The shopping we keep and the shopping we delegate
Pleasure, preferences and permission could shape which parts of shopping we keep and which we hand to agents. Trusted Convos, day two.
Over dinner at Trusted Conference in Phuket, a conversation with several teammates turned to a question businesses will increasingly have to answer: what will agents do, and what will humans want to keep doing?
It is part of the web iv hypothesis I have been developing over the past two years. I expect an internet shared by humans and agents, with different roles for each. The web iv clock tracks the evidence for that transition.
Commerce makes the question concrete. Some shopping is a pleasure. Some is a task we would happily give to someone else. Understanding that difference could tell us where human commerce ends and agentic commerce begins.
Pleasure could define the dividing line
Our working hypothesis is that pleasure will be one of the strongest influences on what people keep and what they delegate.
Someone might enjoy choosing clothes, comparing musical instruments or deciding where to take the family on holiday. For that same person, comparing health insurance might be an obligation. A different person could enjoy investigating those options and have little interest in fashion. The boundary belongs to the individual.
Consumer research distinguishes practical shopping value from enjoyment. Both can be present in the same experience. More directly, in four experiments by Simona Botti and Ann McGill, people were more satisfied with an outcome they had chosen themselves than with the same outcome chosen for them when their goal was pleasure. That advantage was not observed when their goal was practical. Choosing can be part of the value.
There is relevant AI evidence too. Chiara Longoni and Luca Cian found that people tended to favour human recommendations for enjoyment-focused goals and AI recommendations for practical goals. The effect was linked to beliefs about what each recommender could do well. This tested recommendations, rather than permission to buy autonomously.
Enjoying the eventual purchase is different from enjoying the work of choosing it. In recent experiments using imagined purchases, AI delegation could improve reported well-being for experiential purchases by increasing anticipation. Someone can love holidays and gladly hand over the flight comparisons.
The research gives us reason to investigate the boundary, without establishing exactly where it falls. My hypothesis is that the parts of shopping we enjoy doing, or find personally fulfilling, will be harder to delegate than the parts we simply want done.
Comparison websites offer a useful behavioural clue about where people already seek assistance. Agents could take that help further, working to a person’s criteria and, where authorised, completing the purchase.
We will also delegate parts of the journey
The division could run through a single purchase. A person might enjoy choosing a product while having no interest in searching multiple retailers for the best terms.
An agent could compare a much wider set of options against their preferences, then produce a shortlist for a human decision. Once the product is chosen, it could monitor price, availability, delivery and returns against an agreed brief.
One purchase, shared roles
- Person Set the brief. Destination, preferences, budget and limits.
- Agent Compare options. Price, timing, availability and terms.
- Person, approval Choose and approve. Decide which option may be booked.
- Agent Make the booking. Book the approved option on the agreed terms.
- Person Enjoy the trip. Keep the experience that matters.
As I argued in the first Trusted Convos article, an agent needs to understand our preferences and values. Those can be stated explicitly and refined through feedback and observed choices. The cheapest option may fail the brief completely.
For businesses, this means understanding who is doing what at each stage. An experience that helps a person imagine a holiday has a different job from the information an agent needs to establish whether a booking meets their requirements.
Delegation needs a brief and boundaries
I do not expect people to hand over unrestricted authority to spend their money.
The instruction needs to cover what matters, the acceptable terms, the spending limit and the approval rule. Should the agent ask before buying, or complete the transaction when everything fits? If something falls outside the rules, it should return to the person.
The Agent Payments Protocol describes both direct approval and advance authorisation for constrained purchases, including a route back to human approval. These mechanisms give practical form to the distinction between researching on my behalf and spending on my behalf.
The permission changes the agent’s job
I also wonder whether sellers will offer new ways to reserve an opportunity. An agent could find a product at the right price, secure a temporary hold and give me an hour to confirm. That would require the seller to agree to hold the item and terms. It is a mechanism I would like to explore.
The important point is that delegation can have different degrees of authority. Keeping the final decision does not mean doing all the work that precedes it.
By 2030, the intermediary could own the purchase
By 2030, I expect most ecommerce to pass through third-party intermediaries, with customers able to complete purchases without visiting a retailer’s website.
Google, Meta and the companies behind ChatGPT and Claude could become rivals to Amazon for that relationship. A person could remain involved in choosing and approving while the platform coordinates the transaction.
Some of the infrastructure is already being developed. Google’s Universal Commerce Protocol provides a common language for commerce between consumer interfaces, businesses and payment providers. The retailer can remain the merchant of record while the customer shops through another interface.
That separation matters. A retailer can still make the sale while an intermediary becomes the place where the customer discovers, compares and buys.
The convenience comes with a concentration of power
This is the direction I expect. I am deeply uncomfortable with where it could take us.
A platform handling more of the journey could learn what we consider, reject, value and eventually buy. Combine that with its view of our work, searches and media consumption, and it could infer increasingly detailed things about our lives.
The concern extends to the economy. If a handful of businesses control access to customers, retailers become more dependent on the terms those platforms set. Choices about which products get presented, and on what basis, become commercially consequential.
We could end up keeping the parts of shopping we enjoy while handing more of commerce and customer knowledge to a few global companies.
Big tech becoming the intermediary for how we work, search, shop and consume media is a dreadful prospect. The convenience of delegating another task will be easy to see. The cumulative transfer of power deserves just as much attention.
Reviewed 04 November 2026