Scaling Agentic Automation With Open Architecture - with Arun Chandra of NICE

Welcome everyone to the Emerj AI in Business podcast. Today's guest is Arun Chandra, Chief Operating Officer at NiCE Cognigy. NiCE Cognigy provides cloud-native customer experience software and contact center services to enterprises. Arun discusses why most AI pilots never make it to full scale production and what closes that gap. Clean, well-organized data, clear context about decision rights, and systems connected enough for agents to act without a human in the loop. Today's episode is sponsored by NiCE Cognigy.

In this episode, we cover why so many AI pilots stall before reaching full scale production and what closes the gap. To go deeper on this topic and learn how leading organizations approach AI investment more like a venture portfolio, and why interdisciplinary collaboration is critical to defining the right data for AI success, download our free PDF report beginning with AI at Emerj dot com slash AIK1. That's E M E R J dot com slash AIK1 to download your copy.

Now, the conversation with Arun.

Arun, welcome to the Emerj AI in Business podcast.

Hey, thanks, Yolandi. It's a pleasure to be here.

I love having guests come from, uh, the same organization or from the same network. And I want to pick up right where full left things when he was on the show earlier in this series. So he made the case that C X was the first place where agentic A I had to really earn its keep, not just demo well, but perform well, right? So, uh, uh, and the numbers, since then, kind of back that up, but in a very strange way.

So I saw a very interesting stat saying that something like 80% of enterprise apps now ship with an agent built in. That is up with a third from just two years ago, but only about a quarter of teams have gotten one channel fully into production. When you look at that gap between shipping an agent and running one for real, where does the next phase of this go?

So, no, so I think that's a very pertinent question because there's this valley of death between AI pilots and POCs and actually full-scale deployment at the enterprise scale, if you would. So, two things maybe I'll talk about. One, why do we have a valley of death in the sort of like, not, you know, you know what I'm saying? It's more centered around why do pilots don't end up scaling across the enterprise? And then the second is where does this thing go next?

So, in terms of where it's going to go next, I think the ability for the AI agents to handle more complex workflows is going to continue to increase. The cap capacity, the intelligence on tap, the ability for us to be able to drive more complex workflows. It's only gated by the imagination of the people implementing it. The capacity is going to be available. I think it's gonna go into where the future will be more of the human agents and AI agents coexisting in a more harmonious way because every organization is going to need some up mix of both.

So with that, you know, for us for it to be able to fulfill its promise, and that kind of ties back into why we have this, uh, gap between pilots and production. It's in order to do that, there's a few things that have to happen. Number one is the availability of data, knowledge, and even context. And that's sort of where operationally most of these things fail is because once you get past the pilot, you're trying to figure out if I want enterprise-wide scale, my data has to be in a good spot, it has to be clean.

More importantly, or as importantly, my knowledge has to be organized and available because that knowledge and and data is the lifeblood of what AI can produce for you in terms of outcomes. And then the third related thing is context also. So context has to be pretty well provided so that the A I can understand who reports to whom, what decision rights do I have versus what decision rights you have, all of those things become the lifeblood.

And I think that is where the work has to be done. And then the second part, which I'm, you know, which is being addressed also, is ensuring that the A I agents have connectivity into all the backend systems, the systems of record, so that they can take actions autonomously without having to in involve a human being. So I think that's sort of where things are at and where things are going.

I like that. And and we're going to get deeper into that entire part of, you know, getting your A I to connect with with everything in the back end. But so with this transition of getting to a point where we use A I as part of our daily workflows, where does the transition stall out most? Is it is it technical problems or is it more about how the organization is is being set up? And I think thinking about what you just mentioned about the availability of data and knowledge in context, it sounds like it's more an organizational problem.

Well, it can be, and it most likely is. And maybe I'll take a I'll step back and take the broadest lens at it, if you don't mind. The broadest lens is to ensure that the leadership team that is trying to drive the transformation using A I is very clear, first and foremost, about what strategic and financial objectives you're trying to solve for. Most of the time, what has happened in the last, at least twelve to eighteen months that I've seen is people rush into a pilot or a P O C because you've got all these different vendors calling on you or your C-suite is telling you that what have you done with A I?

When are you going to do something with A I? So you kind of rush into a pilot. I think step one for me to get sustainable outcomes is to be very clear. Are we driving towards higher revenue? Are we driving towards lower cost? Are we driving towards both, perhaps? So having that clarity would be the first. What then what that then does is to basically it feeds into the next part of the conversation is in order to accomplish that value, what are the most critical or most impactful use cases that where we can go to start automating?

So that with that comes that clarity, and then you kind of feed into it and saying, okay, so therefore, in order to drive that, what data, what knowledge, what context, what backend systems have to be available? And then how do I measure my R O I on an ongoing basis? In some ways, what I'm describing to you is sort of like classic operations management one on one, but applied to A I's implementation because it is more complex than what we have done in the past.

