The world of commerce is no stranger to new trends and technologies, but the agentic era is transforming the decision-making process for consumers more than ever before. Google has already launched a universal commerce protocol (UCP) enabling AI agents to transact across retailers, whereas Amazon has unveiled tools allowing agents to browse other retailer sites on a customer’s behalf.
A new Signals report from Mastercard has examined the impact of this kind of agentic commerce and what needs to happen before people and businesses are comfortable with allowing AI agents to take on a greater role in their shopping journey.
With agent-assisted consumer spending forecast to reach $3-5 trillion by 2030, as well as business-to-business (B2B) spending possibly reaching three times that figure, the scale of opportunity is vast. But, the challenge of building trust in AI agents to act on behalf of consumers and businesses is just as significant.
According to Mastercard’s “Encoding Trust: The Race for Intent, Consent and Control in the Agentic World” report, autonomy will not drive this growth alone. It will rely more on AI’s ability to simplify everyday choices without taking control away from the user.
Mastercard highlights the importance of trust, saying, “Agentic commerce will properly scale when we can prove that every action is bound by consent, is visible to the user, and reversible if something goes wrong. It’s about designing confidence into the flow.”
The limits of AI shopping autonomy
When conditions are set in advance, AI can help source the right options for consumers. AI agents may yet become the new gatekeepers between brands and consumers. However, Mastercard’s research highlighting a gap between consumers’ willingness to use AI for shopping and letting it make purchasing decisions for them. 85% of consumers said they are open to working with AI agents to find the best product or service for them, while 74% are open to handing over specific commerce tasks to AI. Only one in ten are currently willing to let an agent complete a purchase fully autonomously, however.
This creates a new challenge for brands as traditional marketing funnels could increasingly become a conversation between a brand and a machine, rather than a brand and consumer. AI agents could handle routine purchases within parameters such as a set budget, while consumers retain approval over exceptions such as unexpected price changes
Building trust in AI transactions
Trust remains a key consideration between an agent’s purchasing decision and a payment being processed. Mastercard’s report suggests a trust layer in which systems verify who is acting, what the user has authorised and whether the rules set by the user are being followed, while maintaining a record of the agent’s activity. Rather than giving AI agents unlimited access to payment credentials, agentic tokens could provide specific permissions, such as spending limits, allowing users to define what an agent can and can’t do.
These implications form new challenges for brands and marketers as AI agents increasingly influence what products consumers consider and buy. As a result, brands must ensure their products are visible and accurately represented by AI systems, with machine-readable product information potentially becoming a crucial part of digital marketing.
As Mastercard’s report finds, autonomy will not define the next era of commerce alone. Accurate permission models and a trusted, reliable infrastructure are needed, alongside controls that people can easily manage.
Marketing’s human playbook doesn’t work on AI agents
Marketers have spent decades refining persuasion tactics on human shoppers, but those same techniques don’t necessarily persuade AI agents, with some having zero effect, and others even working against the brand.
An OSF survey revealed most companies are unprepared for this, with the majority of 50 e-commerce executive respondents saying they have noticed conversion shifts attributed to AI agents and are, therefore, working on improvements to agent engagement. However, the belief that the same cues that persuade humans also influence AI agents remains widespread.
Research published by Harvard Business Review discovered traditional techniques, such as countdown timers and scarcity messages, produced inconsistent results when implemented with AI. Tactics aimed around human psychology, like price anchoring and FOMO, had little to no influence on AI agents compared to humans, suggesting marketers cannot use the same human persuasive techniques on AI shoppers.
To combat this, factors including competitive pricing and strong authentic ratings are key to influencing AI agents. Brands should also recognise that different AI models may respond differently to marketing cues,therefore considering which models drive the most traffic and purchases.
Understanding consumer prompts for AI agents may become a new form of customer research, helping brands understand what shoppers are asking AI to prioritise. As AI models evolve, continuous product information testing and adaptation may become necessary, rather than relying on fixed, traditional approaches.
Upcoming DMWF events will be discussing the questions around AI-driven discovery and changing customer journeys. Experts will be answering questions on how brands can remain visible in an ever-growing automated ecosystem across the marketing industry.
With AI agents making up a growing share of purchasing decisions, brands may need to rethink what it means to be persuasive and fine-tune their strategies. Accurate, relevant and useful information with less aggressive persuasion may be key to helping agents make the best decisions that lead to continued brand success.
