Feature
Research Report
Building the Case for Modern Public Sector Messaging
Key Takeaways
Customers don't want AI — they want faster, easier, better experiences, and poorly deployed AI often delivers the opposite.
The smartest companies are slowing down, starting small and aligning AI deployments to specific use cases rather than chasing hype.
As AI handles routine tasks, human agents face harder problems — which means training, metrics and morale all need to be rethought.
AI literacy has to be built from the ground up with role-specific guidance, not top-down mandates to "just use it."
At Customer Contact Week (CCW) in Las Vegas, the largest customer contact event in the world, one thing became crystal clear: the era of rushing headfirst into AI is over. What's replacing it is a more measured, human-centered approach — one that prioritizes outcomes over hype.
I spoke with three CX and technology leaders at the event, and despite their vastly different industries and company sizes, the throughline was unmistakable: AI only works when it's deployed with intention, not urgency.
Customers Want Better Experiences, Not AI
Leaders at the event were pretty blunt in their assessment of AI: customers don't care about the technology. They care about results.
"Customers aren't begging for AI, they're begging for better experiences," said Chad Anderson, Head of Customer Care & Roadside Relations at Mercedes-Benz USA. Poorly implemented AI solutions often create more friction for customers — requiring more effort, generating more frustration and forcing them to repeat themselves when they finally reach a human agent.
Michael Darwal, Chief AI & Digital Officer at ibex, framed the point through data. He cited an MIT study that found the majority of AI implementations fail or don't deliver the expected results, largely because companies target deflection without understanding the consequences.
"They're not setting the proper expectations and not going at it with a true understanding of what that customer journey needs to look like," he said. "They're targeting in many cases just deflecting a human interaction, and without necessarily understanding what the consequence of that is."
Related Article: What Comes After AI? Leaders Say the Hard Part Is Just Beginning
It's a Spectrum, Not a Switch
One of the most persistent misconceptions in the AI-CX conversation is that automation is a binary outcome — either AI handles it or a human does. Darwal pushed back on that framing.
"Many people are looking at it and saying it's a binary outcome. Did this get resolved with just an AI agent? Did this get resolved with a human agent?" he said. "We're looking at it as a spectrum."
That spectrum approach means AI might handle the front end of an interaction — authentication, basic information gathering, routing — and then hand off to a human agent who's already equipped with context. Darwal said this model can cut 30% or more of a human agent's handle time by automating what he called "the really painful parts at the beginning of a phone call."
ibex's work with Philippines Airlines shows this model in practice. The company deployed AI agents in six weeks — in three languages, including Tagalog and Taglish — to handle seat changes, flight modifications and similar high-frequency, low-complexity requests. The CSAT score results:
- AI-handled interactions: 4.7 out of 5
- Typical human interactions: 4.3 out of 5
Anderson described a similar philosophy at Mercedes-Benz, where AI implementation is filtered through three pillars:
- Saving the customer time
- Saving the customer effort
- Going above and beyond
Any deployment that doesn't meet that standard doesn't move forward.
"A lot of companies are rushing to implement it for immediate impacts, but not really thinking of the long-term impacts," said Anderson. While a lot of organizations are looking for cost savings, he urged companies to think of the effect on brand trust and customer loyalty.
The Complexity Problem
As AI absorbs routine interactions, human agents are increasingly left with the hardest problems — and that's creating a new set of challenges that few companies planned for.
Anderson said this is top of mind at Mercedes-Benz. As agentic AI takes on more routine tasks, frontline agents will need enhanced skillsets to handle the complex issues that remain. "We need to make sure that we're training them to be able to prepare for that and to enhance their skillsets, because a lot of roles are not necessarily being eliminated, but they are evolving as AI gets more implemented."
Darwal's approach at ibex is to use technology to close the gap. His team has built desktop-level tools that listen to calls in real time and surface potential answers and suggested responses before the customer even finishes speaking — removing the need for agents to manually search a knowledge base while a frustrated customer waits.
"The goal is to be able to utilize that technology to make the job of the agent easier, even on those more complex interactions," Darwal said. "Otherwise, being an agent is not always an easy job, and it becomes tougher if every one of those is the more complex interactions."
Genelle Chamberlain, IT Manager at PrimeSource, raised a related concern: the shift could break existing KPIs. Contact centers have long measured success by metrics like average handle time and first-call resolution, but if agents are only handling the most difficult issues, those benchmarks will inevitably shift. She pointed to agent morale as well — employees accustomed to quick, gratifying resolutions may struggle as their caseloads become uniformly complex.
Governance & the "Go Slow to Go Fast" Playbook
Behind the scenes, enterprises are wrestling with the operational realities of AI deployment — and governance was a recurring theme at CCW.
Anderson described Mercedes-Benz's governance process as thorough by necessity. As a 140-year-old global brand, the company can't afford to move recklessly. Regulatory differences between the US and Europe add another layer of complexity, and any AI deployment must account for multiple markets with different customer expectations and legal frameworks.
"We have to really be cautious and not necessarily rush to do something because it's the next big thing, but really being intentional to see how it can further enhance the legacy of our brand," Anderson said.
Chamberlain shared a more tactical challenge. "One thing we found with Copilot is it kept pulling from [external sources], even though we set those guards, we set those boundaries, it was still trying to pull outside." It's a much-needed reminder that setting AI boundaries is an ongoing process, not a one-time configuration.
Her company has since set up a cross-functional AI council with representatives from data, logistics, customer service and marketing to guide adoption company-wide.
Darwal, meanwhile, argued that being thoughtful and being fast aren't mutually exclusive — but that the order of operations matters. His company invests heavily in mapping the customer journey before any automation begins, which he said is why the company was able to deploy for Philippines Airlines in six weeks while competitors spent months failing to get out of hallucination.
How Modern Marketing Is Exposing the Limits of Legacy CMS
Register
Why Some Dealers Are Pulling Ahead With AI
Register
Content Leaders Collective: Proving Content's Business Impact Starts With the Right CCMS
Register
Replacing Tasks, Not Roles. The Changing Nature of Contact Center Work
Register
"The goal is to be able to do it both thoughtfully and quickly," said Darwal. "Anybody can do anything bad really fast, but the goal is to be able to do it once well and getting the best outcome possible."
Related Article: Authorization Before Autonomy: A Governance Model for Agentic Commerce
Building AI Literacy From the Ground Up
All three leaders stressed that successful AI adoption requires meeting employees where they are — not issuing mandates from the top.
Anderson said Mercedes-Benz holds weekly AI meetups where employees share how they're using AI in their roles and discuss industry developments. He added that leaders need to provide specific, role-relevant use cases rather than broadly telling people to "go use AI."
"Unless you're someone that is very techie or just really interested and curious, you're not going to proactively dive into AI," he said. "Leaders need to focus on specific use cases for how their employees can use AI, as opposed to saying, here's AI, use it."
Chamberlain described a similar approach at her company, where the AI council runs training sessions built around real-life scenarios tailored to different roles. For some employees, the use case is as simple as adjusting the tone of an email. The key, she said, is removing the intimidation factor. "A lot of them have a fear of even getting into AI, like they just don't even know where to start."