AI in Customer Experience: A 2025 Update


AI in CX: The Comet Has Returned—And This Time, It Exploded

Back in January 2017, I wrote an article – AI in CX – likening artificial intelligence (AI) to a comet—repeatedly flaring into view with dazzling promise, only to fade into the cold darkness of unmet expectations. At the time, AI had delivered some real value, mostly in the form of automation and predictive analytics, but much of it still felt superficial. The tools were impressive, but not yet transformative. I concluded that while AI had made progress, the real revolution was still ahead.

That revolution arrived in November 2022.

When OpenAI released ChatGPT to the public, it marked a seismic shift in how businesses and consumers perceived and interacted with AI. For the first time, AI wasn’t just a backend tool for marketers or a silent assistant in your smartphone. It was front and center—conversational, creative, and shockingly capable. The AI comet didn’t just return; it crash landed.

From Automation to Augmentation

In 2017, AI in CX was largely about automation and using traditional statistical models for predictions. We saw gains in intelligent routing, basic chatbots, and predictive recommendations. These were (and still are) valuable, but they didn’t fundamentally change the ability to create content at scale for 1 to 1 personalization and to achieve better customer experience.

Fast forward to 2025, and we’re in a different world. AI is no longer just powering decision making at run-time, improving customer experience in digital and physical channels, and powering content production at design time. Large language models (LLMs) like GPT-4 and its successors are now embedded in customer service platforms, sales enablement tools, marketing engines, and in marketing content operations solutions. They don’t just deflect simple queries—they resolve complex issues through dynamic conversations, generate personalized content, and even detect customer sentiment in real time.

Consider the modern contact center. AI agents now handle the majority of customer interactions, not with rigid scripts, but with fluid, empathetic dialogue. They understand nuance, context, and intent. When escalation is needed, they summarize the issue for human agents, reducing resolution time and improving satisfaction. This isn’t superficial intelligence—it’s deeply integrated, high-value augmentation.

The Rise of Personalization at Scale

One of the most profound shifts since 2017 is the rise of personalization at scale – meaning having something unique and differentiated to say and show to each customer – and figuring that out and delivering it in real time. Back then, personalization meant inserting a customer’s name into an email or recommending products based on past purchases. Today, AI crafts entire customer journeys tailored to individual preferences, behaviors, and even emotional states. AI also detects sentiment, context, and intent – synthesizes these insights, and factors them into decisions – again all in real time.

LLMs analyze vast amounts of unstructured data—emails, chat logs, social media posts—to build dynamic customer profiles. These profiles inform every touchpoint, from marketing messages to service interactions. The result? Customers feel seen, understood, and valued.

This level of personalization was once the holy grail of CX – a theory in books. Now, it’s attainable in practice.

Conversational Interfaces: The New Front Door

In 2017, I pointed to early chatbot experiments and voice assistants as promising but limited. They often frustrated more than they helped. Today, conversational interfaces are the new front door to many brands.

Whether it’s a voice assistant embedded in a car, a chatbot on a website, or a virtual agent in a mobile app, customers increasingly prefer to “talk” to businesses. And thanks to LLMs, those conversations feel natural, helpful, and even enjoyable.

These interfaces are also multimodal—capable of understanding text, voice, images, and even video. A customer can snap a photo of a broken product, describe the issue in natural language, and receive a resolution—all without speaking to a human.

Trust, Transparency, and the Human Touch

Despite these advances, the human element remains critical. In fact, as AI becomes more capable, the need for trust and transparency grows. Customers want to know when they’re interacting with AI. When that experience is frustrating – taking too long – they want to be quickly connected to a human – and they want context to be remembered and factored in. They want explanations for decisions – especially AI-driven ones.

Leading brands are embracing this by designing hybrid experiences—AI handles the routine, while humans focus on empathy, quality control, creativity, and complex problem-solving. This synergy is where the real magic happens. Great brands get this, and have expert humans monitoring AI, and quickly course correcting it – always with the consumer’s experience in mind.

The Organizational Challenge

As I noted in 2017, the biggest barrier to AI adoption isn’t the technology—it’s the organization. That’s still true today. Many companies struggle to integrate AI into their workflows, data systems, and culture. They chase shiny tools without rethinking their processes, understanding the use cases, or upskilling their teams.

The winners in this new era are those who treat AI not as a bolt-on, but as a core capability. They invest in data governance, cross-functional collaboration, and continuous learning. They empower employees to work alongside AI, not fear it.

What’s Next?

So, where do we go from here – now that all of the AI children are growing up?

