AI is playing a bigger role in how consumers research and buy, creating a growing trust question for brands: can their AI make accurate, relevant decisions with the customer information it has?
Amperity’s 2026 Consumer Priorities Report found that 78% of consumers are more likely to engage with personalized experiences when they trust how their data is being used, while 57.5% say invasive or “creepy” personalization makes them less likely to choose a brand.
Enterprises are also putting AI to work across customer data at greater scale. Amperity saw sessions with AmpAI, its AI interface for working with governed customer data, grow nearly fivefold year over year in Q2 FY27.
That growth offers an early signal that trusted customer data is becoming part of the infrastructure behind AI-powered interactions.
Derek Slager, co-founder and co-CEO of Amperity, discusses the trend in an interview with Retail Insider.
As AI becomes more involved in product discovery, recommendations and purchasing decisions, what does’trustworthy’ customer data look like from a retailer’s perspective?
Trustworthy customer data is data a retailer can use to make a dependable decision about a real person. It has to be accurate, current, permissioned, and resolved across the systems where that person interacts with the brand. Those are production requirements, not abstract data-quality goals.
That standard matters because product discovery is increasingly happening outside the retailer’s own channels. In our 2026 Consumer Priorities Report, 80% of U.S. consumers who use generative AI said they act on its recommendations, while fewer than 23% said they go directly to a brand they already know without consulting AI first.
When an AI system becomes the first interface a shopper uses, the context behind its answer is doing work once performed by an associate who knew the customer and understood the situation.
Most retailers have plenty of data. The problem is that ecommerce, store, loyalty, service, and marketing systems often hold different versions of the customer. Before AI can make a sound decision, the retailer needs to know which records belong to the same person, which signals are still current, and what the customer has permitted the brand to do. An AI system can execute every technical step correctly and still make the wrong decision if that context is wrong.
What are the biggest risks for retailers when AI systems make decisions using fragmented, outdated or conflicting customer data, and how can those errors affect the customer experience?
The biggest risk is that AI turns a data problem into a decision problem, then scales it. A fragmented profile once produced an irrelevant email. An agent can carry the same mistaken context into recommendations, offers, service interactions, and automated shopping experiences.
Consumers already notice the underlying failures. In Amperity’s 2026 State of Personalization in Retail report, 79% of U.S. shoppers said retailers get personalization wrong through irrelevant or mistimed outreach. With AI, the consequences can compound: recommending something the customer already bought, treating a loyal customer as new, contradicting a service resolution, or acting on a preference that changed yesterday.
The customer does not see a model error or an identity-resolution error. They see a brand that does not know them. As retailers delegate more decisions to AI, they need to raise the standard for the context behind those decisions and be able to reconstruct what the system knew, which rules it applied, and why it acted. Without that observability, teams cannot reliably diagnose or correct a bad outcome.
How can retailers use personalization to improve the shopping experience without crossing the line into what consumers perceive as invasive or ‘creepy’ marketing?
Useful personalization is grounded in the customer’s relationship with the brand and appropriate to the moment. Creepy personalization usually begins when a brand uses information in a context the customer did not expect, even if the underlying data is technically accurate.
The consumer data shows that this is not a choice between relevance and privacy.
72% of U.S. consumers said personalized experiences are important when choosing a brand, and 78% said they are more likely to engage when they trust how their data is used. At the same time, 57.5% said invasive personalization makes them less likely to choose a brand. The same customer can want a better experience and reject a use of data that feels unexplained or disproportionate.
Retailers should ask three questions before using a signal: Did the customer knowingly share it? Is this use consistent with their permissions and reasonable expectations? Does it create obvious value for them?
Remembering a preference, recognizing loyalty, or making a timely recommendation can feel like service. Demonstrating knowledge without a clear value exchange feels like surveillance. Accuracy is necessary, but appropriateness is the real boundary.
What specific steps should retailers take to give customers greater transparency, control and confidence over how their personal data is being used by AI?
Transparency and control have to show up in the customer experience, not just in the privacy notice. In our 2026 Consumer Priorities Report, 75% of U.S. consumers said they are more loyal to brands that are transparent about how they use customer data.
That starts with explaining, in plain language, what information is being used and why. If AI is materially influencing a recommendation, offer, or customer interaction, people should be able to understand its role.
The harder part is making those choices real across the organization. A preference expressed to customer service should not disappear when marketing or another business unit engages the same person. That requires a shared view of identity and permissions, along with controls over what data an AI system can access and what decisions it is authorized to make. For consequential actions, teams should also be able to trace the decision back to the data, rules, and permissions that informed it.
Customers should have practical ways to update preferences, manage communications, and opt out of specific uses. But a control is only meaningful if every system can recognize and enforce it. That is what turns transparency from a disclosure into something the customer can actually trust.
As retailers increasingly deploy AI across customer data and interactions, how do you expect the role of customer data management and governance to change over the next two to three years?
Customer data management will move from being treated primarily as marketing infrastructure to becoming core AI infrastructure.
Many organizations could tolerate fragmentation when individual teams used customer data for narrow, human-supervised tasks. AI changes the failure mode. Once models and agents use customer information to answer questions, recommend actions, build audiences, or interact directly with customers, inconsistencies in identity, freshness, permissions, and business definitions become operational risks.
Retailers will need a governed context layer that people and AI systems can rely on. That layer has to resolve identity, combine durable history with current signals, preserve lineage and permissions, and provide only the context relevant to the decision being made. More data is not automatically better context.
Governance will also become continuous. Policies cannot live only in review meetings and documents. Controls need to travel with the data, constrain the authority of the system, and be enforced when an AI system accesses information or takes action. The production standard will be straightforward: can the retailer reconstruct the decision, correct it when necessary, and show that it produced the intended customer and business outcome?
More from Retail Insider:
