For the second year running, CX Network's annual research into the state of CX found that AI-powered technologies for operations is the most significant trend impacting CX practitioners, with AI agents following closely behind in second place.
AI has already vastly altered the contact center landscape in the last decade and it is showing no signs of slowing its transformative effect, with 46 percent of practitioners expecting spending on generative AI, and 29 percent expecting spending on agentic AI, to increase in 2026.
With vast change, though, comes anxiety. Fears around headcount reduction, diminished customer service quality and changing customer expectations are front of mind for CX and contact center leaders and their staff.
In July 2026, All Access: Future Contact Centers sought to unpick some of these anxieties by digging into the impact AI is actually having on contact centers and CX, and what leaders can reasonably expect from the future.
A throughline that appeared was that all speakers, in their own way, argued that AI's real value in the contact center is elevating the human role and that the organizations succeeding are the ones basing AI transformation around not on cost reduction, but on service quality.
This article rounds up the main take aways from CX practitioners at Orange International Wholesale Markets, Fellers, TikTok Shop US, Back Market, Scotiabank, Rail Europe, Roche Pharmaceuticals and PayByPhone. The CX Network round-up on what solution providers said during the event, is here.
Contact center transformation begins with purpose, not KPIs
Our opening panel featured Tristan Rayroles, head of CX at Orange International Wholesale Markets, alongside John McCahan, chief customer experience officer at Fellers.
They are both heading contact center transformation but in very different circumstances. Rayroles is navigating tight security requirements within limited means, while McCahan is modernizing a contact center that previously ran on minimal integrated tech.
They both agreed that purpose must come before metrics. "Transformation is absolutely not about polishing KPIs, but serving your purpose," said Rayroles. "We use KPIs to measure if we are on the right track."
McCahan noted that some of the traditional contact center metrics – average handle time (AHT), specifically – are no longer fit-for-purpose. Instead, leaders should focus on first contact resolution. This will change culture, he said, and for the better. "Supervisors and team members say 'wait, what? You want me to solve the problem and stop rushing through it'?" McCahan mused.
The result is happier agents and, ultimately, happier customers.
Catch up with the discussion on CX+
TikTok Shop US: Scaling AI in CX for a platform growing 500% a year
At launch, TikTok Shop's US revenue was US$1-2 billion and is now "trending towards $25 billion", according to Sujit Mohanty, senior CX program manager, who joined us to discuss his approach to scaling CX. He said the platform currently sees 40,000-50,000 tickets per week has more than one million sellers.
There are fundamentally different trust dynamics at work, with many purchases being impulse-driven. "There is a lack of trust […] if things don't go right… all of those things lead to customer contacts eventually," said Mohanty. Additionally, on this growing platform, there are often 10-15 new policies or features launched per month, he explained, which created confusion among buyers and required constant updates to the knowledge base.
In this scenario, Mohanty said that low-risk, reversible decisions, such as small refunds or replacement orders, are the safest to automate; those he described as "two-way doors". Converserly, "one way doors", such as suspending a seller's account or freezing funds, demand AI and human collaboration with an appropriate level of human judgement in the final decision.
Agent co-pilots have been a major success for TikTok Shop US, reducing cognitive load and delivering a 150 percent improvement in resolution time. So where did the team start? With the unglamorous work in the knowledge base, of course.
"It's definitely not glamorous… it's reformatting documents so that the AI models can parse information really well," he explained. A good knowledge base must be authoritative, and structured. "AI is all good. It speeds up everything, but ultimately it relies on the foundational data," he said.
Catch up with the discussion on CX+
Back Market: Rebuilding customer care after a public trust crisis
In 2024, Back Market had a PR problem. An efficiency drive in the contact center had caused backlash. The self-service strategy looked like it was working internally but the reality for customers was different.
"The contact funnel looked pretty healthy, churn appeared to be minimal," said Rufus Kahler, head of product design, content and research.
