Home > Columns > Executive Interviews

NiCE Executive Interview

Click image below to view the ebook

Inside the AI Shift: A Conversation with Kevin Lee, CTO & Key Pursuits Leader at NiCE

Artificial intelligence has moved from hype to practical application, but organizations still struggle to separate meaningful impact from noise. In this interview, Sheri Greenhaus speaks with Kevin Lee, CTO and Key Pursuits Leader at NiCE, about the realities of AI adoption, why so many pilots fail, how data unlocks real value, and what companies must do to ensure success.

Across the conversation, Kevin offers candid insights into the evolving CX landscape, the role of composite AI, the importance of analytics, and how NiCE helps organizations move from “admiring the problem” to actually solving it.

Sheri Greenhaus: I wanted to start with a little bit of personal background so people can get to know you. This is a relatively new role for you, CTO and Key Pursuits Leader. What exactly does “Key Pursuits” mean?

Kevin Lee: It is a bit interesting, right? I have this intersection of a CTO role and function and then the key pursuits leader. And that's because my intersection is between product and our customers. So while somebody might say, hey, what about sales or that's pre‑sales, really my charter is to do the things for our clients: care about how we actually drive impact and efficacy for our clients.

For practitioners and operators, it's really cool that it's a model or an LLM or whatever. But when it comes down to brass tacks, they care about reducing handle time, improving NPS and CSAT, and reducing cost to serve. Those are the things they care about. My role focuses on finding that sweet spot between the application of the technology and the results — from the practitioner lens while understanding the technology.

Sheri Greenhaus: So Key Pursuits is really about working with customers.  It is understanding what they need and how we can help them, rather than pursuing new technologies or filling gaps in the product line.

Kevin Lee: Yep. We have a large R&D and product organization that’s always innovating. My role is the intersection of customers and technology. When our teams build new products and bring frontier models into the platform, my group focuses on helping customers and partners understand how to use those capabilities to their fullest and get the results they’re looking for.

Sheri Greenhaus: Okay, so previous to this role you headed digital sales, correct?

Kevin Lee: Correct. I led the digital go‑to‑market organization which encompasses the sales organization, the pre‑sales organization, and everything in between. Anything that helped customers understand our digital capabilities and how to utilize them was inside my organization.

Six years ago, a lot of things changed. We acquired five organizations in that time. One benefit is that we were always ahead of the curve, looking at what’s in front of us. When we look at the competitive landscape, we were always the leaders.

For example, we acquired MindTouch, the top‑right leader in knowledge management. At the time, people asked, “You’re a CCaaS or WFO/WFM organization, why do you need knowledge management?” But we were looking at where the puck was going. Knowledge is foundational to how you leverage LLMs.

So acquisitions like Moxie, MindTouch, ContactEngine, and others, culminating inside Cognigy, were part of that strategy. We grew a large business, and I left that in the capable hands of the agentic team last year when I took on this new position.

Sheri Greenhaus: What are some of the things you learned in that position that you brought into this one?

Kevin Lee: Understanding the core technology, both its capabilities and limitations, has translated well. Practitioners care about KPIs. To get those KPIs while leveraging technology effectively, you need to know what the technology can and can’t do.

It’s also helped me scale that capability across the organization. We have sellers, pre‑sales engineers, architects. Part of my job is helping them shift the conversation away from technical specifics and toward understanding the customer: What are they trying to accomplish? What will move the needle?

We have the most complete platform in CX. That means we can choose the right tool for the right job. Not everything needs an expensive frontier model. For example, “Where’s my order?” is a straightforward query. You don’t need to spend $5 in tokens. Deterministic AI can integrate with backend systems and return the status at negligible cost.

Because we have a full portfolio, we can choose the right approach instead of forcing everything through an LLM.

Sheri Greenhaus: My son-in-law leads AI for a large organization, and he says the biggest challenge now is telling people, “You don’t need AI for that.” People want to use AI for everything simply because it’s available.

Kevin Lee: Absolutely. It used to be “don’t touch AI.” Now everyone feels they must use AI somewhere. It’s become the panacea du jour.

But it’s like using Siri to ask what four times three is. You spend a dollar in tokens, wait longer, and maybe get the wrong answer. We’re applying the wrong tool to the wrong problem.

