The App of the Future Is Voice — Not a Screen. Mitel's CTO Luiz Domingos Explains Why.

[00:00:00] [SPEAKER_00] A lot of organizations want AI-driven automation without properly updating their systems. What modernization do organizations need to undertake to adopt AI smoothly and avoid bottlenecks? What's changed most in how organizations are actually using AI today versus how they talked about it two or three years ago?

[00:00:22] [SPEAKER_01] Since 2022, everybody's now really way more active in the day-to-day of AI solutions. Voice is becoming now again the natural interface for AI. You're going to be talking to AI. AI will be your friend. Enterprise are really freaking out about how I manage accountability and what's the governance in the enterprise. How you address the responsibility at the end of the day when something goes wrong.

[00:00:46] [SPEAKER_00] Okay, well, I usually start by having you introduce yourself. Okay. Give us as much of your background as is relevant. Mitel is a Canadian telecommunications company, as far as I know. So you can talk about how you got to Mattel and what you guys are doing.

[00:01:11] [SPEAKER_00] So why don't you go ahead and introduce yourself and then we'll start asking questions.

[00:01:21] [SPEAKER_01] Mattel is a Canadian telecommunications company, as well. Thanks for having me, Craig. So my name is Luis Domingos. I am Mitel's CTO. I am in the business of enterprise communication and collaboration solutions for quite some time. That's my whole career in general. And I'm with Mitel for two and a half years now. Mitel acquired a company called Unify, also another big vendor in the enterprise communication markets two and a half years ago. So I'm with Mitel for this period of time.

[00:01:49] [SPEAKER_01] And I was previously CTO for the previous company for Unify. I am in this business for more than 25 years. And I've been following the whole evolution of communications from the old times of IP telephony through cloud and now with AI, of course, AI communications as part of it. So my whole career is the year for sure.

[00:02:13] [SPEAKER_00] Yeah. And Mitel, I'm sorry I was pronouncing it wrong. Mitel builds on-prem or cloud and hybrid platforms that integrate voice, video messaging, and contact center capabilities. So employees and customers are using a cohesive system.

[00:02:38] [SPEAKER_00] But how has AI impacted that industry?

[00:02:45] [SPEAKER_01] Yes. Mitel is really broadly providing solutions for communication and collaborations. So we have, we're more than 50 years in the market. We have a very large installed base, 70 million users. And as you said, communication collaborations, contact centers, virtual applications. We are not new with AI. So correct. So we started AI at Mitel in 2018, 17, 18, with the first investigations about natural language process.

[00:03:13] [SPEAKER_01] We brought dialogue flow solutions for contact centers. But certainly we are evolving along with AI. And since 2022, everybody is now really way more active in the day-to-day of AI solutions. And we've been applying AI to communications. And communications is a segment of the industry where AI can be easily justified to the business. Correct. You can gain significant value right immediately.

[00:03:42] [SPEAKER_01] And it's very visible. Correct. It's very tangible. You can measure, you can identify the value that it brings. And that's how Mitel has been evolving alongside with AI in the communications industry.

[00:03:55] [SPEAKER_00] Yeah. But how is it applied? I mean, are you talking about voice, you know, these customer service platforms that now use AI voice or how is it being applied?

[00:04:14] [SPEAKER_01] Yeah. Let's take the example of contact centers. Correct. So we have AI inside the contact centers. AI is the assistant to the agent. Correct. It's a co-pilot function. The contact center agents have an AI assistant for all the functions. Also to evaluate the calls, provide guidance to the agents on how they react or they process the call, how they interact with the customer.

[00:04:42] [SPEAKER_01] And providing sentiment analytics, providing summarization of the call and guiding the agent on getting the best customer satisfaction. And also to get really job done fast. Correct. So summarizing the call, documenting the call. All of that in the contact center front. When I look in the unified communications and communications in general, then we talk about

[00:05:10] [SPEAKER_01] certain aspects of the communication that are very important. For example, when you have call routing, correct? So call routing in a solution for communication is very important. And then when we talk about certain industries, healthcare, how you provide communications in a healthcare environment, or how we provide communication solutions with AI embedded for communications workflows in government.

[00:05:35] [SPEAKER_01] So some of those are important workflows and scenarios where Mitel provides that into our AI solutions. And that really brings value to our customers.

[00:05:46] [SPEAKER_00] Yeah. So you guys were in AI long before the current wave. From your perspective, what's changed most in how organizations are actually using AI today versus how they talked about it two or three years ago?

[00:06:06] [SPEAKER_01] Absolutely. So I recall being on stage in 2019 talking about doing an automated translation with a German colleague. So it was kind of novel at that time for sure. But we've been doing that for quite some time. So a few years ago, let's take 2022 when we had ChatGPT coming to live and visible across the whole planet.

[00:06:33] [SPEAKER_01] So companies were trying to understand what was the value, right? So what can I take, what can I extract from that? So what fundamentally changed from two, three years ago is that before it was more slide work. So companies were trying to create value, understand the value, doing proof of concepts. Today is more about execution. It's about the workflow.