That kind of leads me exactly to where I wanted this next question to land as well because our listeners might not have stalked you as much on LinkedIn as I have. But, listener, if you go and look at Arun's LinkedIn profile, you would see that he, he's kind of the right person to have this conversation with because he has been on both sides of the line. Something that I also read about is that, ah, about two thirds of C X teams say that they're running some sort of agentic A I pilot this year or they've already started running it, but barely a quarter have gotten even one channel all the way into production.

So we're still in that gap, we're we're still not solving it. Um, and that's a huge chunk of the industry stuck between that pilot and real deployment. And like I just mentioned, you've been on both sides of this. So, when enterprises cross from pilot to production, what changes operationally then, uh, after the pilot? And and what does scaling really require from those who need it?

I so let me introduce uh another phrase here in terms of why things fail or succeed. I think that is also a a much, you know, important phrase, which is change management of the people involved. So making sure that we are paying a lot of attention through that transformation journey or automation journey to ensure what is the new role of humans in the process. Who needs to change their what needs to be different in terms of handoffs?

Where do handoffs disappear and they're fully automated? How do we reskill the workforce of the future? And and and this applies not only to a C X, but this applies to all of A I. It applies to engineering and product management because that is also a huge use case for A I, if you would, which I know is not the primary topic of our conversation today. In my role, I'm also responsible for the internal A I transformation of the company, of Nice A I for Nice, basically, and our our ambition is to make sure that A I becomes a new operating system for Nice as a company.

So, I think that entire change management, workforce management, trying to understand what we do, how our workforce will evolve, becomes an important component of being able to scale it. Having said that, there are customers like we have a customer called Henkels, it's a global consumer brand. And they have sort of been very rigorous in the in terms of the approach that we just that I've been outlining in terms of approaching this strategically and operationally. And they have twenty five A I agents that are spanning eleven countries and seven channels, and they're all geared towards, you know, delivering exceptional experiences, but in multiple languages and geography.

So, I guess that's one example of where things can scale across geographies and channels at scale. But I think it requires the kind of disciplined execution that I was talking about earlier.

So, exactly that. If we're going to hammer on the the disciplined execution, i if we're thinking about your customer that you just mentioned that had to scale it and they had to make it work. And obviously, it sounds like, uh, the transition went well or is going well. What are they doing differently? Why is this a success story?

Maybe maybe I'll use my own example at Disney. I spent four years at Disney, ah, one of the best, uh, leading consumer brands in the world, uh, most iconic brands in the world. And I was, uh, responsible for modernizing the customer experience function for the streaming part of the company. The streaming business is just to give you a sense of scale, one hundred, 180 plus million subscribers, 110 geographies, twenty five, thirty languages, if you would.

So great big scale, not to drop too many numbers. And the goal was to be able to a. A I was a component of it, but the entire focus was how do I modernize the C X function completely end to end. The objective so it I think it start my journey started by first and foremost understanding the business number one. Number two, I think most importantly is not approaching the C X function as a functional leader, but approaching it as a general manager.

That says, what is the business value we can create from this function versus saying, hey, I manage a contact center, how do I reduce my handle time? How is the? So there are people who've been running contact centers who are now being asked to kind of apply A I to transform the function, but they haven't seen anything else per se. So approaching it as a general manager, establishing a broad vision that we did at Disney, that this is going to be not just a cost center, it's going to be a value creating center in terms of being able to do cross-sell, up-sell.

And, of course, we have to deliver exceptional experiences at the lowest possible cost to serve. So I believe that one of the things that really helped us in our transformation of modernizing the C X function at Disney was having a very clear vision of where we want to go, having very clear objectives behind it as to what we're going to accomplish, how much cross-sell up-sell will be delivered, what will be the C-SAT scores, what will be the cost structure.

And ultimately, my goal was to be able to bend the cost curve. So that means as subscribers are growing, the holy grail is that your cost structure is kind of flattening out. So for me, I think the difference is clarity of vision, very clear execution plan, and then over a multiple long period of time, multiple months, disciplined execution that follows after that and course corrections because you will learn so much in this process.

Absolutely. And, and obviously, the term value, I think that's where most of our listeners go, oh, wait, I need to, I need to really pay attention here. One of the biggest hurdles that I've heard of is the C F O and convincing the C F O, uh, with the matrix. So how do you measure R O I on this once it's live? Where's the number and and how? What is the number that will convince the C F O?

I think the numbers, there's only two numbers that convince CFOs. Number one is, are you being accretive to the top line? Are you helping drive revenue? Okay, can I measure that? And the second, which is more the more favorite, not not to oversimplify the CFO's point of view, but the more favorite favorite option is where is the hard cost savings? What have you done for me? Because you're asking for investment. Let's take the, for, you know, one thing we can maybe segue into is the cost of tokens.

