12 Takeaways on the AI revolution in CX
What we learned from All Access: The AI Revolution in CX 2025
Add bookmarkListen to this content
Audio conversion provided by OpenAI

When we polled our audience at All Access: The AI Revolution in CX 2025, a whopping 93 percent said that they believe that using AI as part of their CX journey will be critical to their company's success. With seemingly endless opportunities to improve CX with AI, including increased personalization, more efficient customer service and automation of back-end operations, we sat down with experts from around the world and across industries to talk all things AI in CX.
Don't miss any news, updates or insider tips from CX Network by getting them delivered to your inbox. Sign up to our newsletter and join our community of experts.
If you missed the event, read on for key takeaways. You can also watch all full sessions on demand here:
Inspiring customers with generative AI at Swiss Airlines
The event opened with Edward Pauls, product owner for digital adoption and innovation at Swiss Airlines, unveiling Heidi, the airline’s AI-powered virtual assistant travel guide.
Heidi serves as a state-of-the-art travel guide with personalized travel recommendations that align with business targets based on what the user inputs. For instance, for someone hoping to travel with their family, Heidi might recommend Barcelona. On the other hand, a solo traveller seeking to travel to the UK might receive a recommendation to visit Manchester, an underserved airport, rather than the typical (and well-served!) choice, London. Heidi is a sophisticated system, operating in real-time and with access to contextual information, such as weather forecasts.
Edward highlighted Heidi’s rapid development process, which prioritized user trust and safety through content sanitization, alignment with Swiss’ values and mitigation against hallucinations. There were challenges along the way, including positioning Swiss Airlines as more than an airline: as a comprehensive travel inspiration source, especially against competitor platforms such as ChatGPT.
In the future, Swiss is going to include greater personalization by integrating Lufthansa Group data, allowing Heidi to tailor suggestions for customers based on previous behavior and travel destinations. The team are also seeking to include pricing integrations and more human-like interactions soon.
How a unified AI strategy improves CSAT and boosts NPS
Lead series sponsors, Coveo, highlighted that $3.7 trillion in global annual revenue is at risk because of poor customer experiences. Patrick Martin, VP of CX at Coveo, emphasized an urgent need for effortless interactions to minimize customer effort and improve CX.
Patrick contrasted traditional, siloed approaches to support with Coveo’s AI-driven, unified approach in which customers intuitively move across different touchpoints. Rather than having to navigate multiple content repositories, AI-powered solutions provide direct answers to customer queries.
Coveo’s platform serves as a bridge between content and digital experience by implementing generative AI. The closed-loop system leverages behavioral analytics to continuously improve personalization and relevance for customers, addressing the challenges posed by fragmented channels and native search.
To demonstrate the effectiveness of the approach, Patrick pointed to case studies such as Xero’s and SAP Concur’s. Xero achieved 21 percent better deflection while SAP Concur realized €8 million in saved revenue. Not only does the approach improve traditional CX metrics such as customer satisfaction (CSAT) and Net Promoter Score (NPS), but it significantly reduces customer effort scores.
Autonomous agents to transform contact centers
Sundar Raghavan, VP, Intelligent Business Applications, Microsoft and Jon Rastia, global director, Innovation, Argano, contrasted the eagerness of companies to adopt AI with the general unreadiness for implementation across industries. They suggested that addressing this starts with seeing AI beyond its cost-savings potential and focusing equally on how it can improve CSAT and, ultimately, generative revenue.
Covering key implementation strategies, the speakers discussed aligning AI with business objectives and focusing on high-impact use cases such as chatbots, predictive analytics and proactively resolving customer issues. They also emphasized the importance of robust data foundations and change management. Exploring how to action this within contact centers, the speakers floated the idea of holding AI workshops with seasoned agents to capture tribal knowledge, levelling the playing field for new hires and helping to reduce average handle time.
AI’s role is expanding from assistant to decision-maker, enabling more personalized customer interactions and reducing customer effort scores. AI-powered recommendations can even help to turn contact centers into revenue generators.
But how will AI agents work with humans? Jon and Sundar predicted that AI agents would significantly reduce human toil, delivering faster, more relevant answers and improving overall support experiences while ensuring compliance with regulations such as GDPR and CCPA.
Reducing risk and building trust when implementing AI
While AI implementation has compelling benefits, many projects do not go beyond proof of concept because of concerns around accuracy, privacy and bias. The session with Cyara focused on AI governance strategy that inspires confidence and trust while reducing risks.