And yes, it’s pervasive across every industry, not just CX.

Sheri Greenhaus: At CCW, everyone says they “do AI.” Two and a half years ago, people were afraid AI would take their jobs. Then they realized it could help with wrap‑up and summarization. Now they’re asking, “What else can it do to make my life easier?” Acceptance is high, but the challenge is finding the right use cases that save time instead of costing more.

Kevin Lee: I totally agree. It’s the quintessential pendulum swing. We went from “absolutely don’t want to use AI” to aggressive adoption without being thoughtful. But people are starting to understand it. Auto‑summary is now table stakes.

We’re also doing a better job of understanding capabilities and limitations. As purveyors of the technology, it’s our job to help customers understand what AI does well, what it doesn’t, and where automation versus augmentation makes sense.

In certain industries, you’ll never automate the entire interaction. Augmentation — taking cognitive load off human agents — is often the right solution. Disney World is high‑touch and high‑empathy. A life insurance carrier speaking with a beneficiary needs human empathy. But “Where’s my order?” can be automated all day long.

The world has shifted. At CCW, you saw the move from selling features to selling outcomes. To navigate that, we need to help customers understand what outcomes are achievable with the right technology. And again, it’s not always an LLM. Sometimes deterministic AI is the better answer.

Sometimes it’s as simple as surfacing the right knowledge at the right moment. You don’t need a lifelike avatar to tell someone the return policy. Just show them the policy. That prevents escalation to a $10 human‑led call.

The most elegant solution isn’t always the most technologically heavy solution.

Sheri Greenhaus: Do you feel your group works closely with sales, consulting, and pre‑sales to dig into the customer’s needs? Helping guide customers toward the right use cases and building confidence that you’re truly their partner?

Kevin Lee: You saw the explosion of AI vendors at CCW. Three years ago, the number of sponsors was small. Now it’s thousands. Everyone is suddenly an AI company, CCaaS, WFM, WFO — all rebranding themselves.

Why? Because enterprise technology used to be hard to deploy and extract value from. Vendors said, “We’ll make sure it doesn’t become shelfware. We’ll do the integrations ourselves.” That shifted the landscape.

But helping organizations understand where the real impact is matters more than ever.

Right now, there’s a big shift toward throwing FTEs at the problem. “We’ll give you bodies who will apply the technology.” But then they still ask the customer, “What do you think we should automate?” Consulting used to step in, whiteboarding, journey mapping, and they’d come back with the usual suspects: “Where’s my order? Where’s my refund? Change my address.”

Those are low‑ROI use cases. They’re easy, but they’re not expensive interactions.

What we do differently is grounded in data and analytics. Nexidia, and the work John Willcutts pioneered, is what sets us apart.

We can process 100% of interactions at scale — voice, digital, everything. We sift through the signal and the noise. Yes, we find the simple calls. But we also find the complex ones: “I want to set up a payment plan,” multi‑threaded, multi‑intent conversations. Those cost brands $10–$20 each and happen far more often.

If you’re a COO or an experience leader, you care about the expensive calls that happen frequently — not the easy ones.

Because of our analytics, we can identify those high‑impact opportunities. Then we pair that with our CX‑specific AI models to understand what’s happening. We come back to the practitioner and say:

“Here are the opportunities inside your organization that will drive the value you’re looking for.”

And now, because we also have FTEs, we can combine awareness + people + the complete portfolio and actually solve the problem.

That’s the winning equation.

Sheri Greenhaus: NiCE always says their LLMs are built for CX, so they hallucinate less. But it sounds like the real differentiator is how you apply the LLMs.

Kevin Lee: One hundred percent. There are thousands of AI vendors now, all using the same frontier models — Gemini, OpenAI, Anthropic, etc. We’re all cooking with the same ingredients. But there’s a difference between a Michelin‑star chef and a street vendor.

Sheri Greenhaus: It’s how you put it together.

Kevin Lee: Exactly. And we have no desire to build our own LLM. That’s an arms race to zero. What matters is how you harness the models, the frameworks you build around them, and how you expose them to consumers, agents, and operators.