[00:06:55] [SPEAKER_01] Companies have moved from AI strategy decks to innovation labs and now to operational AI embedded directly into the real business processes. So I think we shifted from generic chat bots to AI side, contact centers, meetings, compliance workflows, where the outcomes are certainly measurable.

[00:07:16] [SPEAKER_01] So I think the conversation now shifted from which model is the smartest, is the most relevant, is about which deployment brings ROI, which deployment respects the enterprise governance and reduces friction in the daily operations of your business. So I think the leadership, the RxX are always asking, how can I improve the business and still be compliant, right?

[00:07:45] [SPEAKER_01] So those are the two main factors. And critically, enterprises realize that AI, the cloud-only AI doesn't solve everything. So that's another point that we always make at Mitel, that hybrid and edge architectures are becoming essential, especially for real-time voice and regulated industries.

[00:08:05] [SPEAKER_00] Yeah. Can you give a use case to illustrate that?

[00:08:10] [SPEAKER_01] Yes. Edge AI is very important in communications. So if you think about communications, milliseconds matter, correct? So latency is a problem. And if I'm in a contact center, somebody calls me and I'm using AI assistant to transcribe the call, understand the meaning, and give me the recommended reply. If it takes two seconds, it's too late. I'm in a live conversation with the customer, correct? So it has to be really more immediate.

[00:08:41] [SPEAKER_01] So latency is a problem. And if I have edge AI, if I have something closer to my day-to-day, to my business deployment, and don't depend on network access, bandwidth, et cetera, it's way more efficient to operate. And I think that's one value for contact centers for sure.

[00:08:58] [SPEAKER_00] Yeah. And in unified communications, is it the same thing? Is the application of AI the same?

[00:09:05] [SPEAKER_01] It's slightly different because while a contact center is all about every second matter, correct, in the communication, in the contact center operations, how much you can offload from the agent to the AI virtual assistant or virtual agent. And in the unified communications is about maximizing the value that the enterprise created with their knowledge, correct?

[00:09:31] [SPEAKER_01] So in the unified communications is about sharing knowledge, finding knowledge in a more efficient fashion. And that's where our knowledge bases, regs, and taking the most value out of your enterprise culture and knowledge makes all the sense. And that's where we have the ability to connect to those enterprise knowledges, understand the value, identify the right workflows that the enterprise possesses, and create value and efficiency within the communications of the business.

[00:10:00] [SPEAKER_00] Yeah. And you'll forgive me because I'm not in communication. But what does unified communications mean?

[00:10:10] [SPEAKER_01] Unified communications comes from the view that communications in general is not simply voice calls, correct? Communications is all sorts of individual engagements within the company that leads to execution of the day-to-day of the business. For example, I can start, I'm having this conference call with you or I'm having a call to you.

[00:10:32] [SPEAKER_01] And we decide to share information through a collaboration session or I decide to have a chat with you or I decide to bring a third person because that third person is an important collaborator in the business. So it's all means of communication. Certainly started with voice, with telephony, but evolved through collaboration capabilities, evolved through omnichannel and multimedia, multimodality of communication.

[00:10:57] [SPEAKER_01] So then you have the most efficient way through any channel of communication to achieve the objectives of the enterprise and basically be efficient, satisfy customer needs, interact with the day-to-day of the organization.

[00:11:11] [SPEAKER_00] Yeah. So in AI, I mean, again, give me a use case there.

[00:11:17] [SPEAKER_01] Okay. Oh, yes. For example, we have my support organization, correct? So we support customers and we provide all documentation about our products for customers. So one scenario is where my support engineer or my support customer support person will receive queries through text. And these queries will normally be asking about any sorts of information about our products, how I deploy, how I operate, how I provision it.

[00:11:47] [SPEAKER_01] And this person has the responsibility to provide the right answers. In the past, he would come to me, correct? Hey, Luis, what does this mean? How does this work? And I would write back to them through a text chat. And he would provide that feedback back to the customer or to the client, whichever that is. So now with AI, we can take all that knowledge into a knowledge base and have AI as an assistant on top.

[00:12:15] [SPEAKER_01] And the AI will provide that information back to you in the forms of guidance, insights, tidbits that you can use in the communication. So that's how we do it. The other element is the routing capabilities. Remember that before, you would call somebody in the enterprise and you would fall into a particular place. And with intelligent routing, you can identify what's the best person with the most knowledge to answer certain calls and certain queries. And calls in this case can be simply a telephony call.

[00:12:46] [SPEAKER_01] It can be a chat session. It can be a social media inquiry. So all these media come together into the UC solution to provide the best outcome to the enterprise.

[00:12:57] [SPEAKER_00] Yeah. And the AI, then you're not talking about agentic AI. You're talking about RAG. Yeah.

[00:13:09] [SPEAKER_01] Augmented. Yeah, exactly. So I'm talking about generative with augmentation. Because generative without augmentation is still not adapted to the enterprise. You need to understand your enterprise value and extract what you need from it to be successful. Correct? So that's why we do RAG to augment the generative AI and have the spice component of the enterprise. That makes a little difference. Correct? Otherwise, it's just plain vanilla AI.

