Two data points published this week, read together, describe the central tension in retail AI strategy for the rest of 2026. The first comes from the PYMNTS Intelligence report published July 1, based on surveys of 1,185 merchants across the United States, Brazil, and the United Arab Emirates. AI shopping assistants are now the most frequently cited digital capability retailers plan to invest in during the next three years. 37 percent of merchants named them their top planned investment. The second data point comes from The Agile Brand Guide on July 5: 47 percent of U.S. consumers are comfortable using AI for price comparison, while only 24 percent are comfortable with AI-powered payments.
Read those two numbers together. Retailers are rushing toward AI shopping assistants as their primary digital investment. Meanwhile, three out of four consumers will not let the agent complete the transaction. The infrastructure layer is moving faster than the trust layer. And the retailers who do not understand that distinction will build deployments that deliver excellent discovery and stall at checkout.
The bottleneck in agentic commerce in 2026 is not the API connection, the catalog quality, or the checkout infrastructure. It is whether the consumer trusts the agent enough to stop being a human in the loop. That is a different problem, and it requires a different solution.
What the Data Actually Shows This Week
The PYMNTS report documents a clear reallocation of retail technology budgets. AI shopping assistants have jumped to the top of the investment priority list. Consequently, investment in several features that became standard components of digital commerce has declined: cross-channel shopping capabilities, stored payment methods, and mobile apps are all receiving less development emphasis. Retailers are making a bet that AI assistants will deliver more return than the capabilities that came before them.
The consumer trust data complicates that bet significantly. The 47 percent comfortable with AI price comparison is an encouraging number. It means consumers have already accepted AI in the research and discovery phase of the shopping journey. However, the 24 percent comfortable with AI payments reveals a structural limitation. Acceptance of AI in discovery does not automatically extend to acceptance of AI in execution. The same consumer who happily uses an AI assistant to find a product may have no intention of letting that assistant complete the purchase. Furthermore, discomfort with AI-powered payments appears across age groups and markets. This is not a generational gap that will resolve itself over time.
Why Consumer Familiarity With AI Does Not Create Trust in AI Payments
The instinct for most retail technology teams encountering this data is to frame it as an adoption curve: consumers are not yet familiar with AI payments, but familiarity will drive comfort over time. That framing is incorrect, and acting on it produces the wrong investment decisions. Familiarity and delegation trust are not the same thing, and they do not converge automatically.
Using AI and Delegating to AI Are Categorically Different Decisions
When a consumer uses an AI assistant to compare prices or find a product, they retain decision authority. They review the recommendation, evaluate it, and choose whether to act. When an AI agent completes a payment on their behalf, they delegate decision authority. That is a fundamentally different relationship with the technology. The trust required for delegation is categorically higher than the trust required for consultation. Specifically, a consumer can become highly familiar with AI as a research tool. That familiarity does not extend to authorizing it to spend their money. Those are two different trust thresholds, and they do not converge without deliberate effort from the retailer to build the second one.
The Visa Protocol Solves a Technical Problem. The Trust Gap Is a Human One.
On July 3, Cleverbridge announced it had joined Visa’s Agentic Ready program. The company adopted the Trusted Agent Protocol to test AI-initiated transactions. The Visa Trusted Agent Protocol ensures that AI-initiated transactions carry the same authorization signals as human-initiated ones. That is the right infrastructure. However, it solves the technical trust problem at the payment rail level. It does not solve the consumer trust problem at the moment of delegation. The protocol verifies the transaction technically. What it cannot do is make a consumer feel confident about a purchase they did not explicitly expect to authorize. Technical trust and experiential trust are not the same thing. Retailers who invest only in the payment infrastructure will find that transactions still fail at the moment the consumer hesitates.
The 24 Percent Who Are Already Comfortable Are the Segment That Matters Now
The 24 percent figure is not only a limitation. It is also a segment. One in four consumers is already comfortable with AI-powered payments. That segment is disproportionately high-value, high-frequency, and early in the adoption curve. As I described in the context of the Salesforce shopper agent analysis, the retailers who ran their own agents during the 2025 holiday season grew sales 59 percent faster. The consumers driving that growth are concentrated precisely in the segment already comfortable with AI-initiated transactions. Retailers who build for the 24 percent now will build operational experience, trust architecture, and customer relationships. They will capture the market as the percentage grows. Retailers who wait for the number to reach 50 will enter the market after the advantage has compounded elsewhere.
How Most Retailers Will Misread Their Own Performance Data
The 47 percent versus 24 percent split creates a predictable measurement trap for retailers deploying AI shopping assistants in 2026. Most deployments optimize for discovery metrics: session length, product views, add-to-cart rate. Those metrics will look strong because the market already accepts the 47 percent use case: consumers using the agent for research and completing the purchase themselves. The numbers will be encouraging. The problem is that those numbers do not capture whether the agent is actually closing transactions autonomously. That is the capability that produces the 59 percent growth differential. A retailer can run an AI shopping assistant for an entire quarter and report strong engagement metrics. The autonomous transaction completion rate can still be zero. The metrics they track are not the ones that reveal whether they are capturing the revenue opportunity.
