
Agentic AI Shopping: Why the Tech Giants' Conflict Should Interest Investors
This article was created with the help of artificial intelligence.
Key Takeaways
- McKinsey projects that agentic commerce could generate up to one trillion US dollars in revenue in the retail market alone by 2030, which would make the technology one of the largest AI growth segments.
- Since November 2025, OpenAI, Google, Amazon and other AI chatbot developers have been negotiating with major retail partners over liability rules for agent errors and the sharing of product data and chat histories.
- Banks, credit card issuers and insurers could lower customer acquisition costs through AI agents, but they also face increased customer switching and disintermediation by tech groups that control access to consumers.
- Visa is positioning its Intelligent Commerce Network as the backbone of agentic commerce and offers AI platforms tokenised credentials via so-called AI-ready Cards, while Stripe and OpenAI have developed the Agentic Commerce Protocol.
- In their study for the William & Mary Law Review, Professor Mark Bartholomew and Samuel Becher call for regulatory guardrails such as independent audits, algorithmic transparency labels, data portability and protections against self-preferencing by agent operators.
- According to Accenture, investment in AI is accelerating, but most companies have so far seen only limited measurable results, because the shift from traditional e-commerce to agent-based models requires new operational approaches.
Artificial intelligence that shops on its own – what sounds like science fiction is on the verge of a breakthrough. Agentic AI shopping refers to AI systems that make purchasing decisions autonomously for consumers: searching for products, comparing prices, negotiating with providers, opening or switching accounts and completing transactions – all without human intervention. This definition comes from an analysis by the law firm National Law Review dated 17 September 2026.
In the financial services sector, such agents could select credit cards, car loans, mortgages, insurance policies, bank accounts, term deposits and investment products. They could compare thousands of offers in seconds and execute transactions that consumers would otherwise avoid due to time and complexity constraints, according to an assessment by Consumer Finance Monitor on 17 September 2026.
A Trillion-Dollar Market by 2030
McKinsey & Co. estimated in June 2026 that the potential of agentic commerce could reach up to one trillion US dollars in orchestrated revenue in retail alone by 2030. This figure illustrates the enormous economic weight that AI shopping agents could develop – and explains why technology groups and payment providers are investing heavily in the infrastructure.
For investors, the question is who benefits. The answer is complex. While consumer-facing brands such as Perplexity or OpenAI dominate the headlines, analysts at Value Add Pulse (3 September 2026) see payment infrastructure as a more durable investment thesis. Companies such as Catena Labs and Basis Theory are building identity and virtual card infrastructure specifically for agent transactions, positioning themselves to benefit regardless of which consumer brand prevails.
Tech Giants in Negotiation Poker
OpenAI, Google, Amazon and other AI chatbot developers have been negotiating with major retail partners since November 2025 over key operational questions: How can costly agent errors be limited? Which product data and chat histories must be exchanged for agents to function? This information comes from a WIRED report dated 14 November 2025.
Some partnerships and protocols have already been established. In 2024, Perplexity launched the "Buy with Pro" feature with PayPal integration, providing access to more than 5,000 merchants. OpenAI, together with Stripe, announced an Agentic Commerce Protocol that allows users to complete purchases directly within the ChatGPT interface. Visa is positioning its Intelligent Commerce Network as the backbone of agentic commerce and offers AI platforms tokenised credentials via so-called "AI-ready Cards". This information was published by the MIT Initiative on the Digital Economy on 24 June 2026.
The negotiations also show that standards are still lacking. Bob Hedges, former Chief Data Officer at Visa and Digital Fellow at MIT, stated in June 2026 that while formal rules on chargebacks and fraud liability exist between banks, merchants and payment networks, there are "many negotiations about whether and how these rules should be applied in AI agent environments".
Disintermediation as a Risk for Banks and Insurers
For banks, lenders, insurers and credit card issuers, agentic AI presents a balancing act. On the one hand, the technology promises lower customer acquisition costs. On the other hand, it threatens increased customer switching between providers, intensified competition and disintermediation, according to Consumer Finance Monitor and NewsBreak on 17 September 2026.
The greatest risk: technology companies that control agents could become gatekeepers between financial institutions and consumers. They would determine which products consumers see, how products are compared and which providers receive any business at all. On 17 September 2026, the National Law Review warned that already dominant technology companies could, through control over agent interfaces, "dominate a large part of the financial marketplace".
For investors holding positions in European banks or insurers, this means that valuations must now also consider how well these institutions are prepared for a world in which customers no longer come to the bank directly, but an AI agent dictates the terms.
Regulatory Questions Without Answers
Professor Mark Bartholomew of the University at Buffalo and Samuel Becher, in a study for the forthcoming issue of the William & Mary Law Review ("The End of Shopping"), set out fundamental regulatory questions. These include consumer autonomy, data protection, competition, consumer protection, liability for AI errors, financial conflicts of interest when agents favour certain providers, and the adequacy of existing disclosure, fair lending, data protection, advertising and investor protection rules. Consumer Finance Monitor and NewsBreak published this summary on 17 September 2026.
Bartholomew suggests that policymakers should resolve these questions before agentic commerce becomes established. Possible guardrails include independent audits, "algorithmic nutrition labels", requirements for data portability, protection against self-preferencing, genuine consumer control including an "off switch", and regular reviews or sunset clauses for new regulation.
The MIT Initiative on the Digital Economy stated on 24 June 2026: "It is unclear how this will be regulated and how quickly people will trust that they can leave their shopping to AI agents."
Trust, Transparency and Operational Challenges
Not every startup with an "AI agent" label deserves trust. On 3 September 2026, Value Add Pulse pointed to the case of Phia – co-founded by Phoebe Gates – which was accused of cookie stuffing in its deal-finding technology. Cases like these remind investors that "AI agent branding does not exempt startups from the same trust and disclosure scrutiny to which any affiliate-adjacent business is subject".
A further operational problem: in agentic commerce, "both customers and businesses will find it difficult to define intent when purchases are made by AI agents", which leads to disputes over whether transactions were actually intended. Silicon Republic reported this on 8 April 2026.
Investment Flows In – Measurable Results Still Lacking
Accenture stated on 16 July 2026: "Investment in AI is accelerating, but for most companies measurable gains have materialised only slowly." Success requires companies to move beyond traditional e-commerce models and adopt new operational approaches.
For investors, this means caution regarding exaggerated expectations. Agentic commerce is real and its potential is enormous – but the path there remains rocky. Those investing should focus less on the loudest consumer brands and more on the invisible infrastructure: payment networks, identity protocols, tokenisation providers. And on companies that can deliver not only technology but also trust and regulatory compliance.
The conflict between tech giants, merchants and regulators is not a side issue – it will decide who controls the interface to the consumer in the next decade. And with it, trillions in revenue.
Sources
- The End of Shopping? Agentic AI and the Future of Consumer Financial Services Introduction
- AI Agents Want to Shop for You: The Future of Agentic Commerce
- The End of Shopping? Agentic AI and the Future of Consumer Financial Services Introduction [Podcast]
- AI shopping agent startups raising funding, 2026 | Value Add Pulse
- The End of Shopping? Agentic AI and the Future of Consumer Financial Services Introduction [Podcast]
- You Won’t Be Able to Offload Your Holiday Shopping to AI Agents Anytime Soon | WIRED
- Agentic commerce and purchase disputes: Did you mean to buy that?
- Agentic Commerce: Make Your Brand Unmissable | Accenture