[00:13:39] [SPEAKER_01] That's not what enterprises in various business segments need. They need their own aspects into the day-to-day of AI.

[00:13:47] [SPEAKER_00] Right. Is there any application of agentic in your business?

[00:13:54] [SPEAKER_01] Absolutely. Absolutely. And agentic is something that we've been investing a lot in. So we created a solution called Workflow Studio. Workflow Studio is basically creating workflows, of course, his name says, but with the objective of creating workflows for communications. Correct? So I can start, for example, I can call a colleague in the office. And this colleague is the one that will provide valuable information for me.

[00:14:22] [SPEAKER_01] And that information can be retrieved through different means. Correct? So we can have the necessary MCP servers that collect all that data. And this allows me to create a workflow that has communication, AI consultation, action-based functions. For example, I can create a ticket. Correct? So somebody calls me. I'm the IT person in the company. I'm the IT support, helpless support. I'm going to understand what they need.

[00:14:50] [SPEAKER_01] And in normal cases, I would just write a ticket. Correct? Now a person will be able to call in and agentic would help me create the ticket automatically in a much faster way and create, log that ticket in the system for further processing. So the workflows of the enterprise and all the steps required are extremely automated in how you communicate with Workflow Studio.

[00:15:15] [SPEAKER_01] So I think that's the new element that was not there before that allows you to bring agentic where it matters, to take actions on behalf of the users in the enterprise day-to-day workflows. And it's not only about support or contact center. It's also about the day-to-day of automating tasks in, let's say, the finance department. Correct?

[00:15:38] [SPEAKER_01] So how finance can use that to embed into communication solutions and deliver agentic flows to the day-to-day of the business.

[00:15:48] [SPEAKER_00] Where do you see enterprises overestimating AI? I mean, you work with a lot of customers. Do they ask for things that are unrealistic? And where should they focus to realize near-term impact?

[00:16:09] Yeah.

[00:16:10] [SPEAKER_01] So they ask for realistic things, correct? So customers always ask for realistic things. It's just that many times they want a certain technology applied to it. And the solution might be slightly different, correct? So you have to work closely to your customer to understand what is the use case. That's what starts the dialogue, correct? So what you're trying to sort, solve. And we offer how our products and how AI within our products can contribute to them.

[00:16:39] [SPEAKER_01] Take the example of frontline workers, correct? So nurse in a hospital that need to have a communication device or a communication application that gives them guidance on or interaction in the hospital in the day-to-day of the system, correct? So alerts, alarms, messages from the doctors or addressing emergencies, emergency room needs. So how do they communicate?

[00:17:07] [SPEAKER_01] I think that's an important question in how AI plays a role. Frontline workers are a little different, correct? So in my view, they are the most underserved by the digital transformation. So we are knowledge workers, correct? So we have all sorts of solutions. Frontline workers, they still lack real-time digital tools altogether. And communications AI bridges that gap.

[00:17:31] [SPEAKER_01] So what we're doing, we've been embedding AI with real-time insights, automated documentation, intelligent routing, and the workflows. And that means faster response times, better compliance, less manual paperwork. Because in their case, they are professionals that are not in front of a desktop or browsing through an app. So they need something more immediate.

[00:17:57] [SPEAKER_01] So the way you interact with them is different than the way you interact with a knowledge worker like ourselves. So it has to be event-driven, notifications, critical situations that have to be alarms. So it's a different way of the interaction. And you need to understand how you implement that. So that's really a different scenario on how AI plays a role. But certainly AI, in this case, can do something that we couldn't do in the past.

[00:18:26] [SPEAKER_01] And we are a voice company, correct? But I believe voice is becoming, now again, the natural interface for AI. You're not going to be texting, chatting with AI. You're going to be talking to AI. AI will be your friend. You're going to be talking to AI day to day, bringing the needs. And AI will understand and will interact with you that way. So if you are, as a frontliner, that's fantastic, correct? You're not really now looking at a screen where you have to take care of the patient and the bed. So you just talk to it. So it's much better.

[00:18:56] [SPEAKER_01] And I still believe that in the near future, voice will replace traditional user interfaces in your apps. Perhaps the app of the future is a voice-enabled app. It's not a front GUI that's more traditional as we had before.

[00:19:12] [SPEAKER_00] Yeah. And do you guys have your own text-to-speech models? Or how do you deal with voice?

[00:19:25] [SPEAKER_01] We are consumers of text-to-speech, natural language processing models. What we do is that we created our AI services platform that ensures that we can connect to any language models. So we try to be as agnostic as possible.

[00:19:46] [SPEAKER_01] And we basically have a middle layer that allows you to connect any language model for language processing in general, for text-to-speech, for speech synthesization. So all of that is through one standard interface. And that allows us, our applications, to have only one implementation.

