New Five9 research shows consumers rate AI chatbots well below live phone, email, and in-store support. The retailers closing that gap are not buying better bots. They are hiring the people who make bots understand customers.
Retailers have spent the last two years putting AI at the front door of customer service. Order status, returns, sizing questions, and store hours increasingly start with a chatbot. According to a new Five9 consumer survey covered by Retail Dive, customers are losing patience with those bots, and the reason is not that the technology is missing. It is that the bots keep misunderstanding them.
For retail, eCommerce, and omnichannel leaders heading into peak season, that finding matters. Holiday volume puts more customer conversations through AI than any other time of year. If the bot cannot understand a shopper asking where a gift is, every one of those conversations becomes a live agent call, a lost sale, or a frustrated customer who shops somewhere else next time.
Five9 surveyed 3,000 consumers and 600 business decision-makers across the United States, the United Kingdom, and Germany for its 2026 consumer CX report. Fewer than half of consumers rated their latest AI or chatbot interaction positively. Traditional channels, including live phone, email, and in-person support, each earned positive ratings from roughly two in three respondents.
When consumers gave up on a bot and asked for a person, the single biggest reason was being misunderstood over and over. And the research points to an outcome many retailers might not expect: shoppers were more than twice as likely to trust AI when they knew a human was available as a backup.
CX expert and author Brad Cleveland, speaking with Retail Dive's sister publication CX Dive, explained that a bot can process every word a customer says and still miss what the customer actually means. Real shoppers do not talk in tidy scripts. They bundle a late delivery, a wrong size, and a promo code into one message. They use sarcasm. They leave out context the bot needs. And the knowledge the bot draws on is often out of date, which Cleveland described as the core weakness of AI customer service.
His recommendations are practical: keep knowledge bases accurate and current, test with real and frustrated customers in noisy, distracted settings rather than polite test scripts, measure failure and escalation rates, and make it easy to reach a person. His sharpest line was aimed at the most common chatbot metric: "Don't celebrate containment if customers feel trapped."
Every one of those fixes needs an owner. That is where most retail organizations come up short.
| Where bots fail | What the customer experiences | Who has to own it |
|---|---|---|
| Outdated knowledge | Wrong answers on policies, stock, or orders | Knowledge and content management |
| Scripted conversation design | The bot misses multi-part or messy questions | Conversation design and UX |
| Disconnected systems | The bot cannot see the order, the store, or the return | Retail and digital technology |
| Containment-only metrics | Shoppers feel trapped with no path to a person | Customer experience leadership |
| No failure tracking | The same problems repeat for months | Analytics and data science |
Containment rate is easy to report and easy to celebrate. It is also easy to game: a shopper who gives up counts as contained. Five9's trust numbers suggest the opposite strategy works better. A visible path to a human makes customers more willing to try AI in the first place.
That shift needs a leader with the authority to change how success is defined across digital, stores, and the contact center. Our post on hiring a VP or Director of Omnichannel Customer Experience covers the profile. In interviews, ask candidates which AI metric they stopped reporting and why. The strongest ones will have a specific answer.
A chatbot is only as accurate as what it reads. Return windows change for the holidays. Promotions start and end. Products sell out. If the knowledge base and product data lag behind the business, the bot confidently gives wrong answers, and customers learn not to trust it.
In retail and eCommerce, this work usually sits closest to site operations and digital content teams, the same people who keep product detail pages accurate. Our guide to hiring a Site Operations Manager and our DTC Site Management & Marketing practice cover the profile. Look for candidates who have owned a content governance process, not just published content.
Cleveland's advice to test with frustrated, fast-talking, distracted customers is a design brief. Someone has to map how shoppers actually ask questions, write flows that handle several issues at once, and decide exactly when the bot should hand off to a person.
Conversation design is a fast-growing specialty inside our Creative, UX & UI practice. Strong candidates can show you transcripts of real conversations they redesigned and the before-and-after numbers. Generic UX portfolios with no conversational work are a weaker fit.
Cleveland's call to track failure and escalation rates, and to keep improving, is analytics work. Retailers need to know which intents fail most, which products or policies generate the most confused conversations, and whether changes actually improved outcomes.
That belongs in our Analytics & Data Science practice. Look for analysts who have built conversation-level reporting and tied it to business results such as repeat purchase, refund rates, or cost per contact.
Many bot failures are really integration failures. A chatbot that cannot see order status, store inventory, or return eligibility will misunderstand by default, because it simply does not have the answer. Connecting AI to order management, inventory, and returns systems is core retail technology work.
Our guides to hiring for retail technology roles and hiring a Head of AI Transformation explain how these roles fit together. The gap between adopting AI and scaling it, as we noted in that second post, is usually an ownership problem.
A chatbot does not fail because it is AI. It fails because no one owns what it knows, how it talks, and when it gets out of the way. Those are hiring decisions.
Peak season, and the returns wave that follows it in January, will send more customer conversations through AI than any other stretch of the year. A few steps to take now:
Not very. Five9's 2026 consumer CX research, which surveyed 3,000 consumers and 600 business decision-makers in the US, UK, and Germany, found that only 46% of consumers rated their most recent AI or chatbot experience positively. Live phone, email, and in-person support each earned positive ratings from roughly two-thirds of respondents.
The most common reason is being misunderstood. Among consumers who left an AI channel for a human, 46% said the bot repeatedly misunderstood them, according to Five9. Real customers combine several issues in one message, use sarcasm, leave out context, and ask about things the bot's knowledge base has not been updated to cover.
Yes. Five9 found that 55% of consumers trust AI to resolve their issue when a human option is available, compared with 26% when there is no human option. An easy, visible path to a live agent tends to increase AI adoption rather than reduce it.
Five functions carry most of the work: a customer experience leader who measures resolution rather than containment, a knowledge and content owner who keeps product, policy, and order information current, conversation designers who write and test flows for real customers, analysts who track failure and escalation rates, and a retail technology lead who connects the bot to order, inventory, and returns systems.
Containment rate measures the share of conversations a bot handles without passing them to a human. It can be misleading because a customer who gives up, or who cannot find a way to reach a person, counts as contained even though the problem was never solved. Retailers get a clearer picture by pairing containment with resolution, failure, and escalation rates.
Sources: Retail Dive, "Why customers are fed up with AI chatbots," October 22, 2026 (first published by CX Dive); Five9 2026 Consumer CX Report.