A I is not free. We're talking about, hey, A I can automate something, and so we assume that we've reduced the cost of the humans. But wait a minute, the A I agents have their own operating cost, which is in token cost. And there's a whole discussion and debate we can have about where the token cost is going. So therefore, I think ultimately for a CFO, you've got to show the case. Am I driving top line?

Am I reducing cost? And therefore, I'm impact. It's almost as simple as that from a P and L perspective. And actually, one little data point I will point to you is which is why you are hearing all of these articles in the industry where Uber said, Hey, I spent my entire token budget in a matter of months versus the full year. Uh, you're hearing people pulling back on some A I investments, people are hiring again. And I think that is all driven by the fact.

I believe out of the S P 500, the vast majority, ninety five, 97% of the companies have yet to show a documented R O I in their P and L based on A I. It's just that that is that that is the fact today. So people are still trying to comprehend where this thing shows up in value. And I therefore, for me, success will depend on being very clear about what the value, where the value is being accrued, how are you measuring it and how are you tracking it?

That'll get you the investments you need from the CFO.

Yeah, so uh, I guess it's it's safe to say that value means something different for every part of the company. And as long as we know what value means to the C F O, that conversation is just gonna gonna be easier. I mentioned earlier in the conversation that we're going to get into the unification or unifying these different platforms and, uh, getting our our heads wrapped around that as well. So I've been reading up on this open protocol stuff, M C P specifically, and it's wild how fast it's moved, right?

It went from an anthropic project to something like 40% of software companies saying that they're already running it in some kind of production or in some kind of way. And by the end of last year, Anthropic had handed it off entirely to a vendor neutral foundation, so no single company owns it. So, when a leader is staring down that bolt versus open decision for their own stack, what matters most in deciding how to scale agentic automation across the enterprise?

Uh, I think I think that's a a very, very, very relevant question, um, especially when you're looking at, uh, you know, let's just take some regulated industries like financial industry, healthcare industry. I think, uh, you know, there there becomes a conversation to be had as to first of all, what does your what does your organization or your industry require? And if it's, you know, in ensuring that you can use the open protocols, that the open protocols that you're leveraging.

I mean, first of all, let me back up and say, I'm a huge believer in open systems. And I think the future is future is in open interconnectivity. And I mean, there's going to be a huge pressure on avoiding vendor lock. It's always been the case, but SaaS companies developed a vendor lock. And I think A I is trying to democratize that and saying, I can open up everything for you. So, I'm a huge believer in that.

I think it'll continue. I'm also a huge believer separately on the proliferation of open source L L M s that will come. And that, by the way, will drive down the cost of tokens across the industry, without question. The third thing I would say is the fact that make sure that you fully comprehend your enterprise architecture as you're making decisions about where to deploy open systems, where to deploy proprietary systems. Make sure that your biggest workflows will support that, make sure the interconnectivity is there, the scaling will happen.

But for large organizations, which is where a big a big part of our focus lies and the companies like Disney or Toyota or DHL, who are other customers of ours. The number one thing that becomes important is privacy, compliance, legal requirements, et cetera, et cetera. If you can't, if you can't manage that part of it, then the architecture just doesn't work for anybody in the company. Because you're very, you know, you're extremely sensitive to that kind of risk, if you would.

So, from my point of view, open systems are great, M C P is great. We are using it internally and externally in our business, if you would, we're offering choice to our customers. Number two, make sure that your regulatory requirements, your legal compliance are covered through that open architecture. And make sure that you are able to interconnect across the organization using M C P for your own use cases.

Yeah, so I'm thinking that this, uh, this idea is is definitely catching fine. You always get those vendors that would label anything and everything to what is trending. So, um, if if a customer wanted to know, is this really interconnectivity or interoperability? What is the question that a buyer should ask about interoperability that most people don't think to ask to really determine whether they're it's just a vendor saying it's open and whether it's genuine?

Um, I think just, uh, you know, sh short answer on this one would be make sure you fully comprehend your internal architecture, your needs, your use cases, and ask d pointed questions about what you need in your environment and your what is your needs to ensure that you don't accept the high-level hand-waving answer you get from a uh from a, from a seller and a provider.

Yeah, so it comes back to partly, but back to the readiness of the organization, but knowing your organization, knowing your business problem, knowing what you solve for, is that right?

100% and where you're going strategically. And that's sort of where the some of these, like, like we've said, is you said the CFO understands the P and L language. Honestly, I think at the end of the day, even the CEO, that's the language that, uh, ultimately matters when they when you go into an earnings call and you say, I invested, uh, $500 million in A I tokens, the street is expecting you to articulate, okay, so great, what did you get in return?