[00:20:07] [SPEAKER_01] And then you can decide based on customer demand, based on regulatory, which kind of language model you deploy. And at the same time, you can define where you deploy. And that's the conversation about hybrid AI that we talk a lot about, where you want to deploy whatever the customer needs. Correct? So if it's public cloud, typical, traditional.

[00:20:33] [SPEAKER_01] But we also have what we call secure cloud, which is a sovereign solution that we deploy to certain customers in Europe. Or you can even bring to edge and deliver this on customer premises so the customer has the best solution for latency. Different models of deployment with any language model. So that's what we promote. And that's what our infrastructure permits our applications to provide to customers. Yeah.

[00:21:00] [SPEAKER_00] And how does a customer work with you? I mean, you're talking about these platforms and interfaces. Do you provide them with the platform and then they can configure their communications, you know, however they want?

[00:21:20] [SPEAKER_00] Or is this interface something that you have access to and you configure for the client?

[00:21:33] [SPEAKER_01] Yeah. So we certainly, there are clients out there that are very self-sufficient, correct? We have self-maintainers of our solutions that certainly do everything. But that's a small minority. So the way Mitel operates is that we have two models. We have a large base of channel partners, 6,000 globally. And they have their service experts that go there, understand the use case of the customer, configure, update, operate.

[00:22:02] [SPEAKER_01] We also have our professional services organization. And our professional services does the same. Normally for larger enterprises, right? So we have that services that we offer in addition to the products. So we give the products, let's take a contact center with all the capabilities of AI built in for summarization, sentiment, assistance, transcription, etc. And we offer the services on top to deploy and later, of course, operate on behalf of the customer.

[00:22:30] [SPEAKER_00] You know, a lot of organizations want AI-driven automation without properly updating their systems. What modernization do organizations need to undertake to adopt AI smoothly and avoid bottlenecks?

[00:22:52] [SPEAKER_01] Yeah, it's a good question, Craig, because I don't think you can pour AI into a legacy architecture and expect transformation. That would be a little bit too much wishful thinking. A lot of companies want intelligent automation layered on top of updated systems, fragmented APIs, inconsistent data. So that doesn't create deficiency. It just raises complexity.

[00:23:18] [SPEAKER_01] So modernization doesn't mean ripping everything out for sure, but you still need to become API first. You need to decouple the communications layer from the workflow layer. You need to build modular AI services like to do data ingestion, retrieval, orchestration. And all of those need to plug into existing systems. So in our view, hybrid integration is often the smartest path, correct?

[00:23:46] [SPEAKER_01] So gradual modernization, clean data pipelines, clear governance. That's what will lead the company to go and modernize and become way more efficient with AI. Otherwise, the structural issues will limit how much you can extract as value from AI in the day-to-day.

[00:24:08] [SPEAKER_00] And are AI capabilities across the Mitel portfolio? Or are there some areas that are more AI forward than others?

[00:24:26] [SPEAKER_01] Yeah. I would list four different segments where we have AI in the portfolio. So I gave you the example of contact center solutions. We have a very modern contact center today with AI intelligent routing, sentiment analytics, real-time transcription, automated summarization, agent assist, virtual agent doing the call screening before it comes to the real agent, to the human agent.

[00:24:53] [SPEAKER_01] So that's one area where we spend a lot, and that's the area we've been working the longest on AI. And we know that has the most immediate return of investment, correct? That directly improves customer satisfaction and operational savings and efficiency for the contact center operator.

[00:25:11] [SPEAKER_01] Unified communications, as I mentioned before, we deliver smart meeting recaps, transcription, context, insights for the context that reduce communication overload. So I mentioned about workflow studio, correct? Orchestrating automation across systems, turning conversations into executable workflows, which for us in communication world, that's a major change. Because in the past, it was very strict, correct?

[00:25:40] [SPEAKER_01] So you had calls come in, calls go to a point, and that's it, correct? So it would require human intervention. Oh, this is not the right person you call. Let me transfer it to somebody else. So all of that intelligence now is with AI. So that makes a fantastic difference in the day-to-day of an enterprise. And with the edge AI capabilities, we support real-time voice intelligence, where latency, compliance matters the most.

[00:26:05] [SPEAKER_01] And compliance is very important as well, because many times we don't see this, but in certain regulated industries, that's a must, correct? Finance, healthcare. And also in some countries where data privacy, sovereignty is key. We have several customers in Europe where we cannot just deploy AI from the cloud. We definitely need to deploy AI in a more constrained environment.

[00:26:29] [SPEAKER_01] So these are scenarios where AI solves real operational pain and brings value to them. So in this case, I see AI, at the end of the day, is the intelligence layer and integrated in everything we deliver at Mitel today and across our portfolio. So that's where we've been working very hard and we continue to drive more, particularly now with Agentec.

[00:26:51] [SPEAKER_00] Yeah. And what are some of the unique challenges faced by regulated industries? I mean, you talked about compliance. Is that keeping a record of calls and that sort of thing?

