What we learned at All Access: Future of CX

CX Network’s last webinar series of 2025 was All Access: Future of CX. Speakers from United Airlines, Air Charter Service (ACS), NiCE, BankUnited, Talkdesk, Dialpad, IHG Hotels and Resorts, Comcast, JustCall, Gamma, and Amazon joined us to discuss the future of customer experience. We covered a lot, including artificial intelligence (AI) and hyperpersonalization, agentic AI and customer communication, data challenges of the future, and more. 

If you missed the series live, you can catch all sessions on-demand on CX Network+ here, or read on to learn the key takeaways from each session…

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. 

Learn More

Personalization is becoming harder, and data-sharing hinges on transparency and customer control

Our series kicked off with a panel discussion between CX Network Advisory Board member and head of Innovation Lab at United Airlines, Jorie Sax, and global head of client experience at Air Charter Service, Oliver Priestman. 

Priestman’s target audience is highly niche, as opposed to Sax’s target audience: the general population. Both, however, share the same goal of improving experiences for customers through personalization.

Personalization requires customers to share their data with businesses but, as CX Network research found, this can be very challenging, with just under half of CX practitioners agreeing or strongly agreeing that customers refuse to share the correct personal data with retailers. 

Sax said: “Customers are smart enough to understand that they’re not going to get a personalized experience unless they share something about them” but will only do this if they trust the business requesting this data. 

To gain customer trust, Sax said: “The most important word is ‘transparency’. The second is ‘control’”. Priestman agreed, noting “[Air Charter Service is] very open about what we’re using that information for… it's about personalizing their experience”.

Watch the whole discussion on-demand here.

AI-first CX is becoming the dominant service model – and 2026 will be the tipping point

“Customer experience is entering a new era”, began Heather Hughes in her session, “and it is AI-first”. 

Hughes, director of product marketing at NiCE, joined the series to discuss the advent of AI-first CX models, and how agentic AI fits into this. “Seventy-two percent of customers say that have experienced benefits with AI”, Hughes said. “And 69 percent of consumers globally said they trust companies as much or more when using AI.” 

Agentic AI, she explained, has reset the limits of what is possible in service, as it can “think, act, and resolve in real time, detecting intent and anticipating needs before customers ask”. In this, AI will collapse front-office and back-office silos as AI agents can operate across an entire enterprise. 

She turned to the example of Lufthansa, a NiCE customer, which introduced an AI agent that helps customers find and book alternative flights when faced with travel disruption. Now, she said, Lufthansa’s AI agents are handling more than 16 million conversations per year. 

On the issue of securing C-suite buy-in, Hughes recommended prioritizing AI observability that “tracks in real time where AI is working, how it is performing and the value it is creating”. Observability, Hughes explained, answers the following questions: 

1. Are we reducing effort?
2. Are we improving efficiency?
3. Are we delivering a return on investment (ROI)?

Watch the full presentation on-demand here.

BankUnited went “all-in” on AI to fix CX – increasing customer and agent satisfaction and improving operational efficiencies

Jeiner Morales, senior vice president and director of data analytics and business systems at BankUnited, was joined by Dan Goldstein, strategic customer success manager at Talkdesk, to discuss how BankUnited transformed CX with an “all-in” approach to AI.

Morales started at the beginning of BankUnited’s AI journey, saying “our journey did not start with AI, it started with a real need to modernize. We were using an extremely outdated platform and our KPIs were not where they needed to be… our abandonment rate was higher than what we were comfortable with”. 

Fragmented systems caused frustration among agents and created a more “transactional service” model, rather than a relationship-based approach. 

To fix this, BankUnited launched a full AI suite, rather than piloting individual tools: “We launched autopilot, Copilot, interaction analysis, QM assist, workforce management – all of it on day one”, Morales explained. 

The transition was surprisingly smooth and the tools drove immediate operational improvements and increased agent satisfaction. “On day one, our agents were smiling… and within a few weeks, our KPIs stabilized and our abandonment rate dropped below our target – and stayed there”, Morales said. Real-time coaching, call summarization and analytics massively improved agent performance, too, reducing escalations and increasing confidence.