So how much did that save us?

Exactly. I mean, or what, you know, so ultimately, I think if you don't connect it to the P and L at a at a scale, pilot wise, you want to spend, you know, half $1 million, $1 million in a sandbox playing with things, just absolutely go for it. You want to do a company-wide rollout in a large organization, and if it doesn't connect to the P and L in some way, it has to be.

Like, even if you say that, hey, my customer satisfaction has gone through the roof, it's gone from 90% to 99%. Okay, great. Somebody's going to ask, okay, what does that translate into?

Yeah. Yeah, what's the dollar amount of the cut customer satisfaction?

That, I mean, that I'm a I'm a huge believer in that because, uh, that's how you get operational support to go drive things and investment that you need to go drive things. If you can show, you know, what value you're generating and they are very crystal clear about what that is.

Oh, this is interesting. Alright, um, we're heading towards the the closing part of this conversation and I want to end on something practical. And I'm gonna throw in another a Gardner number in there. So Gardner put, uh, put a number on the risk of getting this wrong. So they're predicting that this year, about a third of companies will end up damaging their own customer experience by rolling out A I too early. So that takes us back to the readiness theme that we have been hammering on quite a lot today.

And so, for a leader who is convinced that agentic automation matters within the organization and it's needed, but they haven't started yet, the stakes are obviously very high and the first move will matter. So, what is the most effective place to begin inside their own C X organization?

So there's two schools of thought, and I'll give you my point of view. One school of thought is in the industry is pick an easy use case or two or three simple use cases, you experiment with them, you can, because they're, you can get your arms around it, you can build momentum by showing success. That's one school of thought. And, of course, the problem statement is contained, the number of interconnected systems is fewer, data is, uh, the data cleansing is fewer, less than.

The other school of thought is take the hardest problem, take a really complex use case that is most meaningful to the company, build a coalition across the organization that we are going to go tackle this big problem, make sure you have got executive support for it. And then you go tackle that because in doing so, you will unsurface or uncover, you will surface all the problems that you will face down the road when you're scaling. So you hit it head-on, you see the gaps, you see the needs early on, and if you can overcome them, your path to scale across the enterprise is so much, you know, is so much faster and smoother, if you would.

I think an organization has to decide on their own which is a better path for them because it's very much a personal choice in some ways. We have customers like TripAdvisor who have gone on the record and said, we are gonna pick the hardest use case. And they have been very successful with that with us. I think it becomes a combination of the organization picking the path. We recommend pick a difficult use case because that's the fastest path to scale.

And second, I do think it matters with which partner you pick or set of partners you pick to work with on your journey, because they will guide you the right way.

It comes down to to risk appetite, and it's very cliche, but you know, you need to risk it to get the biscuit. So it sounds like a lot of that matters as well. Uh, Arun, thank you. It's been great having you. I'm going to give you, uh, the opportunity to leave our audience with a final thought. If there is one thing that you want them to take away from this conversation today, one thing that gets stuck into their minds and and they cannot stop thinking about it for the rest of the day, what would that be?

The amount of potential benefits that A I can bring to us is unimaginable. It is only going to get better. It is the it is a culmination of technologies that we've been developing for the past, you know, thirty plus years. And, you know, having this intelligence on tap and harnessing it for the benefit of our organizations and our people both is absolutely imperative. And I think it's highly doable if you take the approach that we've been talking about.

And but my point of view, Yolande, is I'm an operator, so I think of it in terms of operational terms. But I think the possibilities are enormous. We can do this gracefully and for the benefit of all of our team members.

Absolutely, Arun. It's been a great conversation about not just C X, but A I transformation and unification and interoperability and leadership. A lot of themes in here. Uh, I think this is a great conversation, one that I will probably listen to more than once. Thank you so much for sharing so much with us today. It's it's been great having you here.

My pleasure. Thanks for having me.

Wrapping up today's episode, let's look at the three key takeaways from our conversation with Arun. First, closing the gap between A I pilots and full production starts with getting data, knowledge, and organizational context in order before scaling. Second, sustainable A I investment requires leadership clarity on whether the goal is higher revenue, lower cost, or both before a single use case is chosen. And finally, disciplined execution and deliberate change management, not the technology itself, determine whether automation scales across an enterprise.

In this episode, we covered why so many A I pilots stall before reaching full-scale production and what closes that gap. To go deeper on this topic and learn how leading organizations approach A I investment more like a venture portfolio, and why interdisciplinary collaboration is critical to defining the right data for A I success, download our free P D F report beginning with A I at Emerj dot com slash A I K one. That's E M E R J dot com slash A I K one to download your copy.

On behalf of the team at Emerj, we'll see you on the next episode.