AI customer service gaps push 34% of consumers to human agents

Singaporeans are sharply intolerant of bots that fail to recognise them.

Singapore consumers are comfortable handing routine tasks to artificial intelligence (AI) customer service agents, but they remain sharply intolerant of bots that fail to recognise them.

According to a Twilio survey, 78% of respondents said they have abandoned an AI interaction due to the bot lacking basic information about them.

As such, 34% switched to a human agent, 18% abandoned the interaction entirely, 21% continued anyway despite the friction, and 5% stopped engaging with the brand altogether.

Only 17% said their AI agents “very effectively” recognise returning customers and access history, well below Asia Pacific leaders.

When bots hand off to humans, only 38% of interactions carry over ‘some’ context, and 64% say the human agent knows at least something of the prior conversation.

Moreover, two-thirds (67%) say brands should disclose upfront when they are talking to AI rather than a person. On the corporate side, 83% of organisations say they clearly identify their agents as AI, though only 20% do so from the very start.

To feel more comfortable with AI agents, Singaporeans ranked being able to switch to a human highest (67%), followed by knowing the AI requires human approval for actions (54%).

Amongst businesses, the top trust-building priority is offering explainability — "why did the AI recommend this" features (51%), followed by an easy opt-out to a human (51%).

Currently, Singaporeans estimate 49% of their customer service interactions are handled by AI, a figure they expect to rise to 59% by 2027.

Seventy percent say AI bots have gotten better over the past year, and on average, respondents estimate AI saves them 76 seconds per interaction.

Still, only 47% say they are more likely to choose a chatbot over a human than they were a year ago.