Knowledge is a great example. Tight integration with knowledge repositories transforms an experience from “talking about the problem” to actually solving it. Because of our complete portfolio, ERP, CRM, OMS integrations, we can move from admiring the problem to fixing it.

NiCE Labs evaluates every model, every day. New models come out constantly, each a few basis points better than the last. Some are great for summarization, where speed doesn’t matter. Others are great for real‑time agent augmentation, where three milliseconds can make the difference between an awkward pause and a fluid conversation.

Different verticals need different models. Hyatt needs something different than GM. WFM needs different models than virtual agents.  NiCE Labs evaluates all of that.

Our IP is two things:

1. The completeness of the portfolio, the right tool for the right job.
2. Our deep understanding of interactions, what it takes to go from “hi” to “bye” with the highest satisfaction and lowest effort.

Sheri Greenhaus: Do you find that when you bring in a brand‑new customer, they usually start with a pilot? Forrester said a huge percentage of pilots never roll out to production. What do you attribute that to?

Kevin Lee: Gartner says 95% of pilots fail within the first year. Number one, they chose the wrong vendor. Number two, they didn’t start with the data.

Organizations have the answers to the test. They have the map. But they get enamored with the shiny thing, the LLM, and are drawn to point‑solution vendors who say, “We’ll throw warm bodies at this and brute‑force the problem.”

Historically, companies couldn’t process all interactions at scale. Now we can, 100% of interactions. That unlocks the ability to understand what’s really going on.

The next issue is relying on CRM ticket data. Agents disposition tickets as fast as possible. CRMs tell you a case happened, but not what happened. They don’t capture the substance or texture of the interaction.

We do. That sets us apart.

The third issue is applying AI in a way that only talks about the problem. They build an FAQ, tie an LLM to it, and end up with a system that tells the customer how to find their order status instead of actually retrieving it.

Without backend integrations, AI can’t solve the hard cases, the ones stuck in a delivery loop. Those still escalate to a human.

So, pilots fail because:

1. Wrong vendor
2. Not starting with real interaction data
3. Lack of backend integrations
4. Forcing GenAI where deterministic AI or knowledge is better

Everything old is new again — knowledge and data remain underutilized.

Sheri Greenhaus: Most companies are dabbling in AI. What are three things you would advise to help ensure the success of their AI program?

Kevin Lee:

First: data.

Second: upskill your teams. Too often, companies hand an SOW to a GSI and say, “Go build this.” But the difference between mediocre and highly successful programs is having an internal center of excellence — people who understand both the business and the technology.

Third: tight interlock with the executive layer. AI is not the goal, outcomes are. Handle time reduction, satisfaction improvement, self‑service, those are the metrics that matter.

These three things separate the haves from the have‑nots.

Sheri Greenhaus: Is there anything I didn’t ask that you think our audience should know about NiCE?

Kevin Lee: If I were to be self‑serving, I’d highlight our Automated Insights capability. Historically, analytics efforts cost half a million dollars and required teams of consultants listening to calls and taking notes. Some vendors try to use LLMs, but LLMs alone can’t ingest all interactions or understand everything.

We can. We process hundreds of thousands, even millions of interactions and deliver an executive readout. What used to cost half a million dollars is now effectively free.

We tell customers:

• What to automate
• What to augment
• What to shift to self‑service
• Where training improvements could save millions

Start with the data. We provide it.

When it comes down to it, if NiCE cannot help an organization accomplish its KPIs, we are useless. We just happen to do it with technology.

The competitive landscape is noisy; everyone sounds the same. Customers ask, “What’s different?” And while it’s not the sexiest answer, the differentiator is data.

I don’t want artificial intelligence. I want actual intelligence. Actual intelligence comes from analytics. Somewhere in your interactions, “good” already exists. If we can find those great interactions — high satisfaction, no callbacks, short handle time — we can replicate them.

That’s how we build prompts. We analyze what the best agents do and build the LLM prompt around that.

It seems obvious, but at scale, it’s transformative.

Closing

AI adoption is accelerating, but success requires more than enthusiasm. As Kevin Lee emphasizes throughout this conversation, organizations must start with data, choose the right tools for the right problems, build internal expertise, and stay focused on outcomes,  not technology for technology’s sake.