What was not initially clear was that customers were arriving to the self-service options "already having spent time attempting to self-troubleshoot on other platforms like YouTube and Google", missing further support options and leaving feeling deliberately ignored.
In the nascent refurbished tech category that Back Market operates in, this can do enormous harm to customer trust.
Looking at years of CSAT data reframed the contact center from cost to growth center. The data showed that there was "a pretty strong correlation between the quality of the customer's after-sales experience and their likelihood to repurchase", moving the conversation "away from cost to serve… towards how can we drive repeat".
The team set about redesigning the contact center, naming the project Care 2.0, which featured three main pillars:
- Ease of contact
- Faster first response via AI
- A wider channel mix.
The team started by intentionally building a broader view of support, so cross-functional teams were "able to really build a clear picture of what the customer sees from the outside in, even when things look good internally".
Care 2.0 is hybrid by design, in which AI takes simple and even personalized order queries while humans keep the more ambiguous, complex or multi-party cases.
"When there is a human in the loop, there's really no better way to restore trust," said Kahler.
We asked Kahler about how he and the team made the decision to build rather than buy – especially as a re-platform project led by a designer in a business that loves to build!
He said that taking a hybrid approach works best here, too. For business-specific capabilities, the best option is often to build. For business-agnostic capabilities, "view it as a commodity" and buy or rent. For when the reality is inbetween, as in Back Market's case – take a hybrid approach. Feature parity was essential during the decision-making process as the team "didn't want to give up what we already had".
Catch up with the discussion on CX+
Scotiabank: Cultural transformation needs to keep up with tech transformation
AI has proliferated contact centers, and tools often have very similar functionality. In this context, "the technology isn't going to be the differentiator," said Heather Arthur, vice president of client experience centers at Scotiabank.
"What is going to be the differentiator is actually the people that are using our technology," Arthur continued.
Arthur went on to discuss the practicalities of diving labor in an AI world. She recommended splitting tasks based on what only humans can do: "Let AI handle the what and the when… let the humans handle the why and the how," she said.
"Efficiency is not the same as experience… AI is definitely going to make us faster… but it's empathy that's going to make us memorable," she added.
There are two essential questions to ask around AI spending: Where are we spending human time on things that don't need a human? And where are we losing human moments and connections because we forgot to design for them?
Arthur said three "doors" will unlock peak contact center performance:
- Curiosity: Which she described as "the antidote to fear" of job loss.
- Empathy: Which "doesn't slow us down, it actually speeds up trust," Arthur explained.
- Culture: "Your leadership multiplied… what people are doing when no one's watching."
Arthur shared an easy-to-apply boost to culture that she has implemented: a 20 minute, weekly "Fri-Yay" meeting aimed at celebrating small wins. This kind of initiative can help to move the needle for contact center culture, considering recognition "improves employee engagement by over 31 percent".
To protect culture and ensure successful AI rollouts, HR, risk, and procurement must be involved from the outset. "AI is only… ever going to be as powerful as the culture that it lands in," Arthur said.
Catch up with the session on CX+
Rail Europe: AI as a path to true specialism, not fewer people
Media coverage around AI tends to highlight stories in which it is used to reduce contact center headcount as tasks are increasingly automated. This, however, isn't always the case. Valentina Cesani, chief operating officer at Rail Europe, joined us to share her altogether different approach.
Rail Europe is a travel tech platform built for non-domestic travelers navigating the European rail networks, which differ greatly from country to country.
The platform is designed to simplify and standardize otherwise diverse national rail systems. The service covers over 200 rail operators and serves more than five million passengers per year. AI translation has taken language coverage from 25 to more than 160 languages.
During the Covid-19 pandemic, the company was forced to downsize from 18 to five contact centers. In this, AI became a survival strategy.
"We understood pretty quickly that AI was the only way to go, to be able to be there for the customer and maintain the promise that we made," said Cesani.
Like Heather Arthur, Cesani defined AI strategy by deciding first what needed to stay human. In this, AI's role is to take the simpler tasks – the "fluff" as Cesani termed it – which "gives an opportunity to your agents… to be the ones really solving the problem".
