March 24, 2026
Ada Research Finds Customers Prefer AI Service on One Condition
While many customers may still feel hesitant about AI-powered support, new research from Ada suggests this reluctance may be fading, provided the technology resolves their issue. This crucial caveat reflects a wider truth that consumers don’t view AI in black and white terms. Instead, they evaluate it as they do any other tool, weighing its pros against its cons.
Ada’s Key Findings
The Ada study, conducted with NewtonX, found that 59 percent of consumers prefer always-on AI support over waiting in a queue for a human. Speed and convenience now hold significant weight, and many customers are willing to default to automation if it gets them to the outcome faster.
Yet the same research shows that only 24 percent of recent AI interactions were fully resolved without human assistance. The other 76 percent either required escalation, delivered only partial help or were abandoned. Appetite for automation is clearly strong, but tolerance for incomplete resolution is not.
Ada CEO, Mike Murchison, believes it is therefore important to keep an eye on resolution rates: “Most businesses are still optimising for the most familiar metrics: deflection, containment, and cost savings. But those are now table stakes”. He continues: “What this new data reinforces is what these metrics don’t tell us: was the customer’s problem actually solved?”.
The survey also asked consumers what matters most in customer service, revealing that accuracy and capability outrank empathy. This finding challenges long-held assumptions about emotional reassurance. When AI fails, the top causes are comprehension failures where the system doesn’t understand the request, capability limitations that prevent it handling complexity, and repetitive, unhelpful responses.
On the enterprise side, the report finds a significant operational gap. Fifty-five percent of businesses lack visibility into AI agent performance, often measuring AI and human interactions together. Without clear separation, it says, organisations risk assuming the technology is performing well when customers are experiencing the opposite.
A Contradictory View
These findings land somewhat differently from from Metrigy’s CX Optimisation 2025-26 study, which found that nearly 85 percent of participants still favoured human interactions over AI. Even if issue resolution was assured, over 80 percent reported that they would still rather speak to a human agent.
The results couldn’t be more opposed but the explanation for this may stem from something that both studies agree on: attitudes towards AI are conditional. Ada’s study, on the one hand, highlights that people are willing to use AI when it works reliably, while Metrigy pointed to trust in AI being an essential proviso. Customers are choosing the channel that delivers the outcome they want with the least friction.
Signs Point to On-Going Evaluation
This mixed sentiment aligns with broader industry behaviour. In retail, customers increasingly engage in AI-assisted shopping journeys, particularly when the tasks are simple and the value clear. Conversely, other research highlights that people remain less forgiving of AI mistakes, which is why incomplete resolutions can feel more frustrating than slower human service. There is also evidence from an 8×8 study that recorded continued resistance to AI chatbots in the UK, particularly when it is about something urgent.
Across all these insights, one theme cuts through. It is all too tempting to try and reduce customer sentiment to an overly simplistic view when, in reality, it is composed of highly rational assessments. People welcome it when it is fast, accurate and capable. They push back in areas where it is not performing well, making AI adoption less about a one-off switch and more about continuous tuning. Organisations, therefore, need to listen to feedback and regularly refine their AI agents to ensure positive customer experiences, which will be as nuanced, varied and changing as the technology itself.
