In CX, the transformative power of automation is changing experience management for customers, agents and CX teams alike. But what do practitioners see as the top concerns when applying AI and automation on customer data? What prevents CX teams from having a truly unified view of the customer journey? What is the crucial first step to getting automation right? And, in three years’ time, how will organizations and customers interact with AI?
These questions were put to attendees at the CX Network webinar Re-writing the Rules of CX with AI Agents and CX Automation, and this article covers how those delegates responded.
The session was hosted by CX Network’s Georgina Wilczek and featured Gary Ovenell, VP of solution engineering for the EMEA region at Talkdesk, and Prashant Lulla, head of CX for the APAC region at Zurich Insurance.
Their wide-ranging conversation covered how AI can best be deployed within the CX landscape, to what extent it can be trusted to handle sensitive customer data, and how it might be used intelligently to tackle fragmented analytics, amplify customer trust and create more unified CX journeys. You can catch up with the full session on CX Network+ via this link.
Read on to find out how attendees at the session using artificial intelligence (AI) and automation in real-world CX, and how they see its use cases developing over the coming years.
Takeaway 1: Most organizations have a restricted view of the customer journey due to trapped data
During the session, delegates were asked what prevents their CX teams from having a truly unified view of the customer journey.
Among respondents, an overwhelming majority of 77 percent selected responses that prove siloed tools and a lack of integration continue to cause significant problems for the introduction and use of advanced AI and automation tools.
As many as 51 percent (see opposite) said their team was prohibited by data being trapped in legacy or siloed systems, for example the CRM or various contact center systems, and a further 26 percent said their primary barrier was a lack of integration tools or budget.
Lulla explained that it isn’t just the location of data that can be a barrier, but the format, too.
Once that barrier is overcome, to then be able to train a machine to utilize that data requires “an adequate amount of historical context”.
“That's where things are going to evolve, but I think several companies have definitely taken the first steps in that particular direction,” he added.
Ovenell commented: “To me, there are three key things to think about when you're going to put some level of AI into your CX flow: the engine, the data and the rules.
“When you're training an agent to come on board and start servicing your customers within the contact center, you teach them to use their brain, to be able to run a process, and you show them where the data is. Now, the great thing about a human is that we know how to look for bad data and how to bring in empathy and things like that. Technology is catching up fast and there is possibility to be able to do all of this, but I would never take the human out the loop, ever. At the end of the day, the human has that ability to be able to see things that just technology can't see,” he continued.
Meanwhile, 17 percent selected governance and security concerns around data sharing as their top barrier. To maintain harmony between humans and machines, ensure AI use is ethical and preserve customer trust, organizations require strict governance in how AI is (and is not) used. Data security is also of the utmost importance, particularly in a world where data theft can pay so generously.
Only six percent of delegates said a unified view of the customer journey was not a current barrier because they “manage fine with current systems”.
Takeaway 2: Customer trust and data security are the top concerns when applying AI/automation
Attendees were then asked to select their biggest concern when applying automation and AI to regulated and sensitive customer data. In response, as many as 42 percent said it was customer trust, “ensuring customers feel comfortable interacting with an automated system regarding sensitive matters”.
Key to customer comfort is trust and, on this point, the second most selected response was “the risk of AI models exposing or mishandling proprietary customer information” (35 percent, see below). When it comes to compliance and audit trails, 14 percent said their biggest concern was ensuring decisions are transparent and fully auditable, including meeting requirements for such regulation as GDPR and CCPA.
Despite the perceived importance to customers, only nine percent said their top concern was AI bias or fairness, “the risk of the AI making unfair or biased decisions against specific customer segments”.
Talking about AI trust and guardrails in response to the results, Ovenell said: “I would never recommend using sensitive information to train your LLM. Any sensitive information it does have access to in order to make decisions, you would ensure it isn’t retained after the interaction, so it is used only to answer questions and deliver the next step in a process, or pass off to a human agent… The trick is to give the LLM a small and well-defined role and if you need to do more for your business process, network and orchestrate multiple AI agents.”
Takeaway 3: Starting small is first step to getting CX automation right
As the saying goes, a journey of a thousand miles begins with a single step. And when it comes to how AI and automation are used for CX, many practitioners know the same applies.
We asked attendees, when you are just starting with CX automation, what is the most crucial first step to ensure you 'get it right'?
With 43 percent of the vote (see below), the most selected response was “start small, focus on automating one high-volume, low complexity use cases, for example order tracking”.
Securing executive buy-in was still a concern for many (36 percent), but when compared with similar data from CX Network’s annual research into the state of CX, the share of respondents was lower than anticipated. In 2025, when we asked our network members to name the top three challenges in their work only 10 percent of 353 respondents selected senior management buy-in.
Another surprise was the low number of attendees who see data clean ups as a crucial first step (selected by 14 percent). The least selected response was agent training (seven percent).
Ovenell said: “I'd say all four responses are relevant. Ultimately, start small and understand what outcome you're trying to measure. Understand what's happening today so that when you do deploy something, you can work out whether it's been successful or it hasn't been successful.
“Don't be frightened to fail. That's why you measure it and you understand it and you learn it,” Ovenell added.
He continued: “If you think about the way humans learn, we make mistakes all the time and we learn from those mistakes. So don't treat it as something that's got to be perfect. You know, AI is moving forward so fast and it's going to get to get better and better and better.”
Takeaway 4: AI will act as a true digital partner/advisor within three years
Asking delegates to look to the mid-term future, the last question put to attendees concerned their outlook for the next three years. We asked: “In the next three years, which phrase best describes how you and your customers will primarily interact with AI?”
Only 10 percent said they expect the customer relationship to remain “human-led”, yet 41 percent said they see a future where AI handles most routine issues and humans handle complex exceptions.
When it comes to the hybrid workforce of the future, 29 percent think AI will act “as a true digital partner/advisor”, providing autonomous decision-making and 19 percent said AI will primarily augment the human agent behind the scenes.
Ovenell said: “AI isn’t about replacing it’s about unlocking. What previously took 10 months to do you could do in potentially a few weeks because the technology is now there to drive that. Every day my team come to me with new use cases bult with this technology and it’s exciting to see.”
Lulla added: “Very often when I’m planning a trip [as a consumer], you can use AI to put together an entire itinerary and you would be amazed with the quality of the output. The usability has increased by such an extent… and it will be a integral part of life moving forward… as it is getting easier to use it will transform to a larger extent.”
You can catch up with the full session on CX Network+ via this link