simplify routine purchases but not major ones | 12% |
| Will improve personalization | 11% |
What would make you more comfortable using AI platforms for shopping or payments?
Spending limits (61%) and human approval checkpoints (59%) would do the most to make people comfortable.
| Survey response | Respondents |
|---|---|
| Spending limits | 61% |
| Human approval checkpoints | 59% |
| Fraud protection guarantees | 27% |
| Tokenized payment details | 24% |
| Easy refunds or reversal | 19% |
| Real time alerts | 14% |
| Merchant verification | 12% |
The single most important safeguard
Asked to choose one safeguard, consumers pick human approval requirements (19%) and real-time approval notifications (14%) ahead of other measures, but a third say none of the listed safeguards would fully convince them.
| Survey response | Respondents |
|---|---|
| None of these | 33% |
| Human approval requirements | 19% |
| Real time approval notifications | 14% |
| Instantly reverse purchases | 9% |
| Fraud protection guarantees | 9% |
| Biometric verification | 7% |
| Spending caps | 6% |
| Tokenized credentials | 3% |
How important is the ability to reverse or cancel an AI purchase ?
84% say the ability to reverse or cancel an AI-initiated purchase after checkout is extremely or very important.
| Survey response | Respondents |
|---|---|
| Extremely important | 64% |
| Very important | 20% |
| Somewhat important | 10% |
| Not very important | 2% |
| Not important at all | 4% |
Who should be financially responsible for a wrong AI purchase?
If an AI platform makes a wrong or unauthorized purchase, 52% say the AI platform should be financially responsible.
| Survey response | Respondents |
|---|---|
| The AI platform | 52% |
| Not sure | 16% |
| Shared across parties | 9% |
| The retailer or merchant | 8% |
| My bank or card company | 6% |
| The payment company | 5% |
| The user | 5% |
Who do you trust to handle AI-powered shopping or payment tools securely?
Consumers trust banks (around 49%) and payment providers (47%) to handle AI-powered shopping and payment tools securely more than retailers, Big Tech companies, or AI companies.
| Survey response | Respondents |
|---|---|
| Banks and card issuers | 49% |
| Payment providers | 47% |
| Brands you already shop with | 44% |
| Retailers | 37% |
| Big Tech | 32% |
| AI companies | 30% |
What matters most when allowing AI-assisted purchases?
52% say trust depends equally on the AI platform, merchant, payment provider, and bank.
| Survey response | Respondents |
|---|---|
| All equally important | 52% |
| The AI platform | 21% |
| The merchant | 13% |
| The payment provider | 8% |
| The bank | 7% |
Where would you most likely want to complete an AI-assisted purchase ?
58% would still prefer to complete the purchase on the retailer or brand's own site or app; 12% would complete it inside the AI platform.
| Survey response | Respondents |
|---|---|
| Retailer or brand site or app | 58% |
| No preference | 22% |
| In the AI platform | 12% |
| Through a marketplace | 8% |
Would you feel comfortable allowing AI to complete routine purchases without a traditional checkout page ?
61–62% are not comfortable letting AI complete purchases without a traditional checkout page.
| Survey response | Respondents |
|---|---|
| No | 62% |
| Only for low cost or repeat purchases | 18% |
| Yes | 14% |
| Only with spending limits in place | 7% |
Which types of purchases would you be comfortable allowing AI to complete on your behalf ?
Where people would let AI buy, it is mostly routine: groceries (28%), household essentials (21%), and subscription renewals (19%); luxury and financial products sit lowest, and 51% would allow none.
| Survey response | Respondents |
|---|---|
| None of the above | 51% |
| Groceries | 28% |
| Household essentials | 21% |
| Subscription renewals | 19% |
| Clothing and apparel | 16% |
| Electronics | 11% |
| Entertainment and gaming | 10% |
| Healthcare products | 10% |
| Travel bookings | 9% |
| Financial products | 6% |
| Luxury goods | 6% |
Research methodology
The findings on this page draw on a survey of 1,000 US consumers, fielded online in 2025 and weighted to be representative of the adult US population across gender, age, region, income, ethnicity, education, employment, marital status, and household composition. The sample composition is detailed below.
Gender
| Category | Share of respondents |
|---|---|
| Male | 51% |
| Female | 49% |
Age
| Category | Share of respondents |
|---|---|
| 18 to 24 | 12% |
| 25 to 34 | 19% |
| 35 to 44 | 21% |
| 45 to 54 | 19% |
| 55 to 64 | 13% |
| 65+ | 16% |
Region
| Category | Share of respondents |
|---|---|
| Northeast | 18% |
| Midwest | 23% |
| South | 36% |
| West | 23% |
Income
| Category | Share of respondents |
|---|---|
| Below $25k | 17% |
| $25 to $35k | 10% |
| $35 to $50k | 10% |
| $50 to $75k | 19% |
| $75 to $100k | 14% |
| $100 to $150k | 17% |
| $150 to $200k | 7% |
| $200k+ | 6% |
Ethnicity
| Category | Share of respondents |
|---|---|
| Caucasian | 68% |
| African American | 18% |
| Asian | 5% |
| Other | 4% |
| Mixed | 3% |
| American Indian | 2% |
| Pacific Islander | 1% |
Education
| Category | Share of respondents |
|---|---|
| Bachelors | 27% |
| High school | 24% |
| Some college | 17% |
| Masters | 12% |
| Associate | 9% |
| PhD | 3% |
| Technical or vocational | 3% |
| Some high school | 3% |
| Professional | 2% |
Employment
| Category | Share of respondents |
|---|---|
| Full time | 45% |
| Retired | 17% |
| Part time | 9% |
| Not employed | 9% |
| Self employed | 6% |
| Homemaker | 5% |
| On disability | 4% |
| Student | 3% |
| Business owner | 1% |
Marital status
| Category | Share of respondents |
|---|---|
| Married | 42% |
| Single | 36% |
| Living with a partner | 9% |
| Divorced | 8% |
| Widowed | 4% |
Children under 25 in household
| Category | Share of respondents |
|---|---|
| No | 63% |
| Yes | 37% |
Explore the original interactive Agentic Commerce report for chart filtering, visualizations and the complete chapter navigation.