Agents have thus become true experts in niche fields within the ecosystem of Rail Europe and as such are highly valued employees. The proof of concept comes in the form of retention, with an average agent tenure of 10 years, some over 25, and churn below 10 percent on average each year, against an industry standard of 30 percent.
KPIs also needed a rejig. "The average handling time is going up. So, what we measure is the first contact resolution… we calculate the customer satisfaction and the entry sentiment and exit sentiments," Cesani said.
Catch up with the discussion on CX+
Roche Pharmaceuticals: Break out of the bubble: how to get WFM a seat at the decision table
Bryce Ackerman, workforce management expert at Roche Pharmaceuticals, opened his session by redefining workforce management (WFM).
Based on a quote by Tim Buzza of Verint, Ackerman said: "WFM is the ability to balance the employee experience with the customer experience. And… no longer is workforce management purely staffing to the right person, right place, right time."
Breaking out of the bubble is the key challenge for WFM professionals, Ackerman said. Being good at what you do alone won't earn the seat at the table. He advised leaders in the audience to become a "linchpin", which he defined as "your company needs you so much, they cannot imagine making decisions without your input".
There is "no map", he said, and "no way to know exactly how to become a linchpin".
Ackerman then turned to practical advice, offering four pieces.
- He first recommended the 90/10 rule: 90 percent of a linchpin's time, "you're doing what you always do… The cool thing about a linchpin is they find an extra 10 percent of time to do those special things".
- Among senior leaders, a lack of time can be an opportunity, he explained. "They often don't have time to figure out what's going on… outside of their own organization", so bring them what you learn from your own research – webinars, podcasts and LinkedIn were all recommended.
- Capacity planning, he said, is "the quickest way to talk to senior leaders". Going to the office in-person matters, he said, and can make you happier at work. Ackerman explained that during the pandemic, when everyone was working from home, "I missed those one-on-one conversations, those face-to-face brainstorming sessions".
- Quality, training and including marketing and CX teams in your work were all also recommended.
Hear more from Bryce by watching his session on CX+ and reading this interview with CX Network.
PayByPhone: You can't measure empathy with a stopwatch
Chasing cost can erode the brand you are trying to protect. Customer experience and inclusivity should lead AI rollouts. This was the core argument of Anthony Cashel, chief operating officer at PayByPhone, global leader in mobile parking payments serving 125 million drivers per year, in 1,350 cities.
The single biggest mistake that leaders make when implementing AI in customer service is "treating it as a pure cost reduction exercise", Cashel said. The risk is that you "put your customer behind a digital wall".
In prioritizing CX, efficiency will naturally follow. "Those efficiency gains will come, those cost-saving gains will come, but it's really important to handle it in a managed way," Cashel said.
When asked how to sell this to a CFO, he said the secret is framing.
"Damaging customer trust, losing business is going to be much harder to pull back," he explains.
Cashel also recommended rolling AI out slowly, noting that this approach has meant that he has never had to roll a piece of automation back.
The AI rollout has changed the shape of agent roles, who now work truly for an escalation platform and can take on client, city and finance tickets. "There's lots of things that agents can do, we just need to make the capacity for them to do it," Cashel said.
Triage happens based on urgency rather than complexity. Someone "trying to park there and then right at that moment" needs the issue solving immediately. "Solving that query in an hour's time is really not helpful," he added.
Cashel recommended breaking large processes into smaller parts. He noted, though, that PayByPhone has not automated everything, and maintains 24/7 human support in the UK, for example.
"You can't really manage or measure empathy with a stopwatch," Cashel mused before explaining that PayByPhone audits interactions for plain English, runs a single-question post-contact survey and scans app and social reviews to glean customer sentiment.
Rather than looking at the, by now, much maligned metric of handle time, the team look at one-touch resolution rates.
Hear more from Anthony by watching his session on CX+ and reading this interview with CX Network.