By contrast, the KPIs that reveal whether an AI shopping assistant is building the trust layer that drives autonomous transactions are different. They include agent-to-human handoff rate at the payment step and consumer return rate after an agent-completed transaction versus a human-completed one. They also include the percentage of consumers who expand their authorization scope after their first agent-completed purchase. Retailers who measure these will identify exactly where in the journey the trust gap surfaces in their own customer base. Those who measure only traditional conversion will optimize for discovery and never understand why the checkout conversion rate for agent sessions is indistinguishable from their baseline.
Three Decisions That Determine Whether Consumers Let the Agent Close
Scope Transparency Before Scope Expansion
Consumer trust in AI-initiated transactions builds incrementally. It builds fastest when the agent is explicit about what it is doing at every step. Each interaction where the agent does exactly what the consumer expected builds the trust reserve. Over time, that reserve allows the agent to complete a transaction without human confirmation. Retailers who design their agent interactions to be transparent about scope and predictable in behavior build that reserve deliberately. Retailers who optimize for conversion speed will encounter consumer resistance at the payment step and interpret it as a technology limitation. It is not. It is a trust architecture limitation, and it has a different solution.
The Human Override Path That Makes Delegation Feel Safe
Counterintuitively, the feature that most enables consumer trust in agentic payments is a clear, accessible human override. As I described in the context of the agentic commerce infrastructure analysis, the agent layer does not replace the human layer. It specializes it. Consumers are significantly more willing to delegate to an agent when they know exactly how to take back control. The retailers who design conspicuous, friction-free override mechanisms into their agent experience will see higher autonomous transaction rates. Those who bury the human escalation path to reduce agent abandonment will not. The override is not a failure mode in the deployment. It is the trust architecture that makes the autonomous path viable for the consumer in the first place.
Starting Narrow and Expanding Permission Quarter by Quarter
The fastest path from 24 percent consumer comfort to a higher number is not a broader AI payment capability deployed all at once. It is a narrower one that expands over time. Retailers who start with a highly constrained agent authorization will see a key pattern emerge. Consumers who complete one authorized transaction are significantly more likely to expand the scope for the next one. A constrained authorization might allow the agent to complete purchases only for pre-approved categories, within a pre-set spending limit, using a pre-confirmed payment method. That incremental permission model mirrors how consumers built trust with stored payment credentials a decade ago. The retailers who design for incremental expansion will compound the 24 percent into a larger segment quarter by quarter. Those who launch with full autonomous payment capability and no permission scaffolding will hit the trust ceiling immediately and stay there.
What This Means for LatAm Retailers
The consumer trust gap in AI payments has a specific shape in Latin America that differs from the U.S. pattern. In markets where WhatsApp already mediates significant purchase activity, as I described in the context of the Meta Business Agent analysis, consumers have already built a form of transactional trust through informal conversational commerce. A shopper who regularly negotiates price through a WhatsApp conversation with a store associate has already accepted agent-mediated commerce in practice. That is a different trust baseline than a U.S. consumer who has only used AI for price comparison on a desktop browser.
For LatAm retailers deploying conversational commerce through WhatsApp, the trust gap between AI discovery and AI payment may be narrower than the global data suggests. However, the payment authorization infrastructure is also less mature in many LatAm markets. Consequently, the opportunity is to build consumer trust in AI-assisted transactions now, before the infrastructure fully matures. When the payment rails are ready, the consumer relationship and trust architecture should already be in place.
The Two Retailer Profiles That Will Define the 2026 Holiday Season
The Retailers Who Measure Discovery and Miss Revenue
Some retailers will enter Q4 2026 with an AI shopping assistant that performs well on discovery metrics. They will have an engaged user base that uses the agent to find products. Their conversion rate at the payment step will look identical to their non-agent baseline. Those retailers will conclude that AI assistants are a strong top-of-funnel tool. They will be correct. But they will also have missed the revenue opportunity entirely. Their agents will have generated excellent discovery engagement and handed every transaction back to a human at checkout. That is an expensive way to run a product recommendation engine. The gap between their results and those of retailers who built the trust architecture will appear in the Q4 revenue data.
The Retailers Who Build the Trust Architecture First
Other retailers will enter Q4 2026 with an AI shopping assistant that has been building consumer trust incrementally since Q2. Their agents will have a defined authorization model and a clear human override path. The permission scope will have expanded as consumers complete transactions and return for more. Those retailers will have a segment, small but growing, of consumers who let the agent complete the purchase. By Q4, that segment will represent a disproportionate share of their high-value transaction volume. The trust architecture they built in Q2 and Q3 will compound their advantage into 2027. Their competitors will still be trying to understand why their engagement metrics looked strong but their agent revenue did not.
The 24 percent who are comfortable with AI payments today are not a limitation. They are the segment that determines which retailers will own agentic commerce when the number grows. Building for them now is not optimistic. It is the only retail AI investment with a compounding return.
If you are evaluating your AI shopping assistant architecture or your consumer trust strategy for agentic commerce, connect with me here or reach me on LinkedIn. I am happy to walk through the framework we use across the U.S. and Latin America.
Adriana Rivas is a retail technology executive and AI strategist, and the founder of a U.S.-based hardware company specializing in self-service kiosks, POS systems, electronic shelf labels, and digital signage deployed across the United States and Latin America. She is the award-winning author of How to Implement Self-Service Without Failing (Amazon #1 Hot New Release, Silver Nonfiction Book Award 2025) and recipient of the Gold Stevie® Award, Thought Leader of the Year 2026. She is also recognized by Thinkers360 as a Top 10 Thought Leader in Retail and a Certified Master Expert in Retail.