[00:27:07] [SPEAKER_01] Yeah. Let me give you an example. Finance, correct? So finance, if you are in the business of trading or some business that are transactional driven, you need to record all the calls. Okay? So that's mandatory. And it's not only for training purposes. It's really for compliance reasons. And they record all the calls.

[00:27:29] [SPEAKER_01] So, and as they recorded the calls, they need to be able to go back and say, okay, there was a potential issue with the transaction or bank transaction or stock market transaction that has to be reviewed by the regulator. And they need to provide that information there. So in the past, it was really difficult to locate the call, understand what happened and provide the best information. Now with AI, you can scan through, you can transcribe all the calls and understand what happened.

[00:27:58] [SPEAKER_01] So that's one scenario where compliance is very important. And companies are liable for that, correct? So there are fines, particularly if you don't do that, if you miss or if you don't have that in place. So that's one key case of compliance from an industry perspective. And then there is the regulations, correct?

[00:28:17] [SPEAKER_01] So if you talk about European GDPR or European AI Act, all of them have specific requirements for AI providers and companies with AI products on how they treat customer data or individual data, correct? Data information. And how you process that.

[00:28:38] [SPEAKER_01] So the process and utilization of data, as well, how you protect the data from inadequate authors or people that are trying to take advantage of it, scammers, et cetera. So there is a certain element of protection and compliance that we need to put in place. Many companies today try to address that, certainly with solutions like what I mentioned with the recording. But also you need to be prepared to have auditing.

[00:29:08] [SPEAKER_01] You need to have the ability to demonstrate that what you implemented is really compliant. So all of that with AI makes it much harder, correct? Explainability, auditability of the decisions and transactions. All of that needs to be captured. And that leads to another interesting discussion. Because as we introduced agentic, now where is the decision point? When where is the liability?

[00:29:32] [SPEAKER_01] So enterprises are really freaking out about how I manage accountability and what's the governance in the enterprise. How you address the responsibility at the end of the day when something goes wrong. So I was attending IT Expo a few weeks ago. And the talk of the town was not about all the beautiful AI agent capabilities. It was about governance and liability, auditability.

[00:30:00] [SPEAKER_01] Because CIOs today are trying to understand. So how I deploy this in a responsible and reliable way? And many times you don't have that, correct? Many times you don't know how to do it. So that's where we're trying to support our customers with solutions that care and understand that landscape. And at the same time support deployments into regulated markets and in countries.

[00:30:24] [SPEAKER_00] And do you have specific products that you sell to address these issues? Or is it on a case-by-case basis that you consult with the customer and design something that meets the compliance needs?

[00:30:46] [SPEAKER_01] Yeah. On the product perspective, it's not that immediate, correct? So our products have capabilities that enable that. For example, I can deploy my AI application in an infrastructure together with the language model. So that's one scenario, correct? So I can't attend the requirements of data privacy or sovereignty with that. But it's not a product per se. It's a feature, a capability of our solutions that allow us to be compliant.

[00:31:14] [SPEAKER_01] Now, in addition to that, each customer, each vertical is unique, right? We need to understand that. And what we do normally is that our service organization provides advisory services. We sit with the customer, understand their particular requirement and what is compliance in their minds. We help them create the compliance framework. And in many cases, we implement that as well.

[00:31:41] [SPEAKER_01] So either we recommend certain applications or we implement it for the customer. So I think it depends from case to case. But we don't have a tool per se, but we have certainly the capability of creating specific solutions for customers that are compliant and deliver the governance that the customer expects.

[00:32:01] [SPEAKER_00] Given the growing interest, how are you using agentic AI in your solutions?

[00:32:11] [SPEAKER_01] Yes. Yes. Today, we have an initial implementation for contact centers. So our contact center has enabled workflows. As I mentioned, Workflow Studio is the application that delivers that. And we are able to take agentic features in order to do basic functions in the contact center, correct? Summarize the call and write back into the database so the agent doesn't have to do it.

[00:32:35] [SPEAKER_01] Write a ticket in ServiceNow, for example, to document the conversation or to actually require a follow-up from the organization. So moving some of these workloads and understanding what's the next step is what we've been doing, correct? So it's initial at this point, but certainly makes a difference already. Because in a contact center, 20 seconds that the agent will be there typing a ticket makes a difference. So 10 seconds here, 20 seconds there to summarize the call.

[00:33:04] [SPEAKER_01] So that's fantastic and unique. And that's what we've been using as initial agentic workflows. So we want to expand that, but we need to research more. And another point that comes in under agentic is how you ensure that agentic doesn't make mistakes by being too unsupervised. I think supervision is the question that many enterprises out there are trying to solve.

[00:33:30] [SPEAKER_01] And how much autonomy you give to AI agentic functions and how much you control it. Where do you need human in the loop? And again, going back to my compliance case, Greg, so nobody will allow an AI workflow to go on its own without a human in the loop or somebody to say thumbs up, do it. Because it's too risky.

[00:33:53] [SPEAKER_01] And no enterprise wants to run the risk because that's reputational risk, litigation problems, liabilities right there. I think that's why enterprises are very concerned.