One of the next frontiers is emotional intelligence. AI is getting better at detecting and responding to human emotions. Sentiment analysis is evolving into affective computing—systems that can sense frustration, joy, or confusion and adapt accordingly. This will unlock new levels of empathy in digital interactions.

We’ll also see more proactive AI. Instead of waiting for customers to reach out, AI will anticipate needs and offer help before problems arise. Think of it as predictive CX—where the best service is the one you never needed to ask for.

And we’ll continue to see AI as a content production assistant, helping brands put more of the right content on the inventory shelf, and then using AI to help just-in-time assemble the right piece of content for the right person, and quickly test and learn which ones resonate.

And finally, we’ll see AI become more ethical and accountable. As regulations evolve and public scrutiny increases, businesses will need to ensure their AI systems are fair, explainable, and aligned with human values.

Conclusion: The Comet Has Become a Constant

In 2017, I ended with a prediction: the AI comet would return. It did—and this time, it’s not fading away into an AI winter – it’s here to stay. AI is no longer a novelty or a niche. It’s a foundational technology reshaping how we live, work, and connect.

For customer experience professionals, the message is clear: embrace the new normal. Invest in real intelligence, not superficial gimmicks. Focus on outcomes, not just outputs. And above all, remember that the goal of AI isn’t to replace humans—it’s to help us be more human to more people – at scale.

The comet has crash landed. Let’s build something extraordinary with the rare materials that came with it.

The Smart Choice: Licensing an Off-the-Shelf RTIM Engine

In today’s fast-paced business environment, companies are constantly seeking ways to enhance their customer engagement strategies. One critical component of this strategy is the Real-Time Interaction Management (RTIM) engine, also known as marketing technology decision-making engines, next-best-experience engines, or customer engagement engines. These engines play a pivotal role in delivering personalized customer experiences, optimizing marketing efforts, and driving revenue growth. However, businesses often face a dilemma: should they build an RTIM engine in-house, buy an inferior one and customize it, or license an off-the-shelf solution? This blog explores why licensing an off-the-shelf RTIM engine is the smartest choice for large enterprises.

Delivering Value to Customers

Business stakeholders responsible for customer engagement solutions prioritize delivering value to their customers. They aim to optimize customer experience efficiently to achieve high ROI. This means considering all costs associated with the solution, including opportunity costs, and evaluating the solution’s value in terms of revenue generation, customer satisfaction, loyalty, and lifetime value. An effective RTIM engine is essential for achieving these goals.

The Pitfalls of Building In-House

Building an RTIM engine in-house might seem like a viable option for some companies, but it comes with significant drawbacks:

  • Higher Costs: Developing an in-house solution often incurs higher costs due to the need for specialized skills, ongoing maintenance, and infrastructure1.
  • Longer Time to Market: Custom development can take significantly longer compared to purchasing an off-the-shelf product.
  • Complexity and Risk: In-house projects can be complex and risky, with potential for delays and cost overruns.
  • Lack of Expertise: Many IT departments may lack the necessary expertise in marketing technology, leading to suboptimal solutions.
  • Maintenance and Upgrades: Ongoing maintenance and upgrades can be resource-intensive and costly.
  • Integration Challenges: Integrating a custom-built solution with existing systems can be challenging and time-consuming.
  • Opportunity Cost: The time and resources spent on building an in-house solution could be better utilized on core business activities.
  • Scalability Issues: Custom solutions may struggle to scale effectively as the business grows.

The Advantages of Licensing an Off-the-Shelf Solution

Licensing an off-the-shelf RTIM engine offers numerous advantages that make it a more attractive option for large enterprises:

  • Specialized Expertise: Vendors provide specialized expertise that ensures the software meets industry standards and regulatory requirements. This expertise is often difficult to replicate in-house1.
  • Reduced Risk: Licensing software from a vendor reduces the risk associated with custom development, including potential delays and cost overruns. Vendors have established processes and support systems in place.
  • Vendor Support and Updates: Licensed software often comes with vendor support, regular updates, and maintenance. This ensures that the software remains secure, up-to-date, and compliant with evolving standards.
  • Cost-Effectiveness: For certain applications, licensing software can be more cost-effective than building a custom solution. This is especially true when considering the total cost of ownership, including maintenance and upgrades.
  • Integration and Modularity: Modern SaaS solutions offer modular architectures and APIs, making it easier to integrate with existing systems and customize as needed. This flexibility allows businesses to tailor the software to their specific needs without starting from scratch.
  • Scalability: Licensed software from vendors is often designed to scale effectively as the business grows. This ensures that the software can handle increased demand and complexity over time.
  • Innovation and Ecosystem Support: Vendors often have robust ecosystems that provide additional tools, services, and integrations. This ecosystem support can enhance the functionality of the licensed software and reduce redundant technical work1.
  • Speed of Implementation: Licensing software from a vendor allows for faster implementation compared to building a custom solution. This speed can be crucial for businesses looking to quickly address market demands and opportunities.
  • Focus on Core Business Activities: By licensing software, businesses can focus their resources on core activities rather than diverting them to software development. This can lead to better overall business performance and efficiency.
  • Compliance and Security: Vendors often have dedicated teams to ensure that their software complies with industry regulations and security standards. This reduces the burden on the business to manage these aspects internally.

