What customers think about AI in support

64% of customers wish companies would stop using AI in support — while deployment keeps rising. There’s no contradiction: companies optimise cost per contact, and dissatisfaction is measured in a different metric.

Key figures

• 64% of customers wish companies would stop using AI in support Industry consumer surveys, 2026
• Meanwhile AI deflects over 45% of contacts away from human agents Industry AI support benchmarks, 2026
• Only around 14% of contacts reach full self-service resolution Industry AI support benchmarks, 2026

How strong is the resistance?

Stronger than most companies deploying automation expect. 2026 surveys put 64% of customers wishing companies would stop using AI in service Industry consumer surveys, 2026 .

The methodological boundary matters: this is a stated attitude, not behaviour. Question wording strongly affects results in surveys like these — «would you rather speak to a human» and «are you satisfied with the response speed» produce different pictures from the same sample.

Why is deployment rising anyway?

Because companies and customers optimise different things. Deployment is driven by cost per contact, which is measured precisely and immediately. Customer attitude is measured indirectly and with a lag.

The data shows exactly that divergence: AI deflects over 45% of contacts away from human agents Industry AI support benchmarks, 2026 while only around 14% reach full resolution Industry AI support benchmarks, 2026 . The difference between those numbers is the part of the experience that forms the negative attitude.

What are customers actually objecting to?

Not the technology but the dead end. Surveys point at specific scenarios: being unable to reach a human, repeating the same question, answers that miss the point.

That distinction matters in practice. An automated answer that resolved the question in ten seconds provokes no resistance — it isn’t even noticed as automated. Resistance comes from a bot you can’t get out of.

What follows for deployment?

One condition that changes the whole picture: an explicit and fast route to a human on request. The data doesn’t support «don’t deploy AI»; it supports «AI must not be a barrier».

The practical test for your own deployment is to compare satisfaction in conversations that passed through AI against those that reached a human directly. If the gap is large, the problem isn’t the automation but how easily people can leave it.

Where do these figures come from?

This report aggregates 2026 consumer surveys on attitudes to AI in customer service and sets them against benchmarks of how AI support actually performs. Attitude data describes a stated position rather than observed behaviour; survey question wording materially affects results, so absolute percentages should be read as orders of magnitude.