Andrew Smith, AI practice lead at Kenway Consulting, suggested a five-step AI governance framework:
1. Plan
2. Build securely
3. Test continuously
4. Release responsibly
5. Monitor and improve
Cyara’s AI Trust as a Service offering similarly implements continuous testing and monitoring throughout the AI lifecycle.
The panel, led by Janet Vito, SVP, marketing at Cyara and also including Rishi Rana, CEO at Cyara and Sean Rabago, senior service expert and capability lead at Kenway Consulting, emphasized the need for empirical, adversarial and stress testing along with benchmarking to ensure the reliability of AI initiatives. Sean discussed the concept of AI drift, explaining how AI models can deteriorate over time, making robust governance systems even more critical.
The panel advised adopting clear accountability structures for when AI systems make mistakes, warning that customers blame brands, not algorithms, for poor experiences.
Lessons learned from launching Agent Assist
Liran Meir Frenkel, Head of Product Marketing at SmartReach and Boris Grinshpun, VP of Product Management at LiveVox by NICE, joined the fourth session of the day to discuss their journey launching Agent Assist, an AI-driven tool transforming contact cemters.
Agent Assist’s key features include call transcription, next best actions, alerts and sentiment analysis – all in real time. The tool’s versatility allows automation of processes, scaling top-performers approaches and helping regulated industries minimize risk while maximizing productivity.
One of the key pieces of advice was customizing based on agent experience, with newer agents requiring more guidance than more seasoned counterparts. The speakers also advocated for prompt engineering, multi-parameter validation and keyword detection to prevent model hallucination.
Liran and Boris touched on integrations with other automations, such as Robotic Process Automation (RPA). They explained how Agent Assist is adaptable and can auto-populate ticketing systems, streamlining customer service processes.
Addressing data security concerns, the speakers introduced the concept of a “walled garden” approach to data sharing, emphasizing the importance of data masking and ownership for sensitive information.
Maintaining service quality when implementing AI
Dominik Olejko, CX expert and top 25 AI in CX influencer, joined Chad Anderson, vendor manager of Customer Care and Roadside Operations at Mercedes-Benz USA in a panel discussion to discuss the lessons Chad and his team have learned from their AI implementation journey so far, and how audience members in companies of all sizes can apply these lessons to their own AI strategies.
Chad shared Mercedes-Benz’s success in launching Interactive Personal Assistants and an agent assist pilot, citing the significantly improved handle times and CSAT rates by providing real-time documentation assistance and knowledge-based suggestions. Chad also noted that AI is being used to address persistent issues such as ghost calls and appointment scheduling challenges while expanding omnichannel presence through generative AI agents.
Both panelists were keen to emphasize the importance of balancing the efficiency of AI with maintaining a top-notch customer experience, especially for luxury brands such as Mercedes-Benz.
Culture change was also a focus of the session. Mercedes-Benz USA holds weekly generative AI meetups to foster ongoing education and engagement. Dominik highlighted the need for transparent communication about potential role changes and carefully managed change management to ensure smooth adoption of AI.
AI literacy assessment
Connie Hwong, generative AI communications lead at Siemens, joined the series to discuss COMM.AIT project, an AI literacy assessment developed in partnership with Leipzig University and the Academic Society for Management Communication. The initiative aims to enhance skills across the company’s global communications team of 1,400.
COMM.AIT evaluates three crucial dimensions:
1. Knowledge of AI technologies
2. Practical skills such as prompt engineering
3. Attitudes toward AI
The assessment showed high understanding of AI risks and opportunities but also revealed hesitation in practical implementation. To address this, Siemens adopted a ‘Triple E’ approach: Experience, Exploration and Exchange. The HR department developed AI learning modules and bi-weekly calls to discuss AI applications.
In the future Siemens and Leipzig University plan to make COMM.AIT publicly available, fostering industry-wide dialogue around AI readiness, skills and attitudes.
Building and scaling an AI-powered virtual assistant
Jarno Koponen, product design manager at Zalando, led the charge implementing an AI-powered shopping assistant, revolutionizing personal lifestyle discovery on the Zalando app and website by leveraging natural language processing (NLP) and Large Language Models (LLMs) to reduce cognitive load and enhance engagement.
It started with a hypothesis that an intuitive, NLP-based discovery method would improve CSAT. The assistant, now fully functional, shows deep contextual understanding, handling complex queries such as outfit suggestions for themed parties and budget-friendly lifestyle products. It can also cater to specific cultural contexts, such as Berlin or Stockholm fashions.