Addressing Pain Points in Marketing and Sales Processes

Any RTIM solution must address key pain points related to improving marketing and sales processes, focusing on data integration, customer journey, and campaign optimization. Here are some of the critical areas that need attention:

  • Data Points for Training Models: Understanding the data points available for training models for MQL to SQL conversion, including events, sources, categories, segments, and nurturing activities, is essential.
  • Understanding Upsell and Customer Journey: It’s crucial to understand upsell opportunities and customer journey stages to improve upgrades, downgrades, and churn rates.
  • Challenges in Data and Process Flow: Companies often face challenges in documenting buying patterns and process flows, with significant delays in onboarding and disparate messaging across different channels.
  • Marketing and Sales Treatments: Effective customer engagement involves various treatments, including tickets, events, in-house production, trips, client promotions, emails, premiums, and charitable donations, all managed through Salesforce campaigns.
  • Churn Model and Customer Segmentation: A churn model built by the data science team can output monthly churn propensity for customers, helping to identify those who need specific treatments.
  • Guided Selling and Next Best Actions: Implementing guided selling with next best actions, using data from platforms like Snowflake, can enhance customer journey stages without relying solely on historical data.
  • Operational and Data Integration Challenges: Addressing operational challenges due to disparate actions between brand marketing and field marketing teams, and resolving data integration issues across various systems, is vital.
  • Enterprise Architecture and Marketing Automation: Integrating various marketing and sales systems into a cohesive enterprise architecture, with a focus on automating processes and improving data visibility, is necessary.
  • Customer Engagement and Dashboard: Documenting customer engagements and aggregating client-level data into platforms like Snowflake, with a dashboard to provide insights and guidance on effective treatments, can drive better decision-making.
  • Future Steps and Roadmap: Blueprinting solutions, conducting executive briefings, and scoping sessions to align on the marketing and sales automation roadmap are essential for successful implementation.

The Fine Line Between Customization and Restrictions

When deciding between customizing a packaged app or composing software components, businesses must walk a fine line. On one extreme, highly customizable software components may incur hidden costs, such as maintaining customizations and updating them. On the other hand, licensing components with restricted customizations may limit a company’s ability to differentiate itself from competitors. Striking the right balance is crucial for maximizing value and minimizing costs.

The Role of SaaS Solutions

Software-as-a-service (SaaS) has revolutionized the way businesses approach software customization. Leading commercial off-the-shelf RTIM solutions offer standardized architectures and robust ecosystems, including partners, communities, training, and educational services, to support their integration and use. This makes SaaS solutions a sustainable and cost-effective choice for many businesses.

Conclusion

In conclusion, licensing an off-the-shelf RTIM engine is a strategic choice for large enterprises looking to enhance their customer engagement strategies. The advantages of specialized expertise, reduced risk, vendor support, cost-effectiveness, integration and modularity, scalability, innovation, speed of implementation, focus on core business activities, and compliance and security make it a compelling option. By addressing key pain points in marketing and sales processes and leveraging the benefits of SaaS solutions, businesses can achieve their customer engagement goals more efficiently and effectively.

4 well-intended Marketing Automation BAD HABITS to break

Let’s face it.  No one sets out to botch something up or fall short of reaching a goal.  When marketing automation was in its infancy, and pioneers like Don Peppers, Martha Rogers, Tom Siebel, and Paul Greenberg envisioned marketing and CRM systems in the mid 1990’s, they set the right vision, believing great customer relationships could be initiated, fostered, and brought to scale with the right data and technology.  Essentially, their collective creed was:

  • Focus on the individual customer (e.g., be one-to-one and customer centric).
  • Manage the relationship by understanding customers’ buying cycles, needs, and behaviors across the marketing, sales, and service functions.
  • Use that knowledge to custom-tailor and personalize the experience.
  • Use technology to deal with the scale required by larger businesses.