[00:34:05] [SPEAKER_00] Yeah. And also in routing, you were talking about intelligent routing. Is that agentic?

[00:34:14] [SPEAKER_01] No. Routing is not agentic. It is just regular generative with some logs. Basically, we use the basic capabilities to determine what's the best routing feature. What's the best routing destination for a particular call, for a particular request. That can be done without it. Yeah. And actually, there's always a certain confusion in the market because the fact that we have contact centers and contact center has agents.

[00:34:42] [SPEAKER_01] We created agent features in support of the contact center agent. So AI agents, AI assistants are not agentic by nature. They basically are processing data. They're processing the conversation information and delivering insights and recommendations. Now, when you go into a step forward, then it becomes agentic because it's not only giving recommendations and next best actions.

[00:35:07] [SPEAKER_01] They are also, agentic is now providing precise actions into the enterprise workflow. So I think that's a difference. And sometimes there's an overlap or a misunderstanding, but certain functions can be done without agentic and very well done, by the way.

[00:35:24] [SPEAKER_00] Yeah. As you said, trust and governance are constant concerns. What is responsible AI execution look like with communications and data?

[00:35:47] [SPEAKER_01] Yeah. So responsible AI in this case means that you treat the customer data. Remember, in communications, we're talking about conversations, correct? A chat or a call, correct? There's a lot of enterprise secrets involved. Yeah. So secrets are the information that eventually will go to court one day. And you must ensure that all of that is well protected and handled, correct?

[00:36:14] [SPEAKER_01] And when you bring AI on top of it, you must ensure that AI is treating that information with the same care. So that's one element of compliance and regulations. And also, how we audit all of that, correct? So how do you ensure that that data is being handled properly by AI? So is my knowledge base, is the communication I had with the customer that got transcribed, is that going to the public cloud and becoming knowledge? Or how can I protect that?

[00:36:42] [SPEAKER_01] So it's just for my enterprise, just for my enterprise usage, and does not really become common knowledge. So I think all those concerns is something we put in the implementation to put guardrails and protection. So our data, customer data, private data is not really polluted or not lost into the public cloud. That's a major concern today.

[00:37:07] [SPEAKER_01] And that's another reason why edge AI is something I believe, truly believe, will be very important in the future. And then customers will have the option, correct? So certainly there is different price layers, but I'll be able to say, well, I want better protection. I want more isolation of my data. And therefore, I have a solution that is contained and limited in certain boundaries for me. I think that that will be a differentiator.

[00:37:33] [SPEAKER_01] And that's what I believe companies that are in the day-to-day of communication should be looking at, because that's where customers care about. And in many cases, that's first and foremost in their minds before even thinking about deploying AI.

[00:37:48] [SPEAKER_00] As AI takes on more scheduling, analytics, workflow tasks, are IT teams being restructured? I mean, how is it affecting your customers' organizations?

[00:38:07] [SPEAKER_01] Yeah, it's a good question, because IT with AI is morphing into something different. I think companies are not only adopting AI, correct? It's not just a technology upgrade. It has to a degree is an organizational redesign.

[00:38:23] [SPEAKER_01] So if AI automates, as you said, scheduling, analytics, workflow orchestration, the traditional boundaries, the traditional silos between IT operations and other organizations start to break down. So AI systems, in that case, they will cut across domains. So governance and ownership will do as well. So the IT leaders will have to build AI literacy across roles.

[00:38:53] [SPEAKER_01] Because many companies today, they lack knowledge about AI and how to apply it. So that awareness, that literacy net should be built. And not only with the data science teams, because whenever you are in IT and in technology, you always think about let's train the IT and the technology people. But in many cases, you need to go beyond operational knowledge, because they need to know what's the value, how they interact with AI.

[00:39:21] [SPEAKER_01] So the cross-functional workflow ownership becomes essential. And you need to have some kind of governance council within the enterprise to manage risk, ethics, compliance in a holistic fashion. So I think organizations that win with AI will not just be deploying tools. They will build a whole organizational construct. So collaboration occurs in a completely different model.

[00:39:49] [SPEAKER_00] Yeah. And do you advise customers on that? Or is that just something you're seeing happen?

[00:39:55] [SPEAKER_01] We see that happen today. And we advise customers when we talk about communications, correct? So certainly IT is much broader. And in the communications front, we're already advising our customers how they have to change how communication occurs within the enterprise. But I also observe that it's not only communications, because then you start talking about opening tickets and support systems,

[00:40:21] [SPEAKER_01] interacting with CRM, ERP as part of a communication workflow. That will go across the whole enterprise. So it becomes no longer communications only. It becomes communication and IT systems interacting with each other. So I see that. But so far, our advisory and guidance to our customers is focused on communications, which is our core. But we are looking beyond that. And we'll have something more to share in the future. Okay.

[00:40:49] [SPEAKER_00] You know, as AI scales, there's a concern, growing concern about energy consumption and environmental impact. Do you hear those concerns from customers? Or is Mitel's part a small enough part of their overall energy consumption that you guys don't really have to worry about it?

