THE CHALLENGE
The pressure to automate customer service is intensifying. AI powered CX (from chat-bots to Voice AI) are usually based on transcript analysis and internal best-practice guidelines.
While these are critical inputs, they only capture assumptions about customers and what they say in CX interactions, not necessarily what they really feel, need, or expect across contexts. The wider this gap in understanding, the lower the ROI on AI-powered CX programmes.
The distance between words and truth
Customer behaviour in service interactions is shaped by more than the task at hand. Power dynamics play a role: customers often adapt their language strategically, softening complaints, overstating urgency, or framing their situation in ways they believe will get results. Context also matters, as more urgent or high-risk intense creates a multitude of anxieties when they don’t have a human to reassure them. For customers with strong regional accents or accessibility needs, automated systems introduce additional anxiety, which may not surface in transcript data.
Standard satisfaction metrics compound this picture. NPS surveys tend to over-represent customers at the emotional extremes (attracting the most delighted and the most frustrated) skewing the overall read on experience quality.
How primary research can inspire better AI powered CX
In depth primary research can build a richer psychological model of the customer: their state of mind across different call types, their anxieties, expectations, and the reassurances they need to feel confident handing control to an AI. These vary by call intent, by customer profile, and by brand relationship. A customer paying a premium expects a different quality of AI interaction than one on a basic plan.
Bespoke insight by brand and CX context shapes better execution: the right tone of voice, the moments where human handoff really matters, the language and protocols that build trust rather than eroding it.
We recently worked with a large UK insurance provider on exactly this. Using qualitative online diaries and a 1,200-person quantitative survey, we mapped customer psychology across key call intents — producing clear guidance on how their AI should be voiced, positioned, and designed to maintain brand premium while delivering genuine efficiency gains (across call intents).
The ROI case
The metrics companies use to evaluate CX automation - containment rate, Average Handle Time, First Contact Resolution, CSAT - can all improve when design is grounded in real customer insight. Better execution means fewer escalations, less pressure on human agents, and faster resolution. Research isn’t a cost sitting alongside technology investment. It’s a multiplier on it.
If you’re interested in learning how primary research can accelerate the ROI of your CX automation strategy and provide deeper insights into customer psychology, please get in touch. Together, we can explore how to leverage these insights to refine your approach and achieve meaningful results.