Thirty years later, sadly, this vision still seems out of reach, or at best, only partially realized.  So why is that?  What’s held back the realization of the vision?  What are we still doing wrong?

Here are four unhealthy habits of nearly every marketer (so the good news is you’re not alone).  Fix these, and you’ll get a distinct advantage, and get closer to marketing optimization and CRM nirvana.

 

Bad habit #1 – Focusing on customer segments and not individuals

Customers are individuals.  Each has unique characteristics, nuances, and contextual needs that define who they really are.  And though we’re awash in a wealth of unique behavior data, it’s a common mistake to continue trekking on the beaten path, making decisions based on segment characteristics rather than individual ones.  For years, we’ve slotted customers into segments because we had no other choice, oversimplifying who they really are.

1 to 1 marketing automation

It’s understandable in the initial stages of relationship management that businesses make broad customer classifications such as:

  • Returning visitors
  • Mobile visitors by geography and device type
  • Registered users by gender and age (leading to segments like Millennials, Gen Zs, and Gen Alphas)
  • Non-responders to an email campaign

Yet after these customers repeatedly interact and transact, clearly stating their implicit and explicit preferences, continually handing over lifestyle and contextual data, there’s no excuse for still making generalized, segment-based decisions.  We’re spending millions collecting, storing, and refreshing specific behaviors and preferences, so we should use this data to drive individualized decisions and to customize treatments.

In a recent paper titled “Crossing the chasm: From campaigns to always-on marketing,” [i] Rob Walker and Matt Nolan contend that “building audiences using segmentation is a process that introduces severe challenges such as compromised relevance, unscaled labor, and collisions and conflicts.”  They go on to suggest using a next-best-action approach, describing it as one that “targets individual customers, rather than segments – leveraging their unique needs, preferences, and context.”

 

Bad habit #2 – Focusing on selling products instead of customers’ needs

Sounds crazy, right?   How else will we make money if we don’t sell products?

Still one of the cardinal sins holding back modern marketers is focusing strategy and tactics solely on selling products.  By doing this, we’re exasperating two problems:

  1. Product owners, incented to relentlessly push their products, bombard consumers with ill-conceived campaigns containing messages and offers that conflict, overlap, or worse, aren’t even applicable. When viewed through a customer’s lens, these promotions have little to do with their actual needs.  As such, marketers often completely miss the relevance mark.
  2. Even when a product fits, companies fail to provide well-timed promotions, convenient services, and a context-sensitive experience. Oblivious to the individual’s situation, they make company-focused timing and interaction decisions, such as blindly promoting a product simply because ad budget might otherwise expire, or failing to promote crucial services in conjunction with the product..  Consequently, tactics are entirely out-of-synch with the customer’s buying cycle and experience expectations.

Together, these problems compound customers’ negative brand perceptions.  Rather than providing a stellar buying service, well-intentioned marketers inadvertently (and increasingly) overwhelm, turn off, and tune out consumers.  Essentially blind to journey requirements, marketers miscalculate customers’ value calculus, timing preferences, and the overall interaction experience they need and expect.

In study after study (year after year), consumers and brands acknowledge these issues, both resoundingly stating their desire for solutions.  For example, in 2012 the Corporate Executive Board (now part of Gartner) surveyed more than 7000 consumers and 200 CMOs, finding that what consumers want from marketers is relevance and “simply, simplicity.”[ii]  That was 2012.  It’s 2018, and not much has changed.

If corporations keep strategy oriented on selling products, customer relationships will remain transactional and experiences sub-optimal for many more years.  Maybe we’ve forgotten what the R in CRM stands for.  It was put there to remind us that what matters most is long-term relationship building.  Our quest should be to unravel the mystery of a customer’s ever-changing needs, their journeys, and what drives their loyalty.  Our job is to use that knowledge to create custom-tailored experiences.

 

Bad habit #3 – Building channel-based versus coordinated intelligence

Shortly after September 11, 2001, the US government came to a stark realization that its various intelligence agencies were massively disjointed and compartmentalized.  This hadn’t happened overnight.  It was years in the making, and although for decades ample resources were poured into each agency, no one agency was responsible for coordination.  Attempting to solve this problem, the government established the Department of Homeland Security.

channel intelligenceIn a similar vein, some firms have built up marketing automation and CRM intelligence in silos for over 30 years.  In each channel (e.g., email, contact centers, web), they’ve poured substantial resources into projects aimed at beefing up customer intelligence.  Each channel amassing data, rules, and intelligence, but no one designated as the coordinator, and information rarely shared.  Subsequently, as more channels emerged, the problem grew larger. Today, many companies have 15 or more channels to manage, and no coordinating function.

