The headline adoption numbers can obscure an important boundary. Zendesk's 2026 data shows around 80% of consumers find AI bots helpful for resolving simple issues, but that acceptance does not extend to complex or emotionally charged cases, where customers still want a human. The line between automate and escalate sits at complexity and emotion, not at channel.
Trust is also conditional on design. 68% of consumers say they are more trusting of AI agents that exhibit human-like traits. Tone, empathy, and conversational naturalness aren't cosmetic; they materially change whether customers trust a bot's answer. Acceptance collapses when the AI either underperforms or feels robotic.
There's an internal dimension to the gap too. Salesforce finds service workers are among the least likely employees to use generative AI even as many fear being displaced by it: 48% worry about job loss if they don't learn the tools. That training and change-management gap caps the returns organisations can actually capture from their AI investments.
The practical implication is a deliberate division of labour: route simple, high-volume queries to AI, reserve human agents for complex and high-empathy interactions, and invest in agent-assist so those humans are faster and better-informed. Consumers already notice the difference: 70% see a clear gap between companies that use AI well in service and those that don't.