AI agents are deciding who captures the value. Shoppers, workers and voters should have a say

The article argues that control over agentic AI shopping, pricing and workplace systems will determine who captures the value agents create, while opaque rankings and prices make that division difficult to scrutinize. It cites a Ninth Circuit ruling vacating an order blocking Perplexity’s Comet from Amazon, a grocery test in which identical eggs cost $3.99–$4.79 for different shoppers, and Oxford researchers’ analysis that Uber drivers’ inflation-adjusted hourly earnings fell from over £22 to just over £19 after a 2023 pricing-algorithm change; Uber disputed the figures. The authors advocate stronger competition, disclosure and loyalty rules for agents, alongside existing measures including New York’s algorithmic-pricing disclosure law and the EU Platform Work Directive.

Analysis

The key exposure is not AI productivity but who owns product discovery and the customer signal. If shopping agents become a meaningful entry point, they can weaken retailers’ ability to monetize sponsored placement and keep purchase data inside their own ecosystems. That is a conditional, multi-year risk to Amazon’s retail-media economics—not an immediate read-through from a preliminary ruling in one case. Retailers may respond by restricting access, charging for integrations, or steering agents toward controlled interfaces; those responses could preserve economics but raise friction for shoppers and invite further scrutiny.

Near term, the legal headlines are noisy: the Ninth Circuit’s ruling was preliminary, the underlying case continues, and the Instacart test was limited and discontinued. The Uber research raises a more direct regulatory risk because algorithmic allocation can shift value between workers, customers, and the platform, but Uber disputes the findings; do not treat the reported take-rate change as established consolidated financial impact. Over 1–3 months, watch court orders, retailer-agent integrations, and any measurable change in referral traffic or ad monetization. Over 6–18 months, disclosure, data-portability, and loyalty rules could increase compliance costs and constrain opaque optimization across platforms.

Contrarian view: agents may not disintermediate Amazon so much as become another paid acquisition channel, with retailers retaining control over inventory, fulfillment, and access. The Google Android action could widen assistant distribution, but does not establish that Google captures commerce economics. The evidence supports monitoring, not a broad short on the named platforms.

Key Decisions for Investors

• No immediate directional trade from the article alone. Avoid shorting AMZN on the ruling: it does not resolve the case or demonstrate scaled agent-led shopping.
• Put AMZN retail-media exposure on watch. Reassess only if filings or company commentary show material agent-driven referral volume, weaker sponsored-ad monetization, or new access restrictions; those are the missing data points that would make the disintermediation thesis investable.
• Treat CART as a regulatory/headline watch, not a thesis-level short: the cited pricing test was limited and ended. Revisit if regulators establish a durable enforcement standard or Instacart discloses a material product or pricing change.
• For UBER, monitor worker-pay transparency rules and litigation for evidence of take-rate pressure or higher operating costs; the study’s findings are disputed. Falsifiers include no regulatory follow-through and stable company-reported economics.
• Track Google’s implementation obligations and rival-assistant access as a 6–18 month distribution catalyst, but require evidence of user adoption or monetization before assigning earnings upside.