Where AI really fits into digital customer experience tech

It's one thing to buy into a shiny-object agentic AI platform. It's quite another to get all the data pieces ready and build the needed IT plumbing to realize the technology's potential gains for efficiency, customer satisfaction and bottom-line gains for customer experience.

Calvin Cheng is a partner at West Monroe, a global consultancy. Specializing in digital strategy, CRM, customer experience and e-commerce, he has a frontline view of the tech challenges that prevent better digital customer experiences -- and what the barriers are to AI fulfilling its promises of improving them. His clients reside in various verticals, including retail, manufacturing, hospitality and leisure, financial services, healthcare, private equity and more. We discussed his perspective on AI for CX and a whole lot more.

Editor's note: This Q&A was edited for clarity and brevity.

What is the typical way in which clients engage you?

Calvin Cheng: Our entry points are very different, but they all meet in the middle. When speaking with business leaders, marketing leaders and revenue officers, generally the conversation is, 'Calvin, how can you help us increase revenue, increase our customers and grow market share? How do I have better customer retention? How do I maximize customer lifetime value and build the right experiences to retain and grow those customer relationships, and turn those customers into advocates so they are extending our brand on our behalf?'

For conversations with CIOs or members of their teams -- sometimes the chief digital officer as well -- the question becomes, 'What are the right technology components? With the pace of change of all things technology, what are the right technology components that we should consider or invest in so that we can be modern and provide the best customer experience?'

What are the most common gaps that need to be filled before you can provide AI-powered digital customer experiences?

Calvin Cheng

Cheng: The most common is siloed data, which is nonintegrated, disparate applications containing snippets and portions of the customer experience.

A conversation I'm preparing with a large pharma client is precisely this. We're talking with their head of data. His ask from the business is to provide a customer 360-degree view across sales, marketing and service so that the customers can get a great experience no matter which touchpoint. They're finding that their email automation platform has some level of customer data, their CRM has another slice of data, their ERP has order transaction data, and their customer service and customer success has yet another slice of the data. So, like many organizations, each department has its own technology stack.

Often, the technology stack reflects the organizational structure, and it's one of the areas where there is a challenge. The CFO, looking at revenue and customer lifetime, value and growth, sees some stagnation, but can't see why that's happening. Or the chief marketing officer doesn't understand why there's so much churn within the customer base. Same thing with sales. Oftentimes it's symptomatic of these application data teams' silos throughout the organization.

Where does AI come in? Just turn it on and everything will be fine?

Cheng: I wish it was that easy. The good news, bad news and good news is that West Monroe and many of our consulting competitors are getting a lot of interest in how to turn it on. What we are finding is, our conversations can start with what I would call the 'new shiny object in town.'

Then, when we start talking about how to bring it to life, it's a crawl-walk-run model where we have to think about what use cases will deliver the most value, not only for the enterprise, but for its customers. Then, we think about the corresponding data sets that would need to be analyzed and automated to pass the workflows through the various application stacks and through the processes of your organization to deliver that experience.

We hear from vendors, analysts and tech buyers that CEOs are pushing AI because they have fear of missing out (FOMO), but they don't quite understand the data governance and IT plumbing that needs to take place before turning on AI. Do you see that?

Cheng: In those situations, we take a product mindset approach, where we want to show that the CEO or C-suite person wants to get on the AI bandwagon because they've got FOMO, like most executives do. There is credibility in the sense that the need, the desire, the expectation -- that is there.

Generally, we'll do a rapid diagnostic and understand what data they have access to, then we'll build a rapid prototype for what that use case can or could be. Will it be the most elegant? No, but it is a way to prove that use case in a rudimentary way, to show a modicum of progress toward achieving what the art of the possible could be in the full-fledged version of that GenAI use case that the C-suite is so infatuated with. There's a wide range of GenAI large language models [LLMs] that we can rapidly prototype and ingest existing data sources from either the client or publicly available to prove the idea. It won't be 100% production-ready, but the idea is, we can at least dip our toe and show some early proof points for how that could come to life if greater investment is awarded.

What do you think is the future of AI agents?

Cheng: I believe that agentic AI for creative brainstorming -- to generate new or derivative content -- will continue to evolve and get smarter. That story, as crazy as it sounds, is well established in its trajectory in the past three years.

What I find fascinating is the potential of the reasoning engine. There are already LLMs, a lot of data algorithms to help ingest, analyze and recommend. Now there's potential combination within agentic AI that will instantiate new workflows and deliver and execute actions based on that reasoning.

But therein lies another set of challenges that GenAI has continued to evolve on, which is, at what point in time do you trust the model without the human in the loop? In some use cases, if it's work that's highly repetitive, maybe low risk to customer experience, low risk to the enterprise, low risk to the brand, you can begin to take bolder and bolder steps.

Are most organizations -- not talking about your customers or early adopters -- two years behind the vendors as far as AI and all these new agentic AI features are concerned?

Cheng: The short answer is yes, especially with -- dare I say -- luminaries and icons like Marc Benioff promising the world with all things agentic AI.

The longer answer is that it's still up to each respective user, team, organization and enterprise to find the right recipe for turning it on in the most valuable way that meets their needs and their customers' needs, and it's not going to be cookie-cutter.

The beauty and messiness of bringing this to life is understanding what are the right ingredients, the right components, the right data sets and the proper use cases that can deliver the most value for those constituents inside and outside of the organization. We're finding that there's a lot of excitement. We're getting pulled into opportunities and, in some cases, speaking truth to power. We see and support where you want to go, but we'll need to take a minute to breathe, pause and take a hard look at what data, what assets, what capability you have inside your organization so that we can incrementally, progressively bring those use cases to life.

Don Fluckinger is a senior news writer for Informa TechTarget. He covers customer experience, digital experience management and end-user computing. Got a tip? Email him.