Ultimately, Morales said, “agents are spending more time listening and understanding, not just processing requests.”

When asked for advice for the audience, Morales said: “My biggest advice is don’t be afraid. You need to start somewhere. There is a cost to not experimenting and a cost to waiting that we usually don’t talk about. Don’t let perfect be the enemy of good – you don’t need perfect data to get started”.

Learn more about this case study by watching the session on-demand here.

Agentic AI will rise to the challenge of increasing customer demands for instant service and delivery

When CX Network asked network members which three customer behaviors influenced 2025 planning the most, there was a clear winner: customers’ expectations for instant service and delivery. This was the point that John Work, head of sales engineering at Dialpad EMEA, framed his session around. “Historically,” he said, “the goal was simply to get an answer. But today, customers are demanding immediate resolution and action.”

Legacy bots have created a “bot purgatory” with static answers and rigid flows, ultimately leading to customer frustration. 

“Legacy chatbots are often just FAQ librarians in disguise”, said Work. “They can retrieve static information, but can’t perform an actual task or transact any business.”

Agentic AI, however, is goal-driven, executing tasks and adapting as they go. “Agentic AI is not just the next iteration of the chatbot,” said Work, “it’s fundamentally different”. Rather than being transferred to a live agent, an AI agent will complete the task the customer needs doing.

Early adopters are seeing measurable results, with containment “jumping from 25 percent to 60 or 70 percent, because now the AI can handle actions”. “Average handle time can be cut in half and we’re seeing increases in customer satisfaction, often resulting in millions of dollars of savings”, Work explained. 

When it comes to implementation, Work cautioned that governance must come first: “The main mistake is rushing to full autonomy without building governance first. Start small, start with oversight and gradually scale up the autonomy”. 

Change management, too, is an essential consideration for those seeking agentic AI. Skipping this can be a deal-breaker for colleagues internally. “You have to establish the role of the human agent, redefine expectations and deliverables”, Work advised.  “What information will they receive from the AI? How will they solve problems together? These are the points you need to address”. 

Watch the full presentation on-demand here.

AI will fundamentally change CX – and maybe everything else!

Hitesh Patel, global digital director at IHG Hotels & Resorts, frames AI not as another tech cycle but as a fundamental leap forward that is reshaping business at speed – equivalent to the emergence of the internet, but even faster:

“The internet brought information to us. AI is going to bring intelligence to us,” he said. 

Organizations will soon operate with both human employees and digital agents, massively expanding operational capacity. Digital agents will, Patel said, likely outnumber employees in the future.

To meet this oncoming wave of innovation, businesses must shift from data-centric to AI-centric operating models. Patel stressed that data is still key, but that it must be put to use with AI now, to stay ahead of the curve. This will require deep changes in most organizations, with Patel warning that “AI centricity doesn’t come with deploying a tool – it comes with structural change”. 

Leaders need to adopt agile mindsets as AI “is not a technology like we have seen before. It is evolving at rapid pace”, Patel continued. “You need partners working in the same direction. Your business will be completely different in two to three years,” he added.

Leaders also should ignore the temptingly snappy “95 percent of AI projects fail” headlines that block progress, urged Patel. 

“That 95 percent report went viral because of fear. It is either fear or hype that goes viral”, he said. 

Traditional businesses were advised to seek quick wins while simultaneously planning two to three years ahead – iterating now while building foundations for the future.

“You need an unstructured data strategy and a structured data strategy”, Patel advised. “everyone else in your sector is thinking two to three years ahead, so you need to as well.”

Patel echoed Work’s earlier sentiment that autonomous agents require extreme care. Excitement around autonomy must be tempered with discipline and safety.

“Autonomy needs to be carefully thought through. Are you carrying the right brand values? There must be the right safeguards and guardrails,” he warned.

