The AI investments are in. The results aren’t. The business leaders seeing the benefits rethought the very foundation of their customer service operations.
For organizations that have deployed AI—and the customer experience leaders who were supposed to benefit from it—an uncomfortable truth is emerging. The pilots ran. The press release went out. And the customer satisfaction scores have barely moved. Somewhere between the technology investment and the customer outcome, something is breaking down, says Paul Fipps, president of global customer operations at ServiceNow, and most organizations already know what it is. Data is fragmented, systems don’t talk to each other and the AI sitting on top of all of it is only ever going to be as good as the infrastructure beneath it. That gap between investment and outcome is becoming one of the defining operational problems of the AI era, according to Fipps. “Every company wants to tell you they’re reinventing customer experience,” he says. “Most of them are just bolting AI on top of 50-plus-year-old technology. If you’re not solving for the foundation first, that can be an expensive mistake.”
The operational reality behind enterprise CX is sobering. According to a recent ServiceNow study, customer service representatives spend only 45% of their time actually serving customers. The rest is absorbed by system toggling, data hunting and managing handoffs across platforms. Service agents report having to log in to as many as five different systems to resolve a single case. Meanwhile customer expectations have moved faster than most roadmaps anticipated. The gap between what customers want and what they receive is widening, according to the study. In the race to adopt AI, customers are optimizing for speed at the expense of human empathy. For CIOs, that means pressure from both ends: customers demanding instant resolution and boards seeking demonstrable return on investment.
“Every company wants to tell you they’re reinventing customer experience. Most of them are just bolting AI on top of 50-plus-year-old technology. If you’re not solving for the foundation first, that can be an expensive mistake.”
– Paul Fipps, President, Global Customer Operations, ServiceNow
At ROSSMANN, one of Europe’s largest drugstore chains, that friction cost an average of nine minutes of human labor per support case across 5,200 stores—time a store manager spent on hold with headquarters instead of with customers. Today, when a store associate finds a damaged delivery, they snap a picture with their phone and an AI agent scans the photo, classifies the issue, pulls the relevant guidance and routes a complete case, all before a human would have finished dialing. Routing accuracy reached 98%, according to the company. The nine-minute case became five seconds.
The pattern behind that result is gaining traction across industries, Fipps says: connecting systems rather than replacing them to link data across platforms in real time without moving or duplicating it, so every workflow draws from the same current picture of a customer.
“Every CIO I’ve talked to has heard the rip-and-replace pitch and been burned by it,” he says. “You can’t go in and take out all the systems of record inside these companies. You connect what’s already there, put AI to work inside those workflows and move fast without completely disrupting the business.”
Fragmentation isn’t just an IT problem. It’s a people problem, one that steals time from the humans who are supposed to be serving customers. Fewer than a third of enterprises have made meaningful progress toward unified customer data. The result plays out on the front line every day: customer records that tell a different story depending on where you look and customers on hold while service agents search for answers that should already be there.
The organizations that have solved the infrastructure problem aren’t running pilots anymore. They’ve moved to a different set of questions entirely: what work should humans still own and what should run on its own.
The shift is from AI as a capability to AI as a workforce, Fipps says. Agentic systems triage cases, analyze customer sentiment and intent, verify information, complete the next step and hand off to humans only when judgment is genuinely required. They don’t suggest an action. They take one.
The implications reach beyond efficiency. When machines handle the routine, the humans remaining in the loop are doing qualitatively different work, Fipps says, and organizations are beginning to staff and train accordingly. But as autonomous systems take on more of the operational load, a second question is rising: Who governs what the AI decides, and what happens when the technology gets it wrong? That conversation has moved beyond IT to the boardroom, he says.
When the foundation is right, the customer outcome follows, and it follows fast. Bell Canada deployed AI agents on high-stakes-escalation intake cases, with significant customer friction. The technology handles the preparation: creating the case, validating completeness, detecting duplicates and routing to the right team. By the time an escalation manager picks up the phone, the work before the work is already done. Customer response times dropped 25%. AI-assisted cases run at 90% accuracy, according to the company.
“We place AI where friction lives,” says Hadeer Hassaan, Bell Canada’s chief information and customer experience officer. “Our agents and customers prove the value.”
That is the dividend at the human level, too. The service agent who is no longer toggling between five platforms—whose AI tool has already triaged the case and surfaced the customer’s full history before the call connects—can focus on the interaction itself, the thing that actually requires a person.
We are still in the early chapters of what agentic AI will do to enterprise CX. But the organizations writing those first chapters are already pulling ahead, according to ServiceNow’s Fipps. The gap between them and the companies still debating infrastructure is not theoretical. It shows up in response times, retention rates and the daily experience of the people doing the work, he says.
The question facing most CIOs now isn’t whether to move, according to Fipps. It’s whether they’ve already waited too long.
Learn more about how AI is shifting the CX landscape—and how to win.
LEARN MORE
This content was created by Custom Content from WSJ, a unit of The Wall Street Journal Advertising Department.
Top Challenges in Using AI for Customer Interactions
Executives think they know the top challenges—customers disagree
49%
50%
23%
30%
48%
16
%
14
%
20%
40%
37%
Lack of empathy or understanding of my concerns
Unclear explanations about processes and policies
Being transferred to multiple people/ departments
Need to repeat issue to agents or reenter data
Limited communication channels
The Unification
Problem
The Agentic
Shift
The Pressure
Is Real
Companies Spent Billions
on AI for CX.
Are They Getting It Right?
Source: “The CX Shift: A study of customer expectations in the AI era,” ServiceNow, 2026.
Organizations’ progress in improving each area through AI*
Top 4 Customer Values in CX
Source: “The CX Shift: A study of customer expectations in the AI era,” ServiceNow, 2026. *Midway or advanced in AI implementation
The Dividend for
Getting It Right
% customers citing
% executives citing
32%
32%
38%
01 Responsiveness
Top 4 Customer Values in CX
Source: “The CX Shift: A study of customer expectations in the AI era,” ServiceNow, 2026. *Midway or advanced in AI implementation
02 Trustworthiness
03 Security and privacy
04 Emotional connection
16%
32%
32%
38%
16
%
2. Trustworthiness
4. Emotional connection
3. Security and privacy
Responsiveness
Customer values
Organizations’ progress in improving each area through AI*