A global study on customer experience confirms something most people in service delivery already suspected: organizations have gotten a lot faster at helping customers, and customers have noticed. What the study also found is less comfortable reading. Getting faster hasn't made customers feel more understood. For a lot of them, it has made things worse.
ServiceNow partnered with the research firm ThoughtLab to survey more than 34,000 people across 18 countries and eight industries for a report called The CX Shift. The sample included customers, frontline service reps, and the executives who run customer operations, with nearly 1,900 American consumers among the respondents, alongside hundreds of U.S. reps and executives. The headline numbers are global, but for anyone running service delivery in the U.S., public sector or private, the pattern will feel familiar.
Speed got solved. Empathy didn't.
More than half of customers, 53%, say faster and more efficient service is the top benefit they expect from AI going forward, and 40% say AI has already delivered on that promise in their day-to-day interactions. That part of the equation is working.
What isn't working: 48% of customers still rate their service experiences as average or worse, and when asked what frustrates them most, the top answer, cited by exactly half of all customers, isn't slow response times or clunky software. It's a sense that nobody on the other end actually understands what they're dealing with. One American consumer in the study summed up the trade-off well: for something quick and simple, AI works fine, quicker and more efficient than waiting on a person. But for the moments that actually matter, most people want a human who can tell they're frustrated and adjust accordingly. Another U.S. respondent put it more directly, saying that talking to a real person gives you the sense that someone on the other end actually has feelings and understands your situation.
That distinction isn't a nuance. It's costing companies customers. Forty-seven percent of people in the study say they've switched providers because of slow or inadequate service, and a lack of empathy sits right alongside that as a reason to walk.
Executives are solving the wrong problem
Here's where it gets uncomfortable for leadership teams. Researchers asked customers what frustrates them most about service interactions, then asked executives to guess. Half of customers named lack of empathy or understanding. Only 23% of executives guessed the same thing. The gap held across nearly every category. Customers said being bounced between departments was a major frustration, 48% of them. Executives underestimated it by 18 points. Customers said having to repeat their issue to multiple people was a top complaint, 40% of them. Only 16% of executives saw that coming.
That's not a small miss. If leadership is solving for the wrong problem, budget and attention go to the wrong place, and the data shows exactly that: only 16% of organizations report meaningful AI-driven progress on building emotional connection with customers, even though it's one of the things customers say matters most to them. Channel strategy tells a similar story. Eighty-seven percent of customers rank phone calls among their preferred ways to reach a company. Only 7% of executives plan to prioritize that channel over the next three years. A lot of organizations are chasing the next digital surface while their customers are still reaching for the phone.
What's actually happening on the ground
None of this really comes down to whether the AI itself is good enough. It comes down to what's happening underneath it. Eighty percent of service reps have to log into three to five separate systems just to work one customer issue. Less than half of their time, 45%, goes toward actually helping customers. The rest disappears into administrative work, chasing colleagues for information, and reconciling data that doesn't match across systems. Forty-three percent of reps call inconsistent customer data a top challenge. Only 28% of executives think it's a problem at all, which says something about how far removed leadership can be from the daily reality of the people they manage.
A U.S. retail executive who took part in the study described what fixing this actually looks like: giving agents ongoing, AI-guided coaching so their skills and performance keep improving over time, rather than a one-time training session that goes stale. A technology company executive described a related shift, connecting CRM and support platforms so agents work from real-time data instead of guessing. Both are describing the same underlying fix. Give the person doing the work one place to see the whole customer, instead of five separate logins.
Where trust and security actually fit
The study identifies four things customers value most in a service experience: responsiveness, trustworthiness, security and privacy, and emotional connection. Of those four, security and privacy is where organizations have made the most AI-driven progress, at 38%. That tracks with how much regulatory and compliance pressure most U.S. organizations already operate under, especially in government, healthcare, and financial services. But trustworthiness sits at just 32%, level with responsiveness, and emotional connection trails everything else at 16%.
That ordering matters right now for anyone building or buying AI-enabled service tools. Security and access controls have to be the starting point, not an afterthought bolted on after a fast rollout, and that's usually not news to organizations that have spent years working under FedRAMP, HIPAA, or state-level compliance requirements. The mistake is stopping there. Getting security right is the entry ticket. It is not the same thing as building an experience customers trust enough to bring their harder problems to.
There's a useful way to think about why this keeps happening. Most CRM systems were never actually built to deliver customer experience. They were built to track sales pipelines and internal metrics. As AI got layered on top over the past few years, it inherited that same DNA, optimizing for what the business could measure rather than what the customer actually needed. The fix isn't more AI stacked on the same foundation. It's treating the platform as a system of action instead of a system of record, one place where data, workflows, and AI are genuinely connected, instead of stacked on top of a database that was never designed for the job.
What to act on now
Three things from this research are worth moving on now, not eventually.
Measure what customers actually say matters. NPS and CSAT won't catch an empathy gap on their own. Sentiment analysis and effort scoring will, and pairing them with the traditional metrics gives leadership a fuller, more honest picture of where service is actually falling short.
Fix the underlying data before layering on more AI. Only 43% of organizations have unified customer information into a single source of truth. A rep can't deliver a connected experience from a fragmented system, no matter how capable the AI sitting on top of it is.
And don't confuse security with trust. They're related, but they aren't the same thing, and customers can tell the difference even when they can't name it. Only 34% of executives report meaningful progress connecting people, data, and processes on one platform. That gap, not the sophistication of any particular AI model, is the actual bottleneck most organizations are working against.
The organizations that close it will look, on paper, faster than their competitors. What customers will actually notice is that someone finally listened.