To provide wonderful experience, brands need a function responsible for coordinated customer analytics, intelligence, and decision making, such as depicted in Figure 1.  Its role is straightforward:

  1. Collect interaction intelligence and contextual data from each touchpoint, and connect it directly to a system that can leverage that information immediately.
  2. Be brain-like, tracking behavior patterns in real-time, sensing needs, and using analytics to dynamically calculate value, comprehend preferences, and predict intent.
  3. Play the arbitrator, weighing an individual’s needs against corporate initiatives, policy, risk tolerance, budgets, and economic goals. Make instant and well-balanced decisions, track the results, and learn from each decision.

 

engagement hub

Figure 1: Engagement hub provides coordinated omnichannel intelligence

Think of this, not as another physical department, but instead as a virtual customer-centric hub. Designed from the ground up to be connected to all customer touchpoints, it’s journey oriented versus channel centric.  Cognizant of what transpired, why, and what’s best to do next, its embedded strategies and rules act as a real-time arbitration committee – making data-driven decisions in milliseconds versus months.

This hub is also more than a customer data platform.  It’s an end-to-end engagement hub responsible for not only gathering and coordinating intelligence, but also gleaning real-time insights and taking action.  To deliver on that, it manages key data, customer analytics, corporate rules and processes, and channel interfaces.  In a calculated and auditable fashion, it makes recommendations, delivers them to touchpoints (the channel apps fine tune the experience), and it learns from a systemic set of impressions and responses.

 

Bad habit #4 – Worrying primarily about marketing automation and technology, instead of experience

Automation, and the technology that enables it, efficiently repeats tasks.  That’s great, if you computerize the right tasks that deliver the right experience.  Look at it this way:  spammers are very effective at marketing automation.

Above all, to achieve lasting loyalty and build value, avoid the temptation to recklessly make existing marketing processes more efficient.  Granted, some existing tactics may work, yet chances are many need to be revamped (or ditched entirely), and recognizing that requires reframing priorities.  Preferably, focus on customer journeys, and ask if marketing tactics contribute to a better experience.  Consider journeys such as:

 

  • Prospects searching for products to discover and learn more
  • Customers seeking out trials to test those products
  • Customers embarking on a buying or upgrade process
  • Customers doing research on price, available incentives, and financing options
  • People filling out an application, making a booking, or redeeming rewards
  • Consumers getting stuck, struggling, or in need of assistance
  • Clients reaching milestones, entering new life stages, or affected by key events

No organization can serve its customers without supporting people.  To illustrate, assume your kiosk has a reasonable self-service experience, but then something goes wrong.  The technology hiccups, and a customer begins agitating.  Without back-up mechanisms, this situation can quickly turn disastrous.  To avoid it, you need reasonable levels of redundancy, well-tested cut-over processes, and intelligent detectors that gauge the need for human intervention, and then bring the right human into the loop.

Brands that thoughtfully consider these scenarios, elegantly weaving together marketing automation, people, and processes, will deliver better customer experience.

But how can you be sure you’re improving experience?  In short, hyper-focus on one journey at a time, pick metrics to measure each, and correspondingly measure overall satisfaction.  Once more, here’s where many firms trip up.  Instead of measuring whether the customer is fully satisfied with, say, the onboarding journey, they only measure certain tactics, like whether a welcome email got sufficient opens and clicks.

Conclusion

Be honest. We all have some bad habits that admittedly we should give up for our own good.  But breaking old habits isn’t easy.  And like any habits, we’re comfortable with our marketing automation traditions because the outcomes are predictable.  Nonetheless, just because they’re predictable, doesn’t mean they’re best for our customers.

When we force-fit customers into segments, push products on them that we want to sell, confuse them with conflicting and poorly orchestrated channel messages, and hyper-focus on our efficiency (versus their experience), the results will be predictable alright – in other words, we’ll get our anemic 0.5% response rates and slow growth.

If you think, however, you can do better, then take a chance.  Collect as much individualized data as you can, use it to personalize customers’ experiences, coordinate decisions with one principle engagement hub, and as Steve Jobs said, “…start with the customer experience and work backwards to the technology.”

[i] Crossing the chasm: From campaigns to always-on marketing, https://www1.pega.com/insights/resources/crossing-chasm-campaigns-always-marketing , December 2017

[ii] CEB Press Release, https://news.cebglobal.com/press-releases, May 2012