Testing is essential, Patel advised, saying: “You have to R&D everything – you don’t know what that experience will look like until you test it live”. Workflows will need to be redesigned, he warned, “I would never look at current workflow as the answer for autonomous agents. Workflows will get re-done as it is a different parameter entirely”. 

We hit upon the advent of AI avatars, which once seemed so futuristic. Patel has seen cutting-edge prototypes, convincing him that the shift towards them is imminent. “Avatars bridge the gap,” he said. “They look like us, talk like us. The experience layer completely switches – you’re not talking to a chatbot anymore”. It is not quite curtains for human-led customer support, though. Human empathy remains the key differentiator which is currently irreplaceable. “As much as I talk about AI, I’m still a ‘human-and-AI’ person. Our greatest strength will always be emotional intelligence,” Patel said. 

Watch the full session on-demand here.

Comcast elevates personalization and reduces contact center traffic with generative AI and predictive modelling

Comcast, the largest home internet and second-largest cable TV provider in the United States, sought to ultimately minimize expensive call center traffic by solving issues in automated channels. 

Seshendra Balla, senior manager of data and machine learning at Comcast, joined the All Access: Future of CX series to discuss this.

“We have been customer-obsessed for 10 years,” he said. “Any customer calls we do are very expensive, so the main goal was to reduce the IVR cost”. 

Using a RAG-powered chatbot, the internet and TV giant has managed to do just that, while increasing agent and customer satisfaction. Large-scale personalization is driven by age, location, and weather, past behaviour, and device signals, with troubleshooting flows and chatbot behavior tailored to each individual customer. “Initially we tried to implement the same solution for all customers, but that won’t fly anymore”, Balla said.

If customers do end up speaking with agents, this experience is personalized, too. Comcast uses five to six hundred data points per customer to power predictive scoring that can identify churn, upsell, cross-sell, and in-store recommendations, guiding agents to deliver personalized information to customers. 

“As many as 200 to 300 models run for every customer, every day,” Balla explained.

These predictive insights give agents a major advantage as they now have access to context for every customer conversation. For example, “there are times when the customer is not aware that there is a package that will help them reduce their bill… that has helped a lot”. Ultimately, these efforts have combined to produce a “huge uptick” in customer satisfaction.

When asked about the future of AI in CX, Balla said: “Maybe in the next two to three years, you will not be able to differentiate between a human and a chatbot. Responses will change based on customers’ moods and you’ll start talking to more chatbots in a real-time fashion, with more emotions.” 

Watch the full session on-demand here.

AI is no longer a “nice-to-have”, but it will amplify not replace agents 

CX teams, we all know, are under pressure to do more with less. Rising volumes and expectations, coupled with flat staffing levels, are pushing teams more and more towards automation. “Manual processes are breaking under pressure. Operational costs and hiring constrains persist,” said Deren Rehr-Davies, senior vice president of sales at JustCall. Because of this, AI agents are no longer optional, but core to business strategy. 

Early AI deployments have been famously hit-and-miss, due to a lack of autonomy, context and the ability to complete tasks; AI agents do not have these problems. The future of AI in CX, said Rehr-Davies, is one that “completes, not just converses. 

“The shift is from AI that talks to AI that walks”. CX architecture, he explained, is moving towards “connected AIs – not siloed bots or siloed agents”. 

After-hours, overflow and troubleshooting are the fastest CX automation wins as use cases that consistently deliver value. “Overflow is low-hanging fruit – areas where you’re already not staffing a huma,” Rehr-Davies said. But teams that see the best results, he explained, are those that are “experimenting, testing workflows, and collecting customer feedback”. 

Rehr-Davies then moved on to a prescient question facing all business leaders currently: will AI replace agents? “I really think the answer to that question is no”, he said. “It’s going to make us all more efficient and help us add more value, but there will still be a place for agents, CX organizations and humans.” 

Watch the full session on-demand here.

