46% of Consumers Rate AI Chatbot Experiences Positively: Five9

Less than half of consumers — 46% — rated their most recent experience with AI and chatbot channels positively, according to a Five9 survey released last week. Live phone, email and in-person support performed far better. About two-thirds of respondents rated their most recent experience with those channels positively.

The gap has a clear trigger. When consumers abandon AI for a human, 46% said they did so because the bot repeatedly misunderstood them, Five9 found. The company surveyed 3,000 consumers and 600 business decision-makers in the United States, United Kingdom and Germany.

Why do bots miss the point?

Few things push a customer to demand a live agent faster than a bot that misunderstands them. CX (customer experience) leaders implementing AI for self-service should first understand why that failure happens, then take steps to remove the friction.

"A bot can understand every word a customer says and still completely miss the point," Brad Cleveland, CX expert and author, told sister publication CX Dive.

The mechanism is simple. Customers do not speak in neat scripts. They conflate multiple issues into a single statement or use sarcasm and irony to convey the opposite intent. Context can be missing. The knowledge base (the stored information a bot draws on for answers) can be outdated.

"This is the Achilles heel of AI and customer service across the board. It's a huge need to make sure the underlying knowledge is sound," Cleveland said.

What should brands fix first?

Cleveland's recommendations for CX leaders are concrete:

• Keep the knowledge bases that feed the bot current and accurate.
• Test with real customers and real conversations, including difficult ones.
• Measure failure rates and escalation rates over time.
• Make escalation to a human agent easy.

Testing deserves special attention, Cleveland said. Polite, perfect test customers are not enough. "Test with frustrated customers, fast talkers, a person with dogs barking in the background or at a busy airport where they're trying to get something done while waiting for a flight," he said.

Does a human escape hatch help AI trust?

The data suggests it does. Five9 found 55% of consumers trust AI to resolve their issue when a human option is available, versus 26% when there is none. The availability of a human backup nearly doubles consumer trust in the automated channel.

That finding challenges a common industry metric: containment rate, the share of interactions handled without a human agent. High containment looks like success only if customers stay by choice.

"Don't celebrate containment if customers feel trapped," Cleveland said. "Track your failure rate and the escalation rate and how that's improving. These tools are always iterative — they should be anyway. So, make sure you're making improvements and then seeing what's happening."

The message for marketers and CX teams is direct. AI self-service will earn positive ratings only when the underlying knowledge is sound, testing mirrors messy real life, and customers can always reach a person. Brands that measure failure and escalation honestly, rather than containment alone, are the ones positioned to close the 20-point satisfaction gap between bots and human channels.