In a Gartner survey of 3,566 customers, run in February and March 2026, 49% said they’d have been willing to use a chatbot if the company offered one. Only 7% used a chatbot or digital assistant in their latest service interaction. Gartner analyst Eric Keller points to past experience, and the same survey found just 27% said they’d be willing to try a chatbot again after a bad experience with one. He calls it a leaky bucket, where one failed conversation can put a customer off even as the bot improves.
Why It Matters
Service leaders are counting on chatbots anyway. In a separate Gartner survey of 199 of them, published in August 2026, respondents predicted GenAI chatbots will become the most valuable service channel two years from now. Keller’s advice is to allow for the gap between willingness and use when estimating adoption, because saying yes is only potential demand. That makes the adoption forecast the number to check first.
The Decision
So the question for your team is how wide the bot starts. Launching on every issue type gives the forecast its best early chance, at the price of more failures while the bot is least tested. Launching only on issues it already resolves reliably protects first impressions, and keeps more of the work with your people for longer. Keller’s call is reliability over reach, though that’s his judgment, not data.
What To Do This Week
- Ask whoever built the chatbot business case whether its adoption number came from what customers said they’d do or from what they did.
- Ask for the issue types the bot will handle at launch, and how often each was fully solved in testing, judged by a person. Gartner sets no pass mark, so have the team set one before seeing the numbers.
- Have someone ask your bot, or the pilot, something it can’t answer, and time how long until they reach a person who already knows the problem.
What Not To Do
Don’t judge the launch on containment, the share of conversations that never reach a person. Keller says chatbots should connect customers to human support, not become “containment traps”. A customer who gives up counts as contained too. And don’t read the 27% as customers rejecting chatbots. Nearly half were willing to use one if offered. The leash is short for the bot that fails them.
Signal Boost
Gartner Survey Finds Only 27% of Customers Would Try a Chatbot Again After a Negative Experience, Gartner’s Q&A with Eric Keller, September 2, 2026. Forward his last answer, on when to widen what the bot handles.