News Analysis
Research Report
The AI Transformation 100: Concrete Ideas for Fixing How We Work
Key Takeaways
- AI is making contact center work more complex by removing routine tasks and leaving humans with the hardest cases.
- CX leaders need to design operations around judgment, escalation and flexibility — not rigid policies.
- Successful AI rollouts start with specific problems, small pilots and fast iteration.
- Traditional contact center metrics fall short when AI handles the easy work and humans handle the exceptions.
Twenty years ago, the bulk of a contact center agent's day was routine: resetting passwords, answering billing questions. The kind of work that followed a script and ended with a click.
That reality of work is rapidly disappearing. As AI absorbs the simple and repetitive tasks that once filled agents' queues, what’s left over for human employees is harder, more ambiguous and more emotionally charged than anything contact centers were originally designed to handle.
That tension — between the promise of AI efficiency and the increasingly complex work left for humans — was one of the dominant themes this year at Customer Contact Week in Las Vegas.
AI Doesn't Eliminate Complexity — It Concentrates It
"What's the standard operating procedure for things that don't have a standard operating procedure?"
- Bryan Stoller
VP, Global Head of Customer Care, United Airlines
Bryan Stoller, Vice President and Global Head of Customer Care, Contact Centers, and Solutions at United Airlines, began this year’s keynote by painting a picture of how the mix of work hitting frontline agents has shifted.
Two decades ago, routine interactions made up the largest share of an agent's day. Today, that slice has shrunk, while moderately complex and highly complex interactions have grown. And as AI takes over more of the routine work, that shift is set to accelerate dramatically.
"What's the standard operating procedure for things that don't have a standard operating procedure?" Stoller asked. That, he argued, is the question every CX organization needs to answer — and fast.
He gave a vivid example: a family traveling with a child, stuck at an airport after a weather delay, about to miss their last connection of the night. The policy is straightforward — weather-related, rebook them on the morning flight. But one traveler's medication is in checked luggage. A child is melting down. Every hotel in the area is sold out.
"Most customers, they're not trying to get around the rules," Stoller said. "They just want someone to look at the situation and make a reasonable call."
Related Article: The AI Accuracy Trap: Why MLOps Needs a Financial Circuit Breaker
Building a ‘Reasonable Department’
Stoller's solution isn't another escalation path or exception process. It's what he calls a "reasonable department" — not a literal department with a sign on the door, but a mindset and operating capability built into the organization.
At United, according to Stoller, that model is built on four pillars:
- Detect: Identify when an interaction goes beyond routine and requires a different level of support. This is critical not just for human agents, but for AI systems that need to recognize when to hand off a case rather than attempt to resolve it.
- Route: Get the issue to the right human capability, not just the next available agent. The person best equipped to handle a complex, emotionally charged situation is likely not whoever happens to be free.
- Resolve: Give teams the context, tools, authority and judgment frameworks they need to actually solve the problem. "You cannot constrain them by black and white policy," Stoller said.
- Learn: Turn hard cases into fuel for insight across the business — informing policy changes, self-service improvements, operations, training and AI development.
"This is about not designing our organizations for the work that AI takes away," Stoller said. "This is about designing our organizations for the work that AI leaves behind."
Start With the Problem, Not the Technology
A recurring theme during the keynote was the danger of deploying AI for its own sake. Multiple experts brought up the same idea: start with the problem, then work backward.
One way to do that? Follow the data. Analyze why a customer is contacting the company in the first place, then target the highest-volume, most automatable interactions.
Jessica Gupta, Chief Operating Officer at InfoPay, described how their team found that a large segment of customers actually preferred voice as their channel of choice, which opened up a clear opportunity.
"We found out that we have a group of customers, fairly large, that really wants to talk to us," Gupta said. The team focused on automating a specific, high-volume process where customers only needed a simple confirmation — ultimately containing a significant portion of those interactions and freeing up experienced agents to contribute insights across the business.
Aaron Johnson, interim Chief Marketing Officer at Penn Medical, described how growing patient demand and unsustainable call volumes forced the organization to rethink its access operations entirely.
The challenge wasn't just technology — it was the integration work required after mergers and acquisitions, where agents in one part of the system couldn't schedule appointments in another, leaving patients unable to get care.
Start Slow, Learn Fast, Don't Be Afraid to Fail
One point of AI advice given by CX leaders? Start small, measure carefully and be willing to pivot fast.
Gupta described a deliberately incremental approach: "We only had 20 calls — when you go to 20 calls, it's nothing, right? We had one [success], but we don't need to turn it on for half of our calls." The idea is to scale only after validating results at each stage, not to rush toward full deployment.
Neville Letzerich, chief marketing officer at Talkdesk, offered blunt advice for executives feeling pressure to move faster: "Get started. Try something. Don't go overboard. Be intelligent about it. Go get a win, leverage that, or lose and say, 'Wow, that didn't work,' try again."
Johnson noted that when Penn Medicine deployed a voice assistant at scale, the diversity of what patients actually said on the phone far exceeded expectations. "In retrospect, we may have wanted to start with a smaller pilot," Johnson admitted. But the team recovered by pairing business-side leaders with technical tuning teams to rapidly iterate, reviewing what patients were saying and adjusting routing logic in real time.
The Hybrid Workforce Demands Transparency
As AI agents increasingly work alongside human agents, panelists stressed that managing the hybrid workforce requires a fundamentally different approach to change management — starting with honesty.
Gupta described an approach rooted in early, transparent communication: years before their AI rollout was fully underway, leadership told employees directly that AI would change the way they work — even if the organization didn't yet know exactly how.
The payoff was employee buy-in. High-performing agents were brought into the process early, testing AI tools and providing feedback. When the system went live, those employees felt ownership over the change rather than being subjected to it. "Now they're actually looking at AI favorably," said Gupta.
The Hidden Cost of Fragmented Customer Communication
Register
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
Not every employee will embrace the transition. But finding people with the right skills to manage AI tools alongside traditional customer interactions remains one of the biggest challenges.
Old Metrics Won't Cut It
Multiple panels touched on the need to rethink how organizations measure success in a hybrid AI-human model. Traditional contact center metrics like average handle time and transfer rates still matter, but they don't capture the full picture.
Jonathan Rosenberg, Chief Technology Officer at Five9, said it’s important to pay closer attention to business-level outcomes: customer return rates, contract renewal rates and customer satisfaction scores.
"Make sure these things remain good, so you're keeping the top-level picture," Rosenberg said, arguing that operational metrics alone can be misleading when AI is handling the easy work and humans are left with the hardest cases.
Johnson said that at Penn Medicine, his team encountered an unexpected measurement challenge: as their AI system improved call routing and connected more patients to care, costs actually increased, because more patients were successfully getting through — a reminder that efficiency metrics can tell a very different story than effectiveness metrics.
Related Article: The Subtle Signals That AI Is Going Off Track
Leadership Is the Connective Tissue
Leadership — especially middle management — plays a critical role in making AI transformation work.
Managers and supervisors sit at the intersection of strategy and execution. They're the ones hearing from frontline agents about what's working and what isn't, and they're responsible for carrying those insights back up to the C-suite.
The CFO, CTO and HR leaders may all be optimizing for different outcomes, and it falls to operational leaders to translate between them. The most successful organizations are the ones where that communication loop — from the front line, to management, to the executive suite and back — is functioning well.
As Letzerich put it, this is about more than technology: "This is a huge opportunity, especially for all of us in customer experience. We can deliver really great customer experiences at a fraction of the cost. We can serve customers faster. We can give them new experiences they didn't even know about."
But realizing that opportunity requires investing just as heavily in the human side of the equation as the technological one.