CX transformation requires a mindset shift

Slow, linear transformation models are now incompatible with the pace of CX innovation. “If you take 18 months to deploy a solution, by that time it may already be legacy”, said Nick Farinha, head of CX solutions design and delivery at Gamma. PQ

“Your CX roadmap is never complete – it’s continuous improvement”. 

Many organizations modernize the front-end without transforming the foundations. “Oftentimes, all [of a business’s] customer and order data is in five disparate systems that don’t talk to each other. They’ve got modernized technology sitting on a legacy architecture stack,” Farinha explained. 

Two of the biggest blocks to successful – and timely – transformation, are poor data visibility and lack of leadership alignment, said Owen Davies, CX strategy and transformation advisor at Gamma. 

“Shortfalls in data, or poor data, or the lack of it, can be one of the biggest things getting in the way”, he explained. “They don’t have the granularity of reporting to understand why customers are contacting them and, without real insightful data, you’re really operating in the dark”. 

Similarly, lack of internal alignment will stall progress. “There’s often an overemphasis on short-term CX metrics such as CSAT or cost reduction” said Davies. “CX transformation is often seen as a project rather than a mindset, and that’s dangerous. Building best-in-class capabilities requires sustained investment and a real change in culture.”

The future of AI in CX, argued Farinha, is enterprise-wide, not contact-center only: 

“We need to move from contact center tools to enterprise-wide toolkits”, he said. “Why can’t the warehouse or back-office teams use the same AI systems? CX and AI are not competing, and AI amplifies CX when deployed holistically”. 

When asked for advise audience members seeking to transform CX with AI implementation, data quality and ongoing governance were highlighted as the two most important principles to ensure sustainability. 

“Garbage in, garbage out”, Nick said. “There is no point investing in AI if you can’t access your data”. On AI governance, he said: “AI solutions are not one-and-done – they need constant monitoring and re-engineering. You need to know when your AI is doing something it shouldn’t be – and act quickly”. 

Watch the full discussion on-demand here.

Amazon deploys generative and agentic AI to prioritize transparency, speed and fairness in seller experiences

“Seller experience in high-stakes industries,” began Manav Kapoor, senior product manager for selling partner trust and store integrity at Amazon, “is supposed to be transparent, speedy and fair. Building a CX which is explainable, seamless and makes sure that tough decision are gently put forward is important”. 

Traditional machine learning (ML) has failed because fraud constantly evolves: “The fraud and bad-actor space is very dynamic. It is never static. Building scalable solutions which remain effective for X number of years is unheard of in traditional, ML-specific solutions”. Generative AI and agentic AI have opened up new possibilities in the fraud detection and prevention space in marketplaces.

“Generative AI and agentic AI… make it more proactive in terms of detection”, he said. “They kind of learn themselves and pick up very quickly where solutions are dropping the ball, evolving with new information which you’re feeding into these models.”

Explainability, however, is the key in trust and fraud, especially when AI is involved. “Natural language rationales are more personalized and provide more chain-of-thought reasoning. This transforms opaque, black-box decision into more transparent, humane and traceable reasoning.” Great CX in enforcement must be two-way, Kapoor said. “We’re open to feedback. It’s not a one-way street. We want to hear when we’re not doing things right and fix them with adequate evidence. Sellers can understand decisions and challenge if needed.” 

Critically, fairness depends on fast communication and human oversight. Systems – and the people working them – must assume they will sometimes be wrong, and design for recovery. 

“We’re working with someone’s livelihood, so we do extra due diligence. Notifications have to be fast, precise, clear and open adequate appeals pathways”, said Kapoor. 

“Human oversight and weekly audits make sure we identify false positives quickly”. AI must earn trust before it earns autonomy: “We want precision rates as high as 95 percent before moving to full automation”, explained Kapoor. “And even then, models need constant watering and learning from new data.”

Watch the full session on-demand here.

All Access: Future of CX returns on December 1, 2026

Quick links

• How the big brands are revolutionizing CX design and journey management
• 7 Takeaways on customer insights and data analytics
• A guide to making the most out of customer insights