The tech industry is hot for shopping bots, but they're years away

Mastercard cited projections that agentic commerce could generate $1 trillion in U.S. consumer-retail revenue by 2030 and intermediate more than $15 trillion of B2B spending by 2028, but industry participants say fully autonomous purchasing is not yet functional. Amazon's blocking of Meta and Perplexity shopping bots, alongside unresolved identity, authorization, liability, website-access and transaction-infrastructure issues, underscores substantial adoption barriers. Near-term commercialization is likely to remain limited to agent-assisted shopping, with true autonomous commerce potentially requiring roughly five years and broad merchant, payment-network and regulatory alignment.

Analysis

The investable implication is not a near-term payment-volume windfall but a longer-duration fight over transaction ownership. AMZN’s closed ecosystem is economically rational: autonomous comparison layers would weaken sponsored-product yield, private-label discovery, and seller advertising pricing before they materially threaten merchandise revenue. Over the next 6-18 months, merchants are more likely to deploy controlled, on-site AI assistants than permit third-party agents to commoditize checkout; this favors platforms with proprietary consumer intent and fulfillment data over open-web intermediaries.

MA and V are option-like beneficiaries only once authorization, liability allocation, and identity standards are interoperable. Until then, agent-assisted purchases may merely shift existing card-not-present volume without raising payment penetration or net take rate, while fraud, chargeback, and authentication spending rise. The nearer-term winner could be cybersecurity/identity infrastructure—such as OKTA, PANW and CRWD—if enterprise procurement begins treating delegated credentials and auditable agent permissions as a required control layer; however, no broad revenue inflection should be underwritten before issuer pilots demonstrate loss rates and merchant acceptance.

Consensus risks confusing AI-assisted conversion improvement with autonomous commerce. AI shopping features can improve discovery and attach rates within months, but delegation of discretionary spending is constrained by consumer trust, merchant resistance to price transparency, and regulatory scrutiny of algorithmic steering. For AMZN, the relevant falsifier is evidence that large merchants or regulators force meaningful agent/API access, or a sustained deterioration in advertising yield; for MA, look for disclosed tokenized credential volumes, authorization approval rates, and fraud-loss trends rather than ecosystem TAM claims.

Key Decisions for Investors

• Maintain/establish a 6-12 month long AMZN versus short META pair: AMZN retains monetizable commerce intent and can gate third-party automation, while META has weaker direct control of transaction completion. Target 10-15% relative upside; stop if Meta secures durable, scaled checkout integrations with major marketplaces or AMZN advertising growth decelerates materially.
• Do not chase MA on agentic-commerce narratives over the next 1-3 months. Use any AI-driven multiple expansion to trim tactical exposure unless management quantifies incremental tokenized-agent payment volume and stable fraud economics; a better entry requires evidence that new flows are additive rather than cannibalistic card-not-present volume.
• Create a 6-month watch basket of OKTA, PANW and CRWD for delegated-identity demand, but initiate only after at least two major issuers, networks, or commerce platforms disclose production-grade agent credential standards. Risk is that identity functionality is absorbed by Visa/Mastercard, hyperscalers, or platform-native stacks.
• Monitor AMZN’s seller-service and advertising metrics each quarter. A decline in ad yield or seller monetization concurrent with externally mediated shopping traffic would be the first tradable signal to reduce the AMZN long and revisit a long MA/V versus short AMZN hedge.