[00:41:19] [SPEAKER_01] Yeah. Well, it depends, Craig. In some places, it is, correct? So in countries that are environmentally conscious, in Europe, for example, so there is a major concern. In other places, less concerned because they care more about the cost. But certainly, energy consumption influences costs at the end of the day. So, and there is certainly an energy footprint of AI, correct? So you cannot ignore it.

[00:41:46] [SPEAKER_01] We see every day in the news that nuclear power plants are being restored and brought back to operations for the energy generation needs. So I know that large models, they run and demand significant amount of energy. So no debate. The question is, how much you need for your function, for your workflow?

[00:42:11] [SPEAKER_01] So we need to think about energy footprint for the outcome, not just for the large model size. So in other words, we need to right size the models for our use case using smaller, specialized models where appropriate. Because, yeah, we don't need always the largest language model for a particular function, correct? So it's a little bit too much for the task at hand.

[00:42:37] [SPEAKER_01] So leveraging edge AI also reduces unnecessary data transport because you have, as well, not only the cost of processing AI, but transporting data, et cetera. So edge AI for sure will bring value there. And we need to design architectures that minimize redundant compute. Basically, just perform the task in a single shot, in a single path, instead of trying multiple approaches. So I believe AI is so powerful.

[00:43:06] [SPEAKER_01] At the end of the day, in many cases, you need a fraction of that power to execute what you need in the enterprises. And if you can translate that into a model that minimizes the consumption of energy, you are helping with sustainability. You're doing your role, you're performing your role as a corporation that is conscious about the environment. I think that's very important.

[00:43:30] [SPEAKER_01] And I believe what we provide today with edge is an attempt to serve that. So that's what we're doing with our AI service platform to support customers that are looking for efficiencies and environment consciousness. Yeah.

[00:43:47] [SPEAKER_00] And when you talk about edge, you're talking about handheld devices in the field. I mean, not necessarily telephones, right?

[00:43:57] [SPEAKER_01] I have two scenarios, correct? So certainly you can execute AI at the endpoint, correct? So I can have my latest iPhone and I can execute some AI functions on it. But I'm also talking about bringing the language model to a local data center, to a premise cloud, because that's where I think the inference processing happens.

[00:44:21] [SPEAKER_01] And if you can bring that closer to the action, you are certainly saving effort, you're saving costs, you're saving energy at the end of the day.

[00:44:30] [SPEAKER_00] I see. Okay. That makes sense. Yeah. And so what are your plans to continue evolving AI offerings in the next two years through 2027?

[00:44:44] [SPEAKER_01] So at Mitel, we are moving from AI features to AI native architectures. I talked to you about our AI services platform. So over the next years, we're going to see deeper embedding of AI across more parts of the portfolio and with foundational integration. So agentic for sure, agentic workflow orchestration will be more mature.

[00:45:13] [SPEAKER_01] It will allow communication to trigger intelligent cross systems solutions or workflows across the enterprise in a more automatic fashion. We're also accelerating vertical specific AI solutions. But vertical at the end of the day is an area that our customers care a lot and I mentioned to you before. So we're accelerating vertical applications with AI solutions because generic AI is not going to solve industry specific challenge.

[00:45:39] [SPEAKER_01] So you need to customize the industry specific AI solution. At the same time, we have our hybrid platform that will continue to evolve and will strengthen edge capabilities. Edge in the sense, as I mentioned to you, correct? So premise-based AI capabilities for real-time intelligence and also to expand governance frameworks built directly into the architecture.

[00:46:03] [SPEAKER_01] So we want to have those in the architecture of our platform because that's how you can ensure that there is always the monitoring capability, the auditability, the explainability, and the human in the loop as much as needed for the particular vertical solution. So I think we want to make AI intelligence layer as default across the enterprise communications, but also evolving and taking advantage of new capabilities there.

[00:46:30] [SPEAKER_00] Up to five years ahead, which AI capabilities do you expect to become table stakes in enterprise communications?

[00:46:42] [SPEAKER_01] I would list two of them. One of them is transcription, correct? So real-time transcription is becoming more common across. In some cases, it's still premium subscription. But I recall in communications, it's the same as we had before with recording. Call recording many years ago, it was a premium feature. Today we can record everything. We're recording this call. It's not something that you pay extra for.

[00:47:10] [SPEAKER_01] So transcription and summarization functions will be everywhere. I think there was going to be eventually commoditized and just part of a standard subscription. So at the end of the day, I think AI-powered routing contact centers also will be default. So there's not going to be a contact center or a customer support infrastructure without that. And much of the functionality we see today as premium will be standard, will be part of the standard subscriptions.

[00:47:39] [SPEAKER_01] I believe that that's what we'll see progress in communications. And the other element, as I listed before, I believe voice interfaces will be the communication element. So you're not going to be browsing through panels of an application or a GUI. You're going to be talking in a very natural way to an assistant. And I'd love to do that because every time I drive to the office, it's 30 minutes.