The development process involved a multi-discipliniary taskforce including product designers, data scientists and engineers, and it focused on five key elements:
1. Identifying customer problems.
2. Forming a dedicated team.
3. Building a scalable platform.
4. Mapping customer journeys.
5. Prioritizing visual engagement
In mapping customer journeys, the team distinguished between exploratory browsing and specific buying intentions. These insights informed the assistant’s proactive product recommendation feature.
Zalando partnered with a well-known LLM provider and also developed internal tooling to assess and improve performance. This approach has encouraged deeper dialogues between customers and the assistant, which learns more about their personal tastes and requirements.
In the future, Jarno envisions a more personalized and dynamic fashion and lifestyle discovery experience that transcends text-based interactions.
AI beyond chatbots: Exploring its full potential
Gerald ‘Ro’ Sinclair, contact center product specialist, began his session evaluating the factors behind the rapid growth of AI in 2025, highlighting advancements in NLP, machine learning and predictive analytics as key catalysts.
These technological improvements have enhanced AI’s ability to personalize customer interactions. Customer expectations for seamless and fast experiences are also driving the AI adoption rate among businesses eager to stay ahead of the curve.
AI’s role goes beyond chatbots to include automation of repetitive tasks, real-time agent support, hyper-personalized product recommendations and even fraud detection. When CX Network polled the audience, 43 percent said AI will have the greatest impact on customer service automation over the next three years. More than 20 percent answered predictive analytics for customer behavior, 19 percent said personalized customer interactions, nine percent said workforce engagement and agent support and five percent said fraud detection and security.
Cut handle time and boost CSAT with AI
Rahim Teja, account executive for Western Canada sat down with Martin Lampman, director of customer support operations at British Columbia Lottery Corporation (BCLC) to discuss BCLC’s use of AI has vastly improved contact center operations by addressing high call volumes while reducing costs and enhancing multi-channel CX.
BCLC’s AI chatbot has achieved a 30 percent call deflection rate and is now used in quality assurance, assessing interactions with customers. Text analytics, automated categorization and CRM integration has enabled better journey mapping, highlighting negative sentiment and allowing for proactive issue resolution across channels.
This data-stitching approach encourages collaboration between the contact center and other teams to enhance responsible gambling initiatives, and informed discussions on product alignment and revenue loss prevention, elevating the contact center’s strategic role.
As AI evolves, BCLC is exploring agentic AI to replace traditional bot flows with autonomous, decision-making systems. This advancement promises highly personalized customer interactions based on historical individual behavior.
The future of AI agents in the contact center
Devon Mychal, senior director of product marketing at Cresta, joined the series to dive into the transformative potential of advanced AI agents in customer service. Mychal traced the evolution from frustrating flow-based systems to today's sophisticated AI, powered by a multimodal architecture combining large language models with cutting-edge transcription and text-to-speech technologies.
Devon showed the audience the power of an AI agent, demonstrating live Cresta’s agent’s ability to handle interruptions, express empathy and make complex decisions, allowing it to manage interactions from basic inquiries to troubleshooting and account management.
Devon recommended a strategic implementation approach, advising that businesses begin with a simple use case, such as feedback collection, before advancing to more complex use cases such as customer retention. While AI capabilities are constantly improving, Devon noted that human-AI collaboration is still key to success, with supervisors playing a key role in quality assurance, high-stakes interactions and training.
Like other speakers, Devon acknowledged the risks associated with AI, such as hallucinations and data security, emphasizing the need for ethical deployment and robust governance.
Using AI in neuromarketing and customer insights
Roger Dooley, author, marketing expert and Forbes columnist, traced the evolution of persuasion techniques from the 1924 Psychology of Persuasion to Robert Cialdini’s 1984 Influencer, which introduced six principles including social proof and reciprocation. He then explained the emergence of neuromarketing in 2000 and introduced the concept of neuromarketing 2.0, which integrates AI with neuroscience and behavioral science to unlock customer insights.
His presentation looked at various well-known AI models, demonstrating their applications in customer service, content creation and feedback analysis. Roger ultimately showed how AI is a persuasion partner, able to generate marketing and CX strategies.
He highlighted real-world case studies, such as that of the University of Chicago’s application simplification process and a UK tax collection experiment leveraging social proof.
Throughout the session, Roger revisited the importance of ethical considerations in AI-powered customer engagement, referencing Cialdini’s focus on ethical persuasion and Zig Ziglar’s approach to sales integrity, suggesting that professionals can leverage AI to gain publicly available insights from experts while maintaining high ethical standards.
All Access: AI + Data 2025

Join us to hear from industry leaders, innovators, and CX experts as they share insights, strategies, and tools for harnessing customer feedback to drive meaningful change.
Register Now