CX and digital leaders can design an AI chatbot handoff that keeps AI on product research and forces order problems to a person with authority to fix them. Baymard’s survey of 2,273 U.S. online shoppers, reported by CMOtech and set out in the institute’s AI Engagement study, shows people welcome AI for product research and reverse that preference once an order goes wrong. Retail groups have been adding chatbots and virtual assistants to handle large volumes of queries, and the research suggests that approach may be less effective in customer care, particularly during periods when delivery mistakes and return requests rise.
Updated November 6, 2026
Key Takeaways
- 63% prefer an AI chatbot to compare product features.
- 69% want a human for damaged or incorrect items.
- 67% want a person for late or missing orders.
- A failed handoff does not save money, Holst said.
When should the AI chatbot stay on the task?
Shoppers accept AI as a research assistant, not as the owner of a broken order. 63% would rather use an AI chatbot to compare product features, 59% prefer AI for checking stock availability, and 56% favor it for personalized recommendations.
Baymard’s guidance is to embed chatbots in discovery workflows and align them with shopper comfort. Analysis: keep AI while the shopper is choosing; move off AI once the order must be corrected.
| Shopper task | Preferred help | Share |
|---|---|---|
| Compare product features | AI chatbot | 63% |
| Check stock availability | AI chatbot | 59% |
| Personalized recommendations | AI chatbot | 56% |
| Damaged or incorrect items | Human representative | 69% |
| Late or missing orders | Human representative | 67% |
| Shipping or billing mistakes | Human representative | 66% |
When should you force a human with authority?
The preference flips once money, delivery, or a return is in play. 69% prefer a human for damaged or incorrect items, 67% want a person for late or missing orders, and 66% prefer human support for shipping or billing mistakes.
Shoppers see AI as a research assistant, not a problem-solver. When a package is late or the wrong item shows up, people want someone with the authority to fix it. A chatbot that can’t hand over to a human at that point doesn’t save money. It turns a fixable problem into a lost customer.
That is Christian Holst, co-founder of Baymard, as quoted by CMOtech.
How should the AI chatbot handoff work?
Treat the survey split as the routing rule. Holst said retailers getting this right put AI in product discovery and make the human route obvious when an order goes wrong.
- Leave feature comparison, stock checks, and personalized recommendations with the AI chatbot.
- Move damaged or incorrect items, late or missing orders, and shipping or billing mistakes to a human representative.
- Hand that case to someone with the authority to fix it.
- Keep the path visible. Do not hide the ‘talk to a person’ option behind bots and prompts.
- If live support is unavailable, provide a transparent self-service option.
- Do not trap customers in a chatbot loop during the busiest weeks of the year.
- Treat customer-care automation as potentially less effective in periods when delivery mistakes and return requests rise.
For enterprise CX, that list is the routing policy: a visible path to a person who can fix the issue, not a faster bot.
How do you weigh a failed escalation?
Holst’s cost line is the measure on offer: a chatbot that cannot hand over does not save money and turns a fixable problem into a lost customer. He also said retailers that trap customers in a chatbot loop during the busiest weeks of the year are putting their most valuable relationships at risk. Baymard says forcing shoppers into automated bot loops when an order goes wrong can introduce severe friction, escalating frustration at the exact moment customer retention is most vulnerable.
Analysis: a completed bot thread on an order error is not evidence of savings. The loss the study describes is the customer. Check the conditions Baymard already sets: a prominent path to a human with authority to fix the issue, or transparent self-service when live support is unavailable. For wider context, see whether service teams are seeing the AI returns they expected and personal AI agents already contacting service teams.
Frequently asked questions
When do shoppers prefer an AI chatbot over a person?
In Baymard’s survey of 2,273 U.S. online shoppers, 63% prefer an AI chatbot to compare features, 59% to check stock, and 56% for personalized recommendations.
Which order problems should reach a human?
Baymard found 69% want a human for damaged or incorrect items, 67% for late or missing orders, and 66% for shipping or billing mistakes.
What does a failed chatbot handoff cost?
Christian Holst, co-founder of Baymard, said a chatbot that cannot hand over does not save money and turns a fixable problem into a lost customer.
What if live support is unavailable?
Baymard says retailers should keep escalation to a human prominent, or provide transparent self-service options when live support is not available.