[00:48:06] [SPEAKER_01] And I'd like to be talking to somebody or at least interacting, not talking to somebody, but interacting with my assistant and basically do anything that I want to do before I get to the office. That would be extremely a good productive tool for me, for sure. That's right.

[00:48:26] [SPEAKER_00] If a tech leader, a CIO, CTO, et cetera, wants AI to deliver real ROI, not just pilots, what do you say they should prioritize?

[00:48:41] [SPEAKER_01] Yeah. I think ROI certainly doesn't come from pilots, correct? So pilots do not always surface all the complexities of a deployment. So you need to prioritize and understand all the aspects of deployment. And there's a lot of hidden facts that you don't observe. So many companies run dozens of AI experiments before they align with measurable impacts. So what are the KPIs that define this will be successful?

[00:49:11] [SPEAKER_01] So many enterprises don't have it. The first step should be understanding what is the workflow friction in your company, correct? So if you're going to implement a workflow with 10 steps where AI can perform some of them, where are the delays? What are the manual steps? Where are the compliance risks? How you handle that in your workflow? I think that's what many, many companies don't see, but that's where the ROI will come from.

[00:49:36] [SPEAKER_01] So, and then once you understand the KPIs and how you address the friction within the enterprise, then you define the foundation. And many times today you're doing the opposite. You're coming and say, okay, let's do a pilot of something without understanding what is the value, what is the return of investment is bringing to you. And you end up not understanding what you're coming from. So resolution time, cost per interaction, revenue impact. So all of those are important.

[00:50:06] [SPEAKER_01] More important than some standard metrics that was proposed for you.

[00:50:11] [SPEAKER_00] Yeah. And you go into a company or you were saying that there's an advisory function with your customers. Do you go in and go through their workflows and identify, oh, this can be optimized with AI? And presumably you only do that for their communications infrastructure or workflows.

[00:50:41] [SPEAKER_00] Yeah. How does that work?

[00:50:43] [SPEAKER_01] Yeah. So certainly we start with the communications workflow. But as the communication workflow touches into various aspects of the IT infrastructure, we end up doing more than only communications. But the dialogue is basically a discovery model. So we sit with the customer, understands the use cases that they believe is not working well or requires improvements. Where automation can make a big difference.

[00:51:11] [SPEAKER_01] Where are the friction points? Where are the difficulties? Where are the manual steps? Too many actors in the enterprise that are working on some data. What are the errors that this creates? How you can avoid some of these mistakes that are made? So a lot of conversations about communication and communication integrated with the enterprise applications. That's how we start the conversation. So normally we do a discovery workshop with the customer. Understand that.

[00:51:41] [SPEAKER_01] And that allows us to put targets, objectives. What are KPIs that can measure business impact? What will be the return of investment of such of these solutions? And then we devise some of the pilots. And the pilots then will try to address that. Either with Workflow Studio to create a workflow with upcoming agentic features. Or by deploying a complete new implementation that is very unique to the customer that we'll create. So that's how we start the dialogue.

[00:52:10] [SPEAKER_01] And that's how we bring AI into the communications workflow of the customer.

[00:52:14] [SPEAKER_00] I see. And for you with a customer, what does success look like in their operations? I mean, how do you measure that? They say, oh, we love it. Or do you have metrics that measure things? Yeah.

[00:52:35] [SPEAKER_01] Yeah. Yeah. Certainly we put metrics in place. But some of them are precise, correct? Something that you can precisely measure. Some of them are qualitative, correct? So if you improve customer satisfaction in a contact center, that's based on scoring. Certainly there's the quantitative measure of the score. But behind it, you have the qualitative of customers are happy with our products.

[00:53:03] [SPEAKER_01] Our customers are happy with our services. So that's certainly one aspect of it. But in general, there are a few metrics, correct? So there is time savings. There is efficiency gains. Productivity improvements. But I think at the end of the day, the two main metrics is satisfaction. You want customers to be satisfied. You want employees to be satisfied. So the satisfaction is a measure of success that I truly believe in.

[00:53:33] [SPEAKER_01] And then you have the measurable metrics with time, cost, efficiency, productivity. Those you can measure very well. So we try to create work operating these two layers.

[00:53:46] [SPEAKER_00] Yeah, I see. Okay. Well, we're coming up to an hour. If someone wants to learn more about Mitel, where do they go?

[00:53:56] [SPEAKER_01] Just come to mitel.com online. So we have all sorts of information on our portfolio, on our services, on our customer base, or any references that you might want. So we are here to provide the communication solutions of the future to our customers and to any customers. So glad to have that conversation. And if you want to reach out to me, I'm on LinkedIn, Luis Domingos. So you can reach out. You can have a very good conversation.

[00:54:25] [SPEAKER_01] And see what can be a solution for communications with AI in the future of the enterprise.

[00:54:32] [SPEAKER_00] Okay. That's great. That's a good place to end.

[00:54:36] Thank you. Thank you. Thank you.