Oli Yeates · 12 August 2026

Would You Let AI Do Your Shopping? 74% of People Say Yes

74% of consumers would trust an AI agent more than their best friend to buy for them, but only 9% would let it buy unsupervised. What Accenture's 25,590-person study means for brands.

Would You Let AI Do Your Shopping? 74% of People Say Yes

Here's a stat that stopped me mid-coffee this week.

74% of consumers say they'd trust a personal AI agent more than their best friend to make a purchase on their behalf.

That's from Accenture's 2026 Consumer Pulse, 25,590 people across 16 countries including the UK, surveyed in January. It's a proper sample, not a vendor poll of 400 people who already love the product.

More than your best mate. Let that sit for a second.

What the 74% actually means

Now, before we all rewrite our strategy decks, the honest version.

That 74% is about delegating routine tasks. Negotiating a deal. Chasing a complaint. Renewing a subscription. Reordering the stuff you always order. The boring, repetitive, mildly annoying end of shopping that nobody enjoys and everybody puts off.

When you move up the ladder, the numbers drop fast.

32% would let an agent choose the product, within limits they've set on budget and brand, with the customer still approving the payment. And only 9% would let an agent go all the way, choose and buy, no final tick.

So it isn't "three quarters of people have handed their wallet to a robot". It's "three quarters of people would happily never chase a refund again". Which, honestly, fair enough.

How much shopping would you hand to an AI agent?Share of consumers willing to delegate, by level of autonomyRoutine tasks: negotiate, complain, reorder74%Choose the product, I approve the payment32%Choose and buy, no approval needed9%Source: Accenture 2026 Consumer Pulse, 25,590 consumers across 16 countries including the UK.

But don't let the caveat talk you out of the point. 71% of those consumers expect generative AI to influence at least half their spending decisions over the next twelve months. PYMNTS has 48% of online shoppers already using AI to research their most recent purchase. Influence is running miles ahead of automation, and influence is the bit that decides whether your brand makes the shortlist.

Influence is running ahead of automationConsumers using or expecting AI in buying decisions, vs consumers ceding the purchase itselfINFLUENCEAUTOMATION48%Already use AIbefore they buy71%Expect AI to influence halftheir spending within 12 months9%Would let AI buywithout approvalSources: PYMNTS Intelligence and Visa Acceptance Solutions, 2026 (48%); Accenture 2026 Consumer Pulse (71%, 9%).

The bit that should worry you

Two numbers from the same research that I think matter more than the headline.

56% said they'd tell their AI agent which brands to consider. Good news if you're one of them.

And among behaviourally loyal customers, the ones who've bought from you again and again, 37% said they'd let an agent switch them to a different brand if it found a better fit.

Read that again. More than a third of your loyal customers would let a piece of software talk them out of you, on the basis of fit, price, availability or service record.

Your loyal customers are one comparison away56%would tell their AI agentwhich brands to consider37%of behaviourally loyal customers wouldlet an agent switch them to a rivalSource: Accenture 2026 Consumer Pulse, 25,590 consumers across 16 countries including the UK.

Habit has been doing a lot of quiet work in your retention numbers for years. People buy from you because they always have and switching is a faff. An agent removes the faff. It compares, it checks stock, it reads the returns policy, and it does it in two seconds without any of the emotional inertia that has been protecting your repeat rate.

What people won't hand over

There's a pattern in what people are willing to delegate, and it's a useful one.

They'll give away the high-effort, low-emotion stuff. Grocery restocking. Subscriptions. Post-purchase admin. 61% want an agent that can shop across several grocery retailers for them.

They keep hold of anything tied to identity or enjoyment. The hotel room. The jacket. The experience. Accenture found lifestyle and travel purchases drop off sharply as the level of autonomy rises.

What people will hand over, and what they'll keepWillingness to delegate a purchase to an AI agent, by type of purchaseHappy to delegateKeeping controlGrocery restockingSubscription renewalsReordering the essentialsChasing complaints and refundsNegotiating a better dealThe hotel roomClothingGiftsExperiencesAnything they'll be proud of owningHigh effort, low emotion goes first. Identity stays put.61% want an agent that shops across multiple grocery retailers.71% want an agent that books a whole trip: flights, hotel and activities.Source: Accenture 2026 Consumer Pulse.

So if you sell repeat-purchase, functional, comparable products, agentic commerce is coming for your category first and hardest. If you sell things people are proud of owning, you've got more time. Not unlimited time, but more.

Interesting one to finish that section on: 87% think AI will change the role of physical stores, and 31% reckon stores become more important as places to enjoy yourself. If the admin moves to an agent, the shop gets to be a nice day out again. There's a whole strategy in that for retail clients.

So what do you actually do about it

Make your product data readable by machines. Not just tidy. Readable. Pricing, availability, delivery times, returns policy, specifications, verified claims. If an agent can't parse it, you're not in the comparison. This is the least glamorous work in marketing and right now it's the highest leverage.

Assume you're being compared on things you don't currently publish. Fulfilment reliability. Real stock levels. Whether the policy actually says what the marketing says. Agents don't get charmed by a nice hero image.

Get on the consideration list before the comparison starts. 56% will name brands to their agent. That's a brand awareness job, an old-fashioned one, and it's suddenly load-bearing again. Being known is how you get into the shortlist the agent works from, which is the whole point of optimising for AI answers alongside classic search.

Give your loyal customers a reason that survives a comparison. If the only thing holding them is habit, an agent will unpick it. Loyalty schemes, service, warranty, the stuff a comparison engine can actually see and score.

Check your own data before you believe any of this. Including this post. Forrester's UK read is noticeably more sceptical than the global picture, with plenty of British consumers believing in agentic commerce but few trusting it to buy. We're a cautious market. Look at your own referral and assisted-conversion data rather than taking a global average as gospel.

The honest read

I don't think an army of shopping bots is going to descend on your ecommerce site next quarter. 9% is 9%.

But the direction is unmistakable, and the interesting risk isn't automation. It's that the deciding is quietly moving upstream, into a comparison you never see, run on data you may not have bothered to publish properly, before the customer ever lands on your site. We've already seen the first version of this arrive with product carousels inside ChatGPT.

The brands that win this aren't the ones with the flashiest AI announcement. They're the ones whose product feed is honest, complete and current, and whose name people already know well enough to say out loud.

Boring work. Big payoff. That's usually how it goes. 👊🏻

If you want a hand working out how your brand looks to an AI agent right now, that's rather literally what we do. Get in touch >>

Clicky™

Figures from Accenture's 2026 Consumer Pulse Research, 25,590 consumers across 16 countries including the UK, fielded 7 to 22 January 2026, via Accenture's "Talk to My AI Agent" report and trade coverage. The 48% figure is PYMNTS Intelligence with Visa Acceptance Solutions, surveyed across the US, Brazil and the UAE, so treat it as directional rather than a UK number.

Sources: