Paula Rivera: 

Welcome to the AI Factor, where business meets ai. I’m your host, Paula Rivera, and in today’s CX Explained episode, we are diving into the 2025 Opus research conversational AI Intelliview report with the one and only Derek Top, Principal Analyst and Research Director of Opus Research. We’ll explore how the conversational AI landscape is evolving from flow-based bots to gen AI-driven agents, platform maturity, and what really separates leaders from laggards. If you are evaluating a provider or just trying to keep up, this is the episode for you, Derek, welcome. 

Derek Top: 

Thank you. Good to be here. 

Paula Rivera: 

Great to have you. Before we dive on into the heart of your report, which is a great report, I’ve spent a fair amount of time with it. I have to ask yesterday, the White House released its winning the Race America’s AI action plan. And according to CNN, the plan has three pillars. Accelerating innovation, of course. Building out AI infrastructure in the US, and making America hardware and software the standard platform for AI innovations built around the world. Have you had a chance to review the plan yet and do you have any initial reactions? 

Derek Top: 

Yeah, well, I’d say I’ve reviewed it. I’ve not actually read it word for word all the way through. I know it came through yesterday. And yeah, I think it is a very interesting take on where we’re going with AI, both as a country from the US and also from a global perspective. So it is from in line with a lot of those industry are hoping for in terms of accelerating the pace of adoption, using AI as a strategic asset, and really of what we as I’m based in the US and based in California, actually in the Silicon Valley. In fact, there is a lot of hope for what’s next in terms of how US can be a leader in the AI race, if you will. 

Paula Rivera: 

I concur completely, and I think I was of the same mindset as most of the leaders in the room yesterday thinking this is all very positive and definitely needed. 

Derek Top: 

Yeah, well, I shouldn’t say all very positive. I mean, because there is some questionable aspects to it, but in terms of there was some push towards making streamlining regulations and making data centers built faster, and then using that as a way to build software and export those AI software to the world in terms of their semiconductor manufacturing. That is positive in terms of that level of innovation and that level growth. It also looks at some of the potential issues around open source models and open models, which again, that also is typically thought of as working towards a more common objective and a common goal. 

There was some language in there about the moving training models that use, quote-unquote, around bias and forcing… Or not regulations. [inaudible 00:02:51]. It’s not even legislative is really more of just a recommendation that federal agencies can only use models that don’t use the bias in their training, which I think there is some potential for some challenges to that. But at the same time, I think the goals around making it easier for tech companies to develop and test and deploy AI, that’s a good thing overall. 

Paula Rivera: 

I concur. I totally concur. So let’s get into the report, which very thorough report. I really enjoyed it. One of the big shifts you highlighted in your report is US moving from flow-based bots, rigid scripts and trees to prompt-based gen AI powered agents. Derek, why is this such a significant turning point? 

Derek Top: 

Well, I think what our goal is to look at the world of conversational AI right now with the introduction of generative AI gen AI and the large language models that have happened over the past two or three years, we at Opus Research have been covering what we call, quote-unquote, enterprise intelligent assistance for over 10 years. This report actually began in 2014, so we had to kind of see what has changed in that time. And what had been traditionally intelligent assistance and bots as they were introduced 10 years ago, it was focused on how you can do self-service, automated self-service to help reduce call volumes, to reduce repetitive tasks, and that was true what was called what we call deterministic flows or of your traditional dialogue design, which had been done with IVRs, interactive voice responses, and the first chatbots and the voice bots that came out were all about deterministic. So that means if this, then that. 

And with the role of gen AI and large language models that became more prompt based, which does become a little more flexible in terms of the understanding of how bots can understand what’s being said, both from a voice and text point of view. It’s more flexible in terms of how that can be tasks that are done for organizations, especially within context centers. The deterministic flows that have been in the past were more limited in their flexibility, and the goal is to of use gen AI, to use large language models to use AI to become for self-service to become more like a human. 

Paula Rivera: 

Ah, okay. So how are enterprises navigating the two for the more predictable and the more flexible? 

Derek Top: 

Well, I think that’s part of the process here, is that as I was mentioned earlier, I think IVRs, that’s a very old school technology and so many organizations have had IVRs for 10, 20 plus years and have leveraged these pretty limited use cases in terms of how you call into an organization. So it’s more those kind of limited use cases, which also required this deterministic, this kind of labor-intensive dialogue designed, understand how to really kind of answer people’s questions through an automated way. That was limited in its functionality, also limited in its business outcomes. It didn’t really actually save a lot of money. It’s sure there is some what’s called deflection or containment, these kind of context center metrics that are used. 

What’s happening next is the flow-based and kind of the next generation of customer service through generative AI can be exciting, creative, more flexible, can really create more opportunities for businesses to answer questions, to save money, to actually get things done, have what we call a voice bot or some kind of bot, help a customer go through a transaction or return something or even schedule appointment. There are many different types of use cases that are used that voice bots do use, and so what’s next on the horizon? I think it’s pretty exciting. 

Paula Rivera: 

So let me ask a quick question. Inevitably, whether it’s on Reddit or I’m reading a news article, people are still kvetching about customer service, getting the bot, screaming for the representative. Is this because companies are still using old school technology and have been a little bit slow to adopt more modern technologies? 

Derek Top: 

Short answer, yes. I think that’s part of it. In terms of what had been the technology that folks use, there is a limited understanding from the bot perspective. So people get frustrated or they also based on previous experiences have thought, okay, I don’t want to talk to a bot right now. I just want to talk to a human. So agent, agent, agent, whatever’s kind of the exiting out of that type of path has kind of become commonplace based on previous experiences, based on previous technologies. And so what happens in this new world or potentially new world is that we can look at how AI and large language models and what we call prompt-based agents or however you want to call it really, there’s lots of different words out there that can understand better what’s being said, and that can also make sure that the person, whatever their frustration is, whatever their task may be, won’t get frustrated and actually get something done. So yes, there is I think a new world on the horizon based on that. 

Paula Rivera: 

Yeah, I do chuckle when I read these articles because I’m like, this should not be the case in this day and age. 

Derek Top: 

Yeah, I think that’s exactly, I think that’s part of it too, is that consumer’s expectations are changing, from what had been a very rote… Even the initial natural language that people use when it came to IVRs and just there was an episode of Seinfeld I think back in the ’90s. That was where Kramer was like, why don’t you just tell me the movie you want to see? Because that was the first beginnings of how people learned to talk to a machine. And so, those experiences have become such a common part of our existence that their expectations were pretty low. But I think the possibility of how AI and large language models can understand people better, it opens up new creative opportunities that opens up new business opportunities for customer service. 

Paula Rivera: 

Well, I love and appreciate the Seinfeld reference. My background actually is car rental, and inevitably that one scene with Seinfeld at the car rental counter at least once a year would come up as this is the epitome of what not to do. 

Derek Top: 

Exactly. Yeah. That’s funny. That was also a customer service situation too, where the manager had to tell him he couldn’t do something. So it is funny how that works. 

Paula Rivera: 

Definitely, definitely. So in the report, one of favorite distinctions is what you guys call pragmatists and true believers, A shameless plug, IntelePeer was labeled a true believer. It’s kind of philosophical in how you’re describing this. Do you evolve slowly or go all in on gen AI? Can you unpack what those labels mean? 

Derek Top: 

Sure, yeah. And the report itself, like I said, it’s something that we’ve been doing for many years. As part of our evaluation is that we do look at a lot of the solution providers that are out there, understanding what their current differentiators are with their current deployments, and how they’re looking at the near future for conversational AI and what they’re doing. 

As we started to look at the different approaches, we recognized that there was these buckets that people might fall into, both from pragmatist point of view, meaning something that’s practical, getting things done, and has some experience in leveraging previous technologies and to true believers who we thought were maybe a little more gen AI forward, who had the sense of providing some newer technologies that might leapfrog some of the older legacy stuff. And that was a way for us to categorize and distinguish some of the different folks. And IntelePeer actually, we do think among the leaders that are out there, we think that there is, it’s more of a continuum and it’s more of an approach that doesn’t necessarily mean you have to do one or the other. 

In fact, what we really see is this hybrid time that’s happening where a lot of folks are actually looking to, a lot of businesses and customer contact centers are looking to leverage their previous technology to leverage what they’ve built around IVRs and contact center automation in order to make sure that those investments are helpful in the near future. But at the same time, they want to close the gap from what you’re seeing out there, and they’re getting a lot of push from on from the CEOs and the world itself saying, use AI, use AI. And so there’s this push to try and become more of a true believer. We think both are very helpful and accurate in terms of where people are at right now. So we do think it is a evolution. We do think it is a process. We do think it’s something that’s going to take some time, but at the same time, there is definitely ways to leverage both your previous technologies in order to leapfrog into the future. 

Paula Rivera: 

So that’s actually interesting, and I’m inclined to say vendors who actually straddle both pragmatism and visionary probably are going to succeed in with the largest customer base. You just described the situation so many companies are in. 

Derek Top: 

Yeah. No, and I think a lot of the challenges too from a business point of view is that you’re looking at the state-of-the-art technologies, and like I said, you’re getting this push. I need to be using AI for my customer service needs, but I don’t know how to do that. And so leveraging solution providers like IntelePeer or others is really helping them leapfrog to really take that next step. Because a lot of problems are how do I teach my personnel to learn about AI? How do I use it in my business cases? How do I even know that it’s going to provide the value that it says it’s going to bring? Working with partners, working with folks, and this is part of what Opus recommends, is actually giving you a better deeper understanding to adapt, to scale, to really kind of future-proof your design, do everything you need to do in order to make sure that AI and conversational AI within your organization really gets the job done. 

Paula Rivera: 

Nice. So the technology itself is becoming a little bit commoditized, and what’s really setting leaders apart now is platform maturity. Tell us about the gen AI platform maturity framework that’s in the report. I think you have five pillars. 

Derek Top: 

We do, yes. And we look at, and these aren’t necessarily kind of check boxes, but at the same time it is important. As we start to look for organizations and providers, really, who are providing a framework that you as a business can take advantage of. There’s five different things we look at from that gen AI, again, like the generative AI platform. One is around agent orchestration. Really that means the ability to kind of have a platform that you can build agents or AI agents or bots or like I said, there’s lot of different terms that could be, is used in multiple ways, but that’s the orchestration layer. 

That also then leverages tools, which again are those agents themselves, but doing the right use cases, finding the right things within your organization to do, which is again, usually around self-service, but also around appointments, around sales and transactions, around changing information, customer queries that come. Those are the agents need to provide tools to get that done. In order to do that, they have to leverage knowledge. So that becomes all of the knowledge management and all of the information that’s out there. That’s the third layer’s, knowledge. 

Observability is the fourth one, which looks at this is more about the controls that are within your agent, your AI agent platform, how you can make sure it’s doing the right things, doing at the right time. This requires doing some testing, doing some prototyping, doing some pilots, and then controlling to make sure that your AI agents are doing what you want them to do. 

The last level is looking at evaluation and trust, which is the security. We talked about this a little bit, but there is some resistance to AI based on some of the experiences that are out there around trust and making sure that an AI bot says what it’s supposed to do. If it’s your brand, if it’s your business, you want to make sure that it’s within all the confines of those security protocols and that it doesn’t go off the rails, so to speak. So evaluation and trust and security are really another fundamental part of that. 

Paula Rivera: 

Yeah. And I think that’s been the case all along. In regards to the observability, you mentioned testing, and it kind of popped into my head that testing is almost something that needs to be a constant. It’s not necessarily test once and you’re done. You need to constantly be dipping your toe into your AI agents to make sure things are running well. 

Derek Top: 

Yes, and this is all about continuous optimization, continuous improvement. All of what you need to do from an organization standpoint is not just set it and forget it. It has to be something that you’re constantly monitoring and paying attention to and understanding what your outcomes are. So this I think touches on a little bit about, or could touch on the idea that AI is something that will take people’s jobs in the sense the AI needs humans for it to work better, we need to have a constant sense of evaluation and tweaking and optimizing and testing. And so, that is what part of this process is. And having a framework, which is what we described talked about earlier, is that kind of maturity framework gives you the tools that you might need as an organization to make sure you’re doing all the right things to take advantage of everything conversational AI is supposed to do. 

Paula Rivera: 

Excellent. Are there any areas overlooked or taken for granted? 

Derek Top: 

Well, it’s funny. I would emphasize that trust and security layer. So that’s something that we have also looked at from an organization is making sure that you as an organization embed trust from the beginning, having it deeply integrated into your organization. And even if you’re just starting out, even if it’s a limited use of gen AI, we do recommend that you have a strong commitment to that evaluation layer, to that oversight, making sure that you’re continuously optimizing and enforcing all of the same kind of compliance and security protocols that you would do for any type of other self-service technology you might have. And I think that’s especially true within organizations that are in healthcare or financial services or the typical regulated industries. We absolutely emphasize trust and safety. 

Paula Rivera: 

Excellent. It really sounds like this framework, it would be beneficial if a vendor is like, where do I start? How do I evaluate? What should I be looking for? It sounds like this framework is definitely a great jumping off point for enterprises to use as they start delving into their vendor selection. 

Derek Top: 

Yeah. And I think that is a good starting point. And I think we also recognize and recommend that every organization is different. So when it comes to your business seeds, you’re looking, just get started with a use case and find the right ways to leverage conversational AI. Here are some features that you might want to look forward to start with. We do absolutely recommend identifying a business case, identifying the need, and then choosing to solve towards that need. Don’t just bring in technology for technology’s sake. You really need an opportunity to find where your business can leverage conversational AI. And yes, this is a good place to start. And again, we also recommend working with pollution providers, having that as a way to leverage their expertise. They’re here to help you. 

Paula Rivera: 

Ah, that’s an excellent point. I like that. We talked a little bit about the trust. How should a vendor go about assessing trust, safety, maturity beyond a company’s marketing slides? 

Derek Top: 

Yeah, I think that’s a good point, actually, is that sometimes the marketing says they do trust or at least they propose to, and then as you get down to lower, there might be some things lacking. So again, this does just go back to your ability as an organization to identify your security layers, your compliance needs. We actually do see many times when it comes to AI and conversational AI within customer service or anywhere, is that you work with others in your organization. 

So many times we’ll see a task force or a center of excellence or kind of bring in all the different business units and stakeholders and decide, okay, hey, we’re looking at introducing a new conversational AI experience for our customers. What are some things we should be concerned about when it comes to security or trust? And usually it comes down to things like data and privacy and all the kind of common things that a lot of organizations will already have some protocols in place. But as you unleash AI into what’s next, those same stakeholders from a compliance need, from an IT security need, from even your customer CRM, like your customer databases, make sure that all of those security protocols are in place. And then once you have that to start with, that does become embedded in your solution as you go forward. 

Paula Rivera: 

Nice. I just finished our company’s security training and I’m like, basically it’s like employee security training, except it’s for the AI model. 

Derek Top: 

Yeah, exactly. Yeah. 

Paula Rivera: 

So are there any red flags, enterprises that are evaluating conversational AI solutions? Are there any red flags that buyers should be aware of? 

Derek Top: 

Well, I mean, part of this current world that we’re in right now is this accelerated piece of innovation, and so how are you as an organization going to keep up? And I think for the most part, it is working with your solution providers, having a common goal and having that sense that as new models are introduced, as new capabilities are unleashed, you as an organization are able to take advantage of those in a way that makes sense from your business to be able to scale. So the red flags are somewhat based on your interests, your needs, and so it’s hard to necessarily point out one thing to specifically look for. 

 It’s more about are your business goals being met? And then as you think through where you can scale, are your providers, your vendors, your collaborators working to get those? A lot of the stuff is, it’s funny because we’re talking science fiction just in recent years, but even by the end of this year, next year, capabilities are out there where we’re going to be able to talk to bots in a very casual and uncomfortable way. And so, is your business ready to match that? And I think that’s something that as an organization you have to pay attention to. And if you’re not, then that’s a red flag right there. 

Paula Rivera: 

I concur completely. It is a brave new world. So I don’t want to delve too much into pricing because pricing can be complicated, whether it’s from the vendor perspective or from the buyer that might have certain criteria, but do you have any advice for matching a pricing model to a business strategy? 

Derek Top: 

Yeah, and this kind of gets back to what I just said in terms of it depends. It’s based on your needs. What are your goals? Pricing, and especially this can be for those that are used to models around contact centers, which we’re focused on kind of a price per agent per seat, as it used to be called or is called. So that would give you a sense of how many… It gives you a security in terms of what to expect based on your contact center needs and that scale, sometimes it goes up for whatever your organization may be. If it’s for holiday season, things go up or open enrollment, there are definitely some spikes and peaks to help people price contact centers and customer service. So that’s a kind of known predictable model. 

Consumption-based model pricing, which means, so how can be also for individual computational AI, for bots, for things like that. It could be based on conversations, so the number of words or exchanges it might have, or it might be based on the number of concurrent interactions. And there’s lots of different things that are ways to determine that, but that type of pricing can be highly unpredictable or variable. And so again, these are kind of models that folks are somewhat familiar with. AI also, we talk about the AI models, and they’re also using what’s called tokens, which is a way to get information back and then visit you. That can also be variable by the [inaudible 00:22:12], the type of model. And so these are the different variables that go into it and can change based on whatever your business needs are. 

So again, this is where you do work with a solution provider. Sometimes your providers will actually embed and take in that cost. And so, that might be something to look at. Or if you’re looking for something a little more flexible, a little more granular, then you can look at more of a consumption-based model. 

Paula Rivera: 

You took my last question right out of my mouth, which is, well, first off, you need to do some self-evaluating. But if you’re not sure where to start, work with the solution provider, most vendors will sit down with you and help you figure things out. Really, I think really knowing and understanding your needs will help actually, whether it’s what solution you’re getting or what pricing you’re paying all around, and key to work with the solution provider. 

Derek Top: 

Absolutely. And I think that that’s the success stories that we’re seeing right now are because of that and really understanding that that is a partnership that they can use to move forward. 

Paula Rivera: 

Excellent. Well, listen, we are getting close to the end of our segment, but before we let you go, I like to do, and I do this with all my guests, it’s a rapid-fire around, I like to call it. And I have three questions. You can give me one-word answers or you can expound upon the answers, but this is to get to know Derek a little bit outside of the world of conversational AI. Although my first question actually is about conversational AI. So one myth about conversational AI that drives you crazy? 

Derek Top: 

Well, I would say it’s that it’s not to be trusted. I think we’re actually getting to a point where the security and the guardrails and hallucinations are something that is a little overblown. We hear about some of the worst-case scenarios that are happening for conversational AI based on some high-profile experiences that are out there. We’re getting to a point where the more we use it, the more useful it becomes. And I think that’s what AI is going to be to us all in general, really. But there’s a fear and anxiety that I think overstates what is actually possible. That’s not to say there shouldn’t be. There always should be some security and trust that we need to have, and that requires an ongoing continuous monitoring of it. But I think some of that stuff is overblown. 

Paula Rivera: 

I concur completely. Favorite use of AI in your daily life, whether it’s personal or professional? 

Derek Top: 

Well, I absolutely use it for sure. In terms of consuming information, it’s a way for me to gather and summarize and know how to get the too-long-didn’t-read readout on every single thing in our lives. I think we’re all doing that in a way that makes us understand the world faster and better. I would say the one thing, from what I’ve seen in contact centers, I mean, obviously there is the use of self-service and conversational AI for getting things done, but I’ve actually seen contact center agents and IT help desks and others who are using… Well, there’s Google Notebook LM, which is a way to basically take information and then create a podcast out of it kind of like this. And then they’ll actually use that as a way for them to consume the information, like say, a highly digital technical support manual that actually helps them as a contact center agent do their job better because they understood in a way that is more digestible and easy. So that is exactly what I think is how AI can help us understand our world. 

Paula Rivera: 

Well, it’s so interesting because I’m a big fan of Perplexity, and I just recently, actually earlier today, I just plucked a question in there and I’m like, “Explain this to me and how is it used in this scenario?” And the response was so concise, and I’m like, wow, I would’ve probably had to have read three or four different articles to get what this little machine just spun up for me, and it truly was helpful. It was pretty darn amazing. 

Derek Top: 

Yeah, I agree. 

Paula Rivera: 

Yep. So if you could automate one part of your job, what would it be? 

Derek Top: 

Well, I mean, we can’t clone ourselves. There’s that. I can’t be in different places at once, but I do like the idea of a very powerful assistant. So this is something that really does understand my day, has my calendar, knows my needs, has a way for me to get things done and actually even become a proxy for me, that is the next level that we’re looking at too, by the way, in terms of personal assistance and kind of the bot-to-bot communications, there is a new world that’s going to be happening within that, and I think that’ll become part of your work life too. So in terms of automating, part of your job would be to answer emails or to follow up on particular conversations. We’re not there yet. I mean, my phone calendar doesn’t sync with my laptop calendar. That would be something that’d be nice to have that automated, but I think there are some new ways for us to leverage the technology we have in our everyday lives. And so I’m excited for that to come. 

Paula Rivera: 

Yep. I concur completely i, on a personal level, am waiting for the day that Rosie from the Jetsons is widely available. I love my iRobot, my Roomba, but I need someone who’s going to go a little bit further than the rugs. 

Derek Top: 

Exactly. We’ll get there. We’ll get there. 

Paula Rivera: 

Yeah. So Derek, thank you so much. This has been an incredible conversation. Your report offers such a clear-eyed view of where conversational AI is today and where it’s going. Our listeners, if you’re exploring or evaluating platforms, definitely be sure to check out the 2025 Conversational AI Intelliview over at Opusresearch.net. Derek, thank you again for joining us. I really appreciate it. That’s it for this episode of the AI Factor, CX Explained. Don’t forget to subscribe. Leave us a review and share with your fellow AI curious colleagues. Until next time, stay smart, stay skeptical, and stay ahead.