Walmart is adding AI shopping features that can turn handwritten lists and photos into product recommendations and shopping carts, while Amazon is making the case for AI agents to handle business purchasing within company rules. The shift is from helping people choose towards asking software to take practical steps towards a transaction.
Watch Desk analysis
What happened
Walmart’s Sparky can build a cart from a photo of a handwritten list or a digital list. Shoppers can also share an image of a problem, such as a stained tablecloth, for product suggestions, or ask for recipes tailored to a budget and dietary needs before adding ingredients to a cart. Its Shop to Light feature can locate an item on a store map and flash the relevant shelf label. Walmart said voice capabilities were expected chainwide by the end of the year; that is a planned launch, not a feature the article says is already available everywhere.
Amazon’s business pitch centres on how much authority to give agents purchasing supplies. At an Amazon Business conference, executives highlighted spending limits, company policies, human approval above set limits and records of agents’ actions. Separately, an AWS executive said the marketplace’s agent mode doubled conversion from around 3% to 6%. That figure applies to AWS Marketplace, not business purchasing generally. The article reports a forecast of agentic payments at scale in 2027, not evidence that routine autonomous payments are already widespread.
Why it matters
Shopping assistance and delegated purchasing are not quite the same thing. A system that suggests a product helps with a decision; one that assembles a cart or spends company money needs to respect budgets, policies and approval rules. The further agents move towards transactions, the more important it becomes to know what they can do and how their actions can be checked.
Walmart’s features also connect digital assistance to its physical shops, with shelf labels and store maps part of the experience. Amazon’s business approach puts controls and audit trails closer to the centre of the pitch. Together, these are concrete examples of AI being fitted into existing retail and purchasing systems, rather than left to admire itself in a chat window.
Our read
The most useful test is not whether an agent can make a plausible recommendation. It is whether it can help someone complete a real task without losing track of cost, consent or accountability. Walmart has outlined a mix of shopping and in-store features; Amazon’s business case makes clear that delegated purchasing needs guardrails. The conversion figure is interesting, but one marketplace result should not be mistaken for a verdict on agent-led commerce.
What to watch
- Whether Walmart’s planned voice features launch chainwide by year-end.
- How shoppers can review or change carts and recommendations produced by Sparky.
- Whether Amazon’s purchasing controls are available in practice, and how approval limits and action records work.
- Whether the AWS Marketplace conversion result is matched by further evidence across other purchasing tasks.
Discussion spark: How much purchasing authority should an AI agent get before a person has to approve the transaction?
Sources and evidence
- Source update (10 November 2026, 09:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.