Talkdesk's 2026 AI Holiday Shopping Survey, fielded in September and released November 4, found that 89% of U.S. retailers are preparing AI shopping agents for this holiday season. The catch: only 15% of consumers would let one of those agents buy something using their payment details, even after reviewing the order first. Without that review step, trust falls to 10%. The gap between retailer ambition and consumer comfort is this quarter's real story, not the agents themselves.
What Did Talkdesk's Survey Actually Find?#
Talkdesk commissioned Centiment to survey 1,048 U.S. adults who plan to shop online this holiday season, plus 312 U.S. retail decision-makers at director level or above, all at companies with 50 or more employees already using AI. The field work ran in September 2026.
Nearly 9 in 10 retailers (89%) are building AI agents that shop on a customer's behalf, and 93% plan to deploy AI in customer service this season, 51% of those specifically for handling returns, refunds, or exchanges. On the shopper side, 74% plan to use AI assistants for product discovery and 49% say they would buy from an unfamiliar brand on an AI's recommendation.
At WebEpex, we read this one twice, because it lines up almost exactly with the Meta ads and WhatsApp automation builds we ship for D2C brands across the GCC, Europe, USA, Canada, and Indian markets heading into Black Friday and Cyber Monday this month. Retailers and their customers aren't disagreeing on whether AI belongs in the shopping journey. They're disagreeing on where it's allowed to stop being advisory and start moving money.
A separate, smaller Deloitte survey of 4,000 U.S. consumers, reported by Adweek, found only 16% trust AI agents to complete a purchase at all, and 66% worry about how their data gets used. Different survey, same conclusion.
Does This Change Anything for Your Business?#
If you run a service business with no shopping cart, this barely touches you. Skip to the disclosure point below, then move on.
If you're a D2C brand, a SaaS company billing through checkout, or anyone running Meta ads into a Shopify or WooCommerce store this quarter, it changes where you should spend automation effort. Shopify's own Q3 earnings already showed agentic commerce moving from theory to revenue line, but this survey is the other half of that story: shoppers want AI involved in deciding what to buy, not in authorizing the charge.
That split matters for budget. Spend on AI-driven discovery, like the kind behind ChatGPT's virtual try-on push for D2C brands, is still money well placed. Spend on a fully autonomous checkout bot, this season, for most catalogs, probably isn't yet.
The Honest Trade-Offs of AI Shopping Agents#
| What the survey measured | Retailers | Consumers |
|---|---|---|
| Preparing or using AI agents this season | 89% building shopping agents, 93% in customer service | 74% plan to use AI for product discovery |
| Comfortable with AI completing a purchase | Not asked this way | 15% with a review step, 10% without one |
| Want AI involvement disclosed | 37% don't fully disclose it | 77% want to know when AI is involved |
The upside is real. Returns and refunds (51% of retailers automating this) are a genuinely low-risk place for AI to operate, since the money already changed hands and the agent is reversing a decision, not making one.
The downside is real too, and most vendor pitches leave it out. Among retailers using AI, 31% said it took an incorrect action on a customer's account in the past 12 months, like issuing an improper refund. Of the 58% reporting any AI error, 29% lost a customer over it. That's not a rounding error.
How We're Handling This at WebEpex#
We build WhatsApp and chatbot ordering flows where the agent's job stops at the cart. It drafts the order, shows the total, and waits for an explicit confirmation typed back before anything touches a payment link. We don't let it auto-charge, even for a repeat customer we've served a dozen times, because the cost of one wrong charge is the relationship, not just a refund.
That pattern held up in a WhatsApp lead-qualification build we shipped for a real estate client this year: the bot scores the lead and books the call, then hands off to a human before any payment conversation starts. A comms stack we run elsewhere pairs IVR routing with WhatsApp messaging through role-based access, so no automated step moves a transaction forward without a person signing off somewhere in the chain.
My read, and this is an opinion, not a fact from the survey: fully autonomous checkout for AI agents is probably twelve to eighteen months out for mainstream D2C, not this holiday season. I'd bet on it. I could be wrong. Model capability keeps moving faster than I expect, but the trust math in this survey backs the slower timeline, and AI shopping chatbots were already losing sales before anyone tried handing them a credit card.
What I'd Tell a Client Asking About This Week#
Check your chatbot and WhatsApp flows for the one thing that's free to fix: a visible line telling the shopper AI is involved, right where it's involved. Seventy-seven percent want that and most businesses still skip it.
Don't bolt autonomous checkout onto your Meta ads funnel for Black Friday. Automate discovery, support, and post-purchase service instead; adoption there is already high and the failure mode is cheap. If you're not selling through a cart at all, skip this week. Do nothing. That's a legitimate answer too.
If you want someone outside your own head to look at where your ads and chat flow actually hand off to a human before money moves this season, send me what you're running through WebEpex and I'll tell you straight. No pitch attached.
