Tom Arbuthnot: Hi, and welcome back to the show. This time we have long-term friend of myself and the community, Kevin Keeler. Kevin is kind of a hybrid of analyst and a consultant, and gives us an interesting perspective on what's going on with Microsoft versus Cisco versus Zoom in terms of AI. And as usual, we have a pretty lively conversation about what's going on, the kind of convergence of UCaaS and CCaaS, who's doing what. Many thanks to Kevin for taking the time to jump on the show and the discussion, and many thanks to Momentum, who are the sponsor of this podcast. Really appreciate all their support of the community. On with the show. Hey, everybody. Welcome back to the show. This has been long overdue. Kevin and I have had some great conversations, off the show— but we finally found some time to get together and, talk a bit more on the record. So Kevin, welcome back.

Kevin Kieller: Yeah. No, thanks, Tom. It's great to be here. And yes, now we've at least. We're smart and we've hit record, so.

Tom Arbuthnot: Awesome. So, I think most people in the community know you and know your face, but just give us a kind of, a brief of, you know, your history and, what you're focused on at the moment.

Kevin Kieller: Yeah. So I, say that I'm, you know, part consultant, part analyst, and, you know, I really, you know, started and I would do, consulting for large enterprises. Really, you know, very heavily focused in the Microsoft ecosystem and, you know, love and know, and the Microsoft ecosystem way back from Visual Basic. I mean, I wish they'd bring back something that was as simple and, as Visual Basic. But, you know, and then of course did all the whole Lync and Skype for Business and Teams. But then increasingly I was asked by vendors, and sometimes, the vendors that compete with Microsoft, you know, to assist them with their messaging or some competitive positioning or competitive intelligence. So now I also consider myself a half analyst. In that role, as I said, I work directly with vendors, sometimes being a sounding board or saying, "Hey, I know you think this is a good idea, but look over here what these other people are doing." And sometimes it's help get their message out or help do message refinement. So, you know, I like the, the dual modes, 'cause hopefully that brings, you know, value, and I always try to, you know, be the person that's the no bullshit person and kinda tells it as it is.

Tom Arbuthnot: Yeah, I think you do a good job with that, and, you do a good job of keeping an eye on, kind of the Microsoft, Cisco, Zoom, the big players in our space and what they're all doing. You're always one of my go-tos for, like I hear they're doing this. What's the actual reality of that?

Kevin Kieller: Yeah, and as you know, I mean, it is challenging to keep up on the reality for one vendor, right? But, I think that, you know, I spend a lot of time, you know, kicking the tires and trying things out, and obviously it's changing very quickly. But, I think that we all, like, if we're IT pros, we all can learn, from looking at the other platforms a little bit 'cause then we can ask our vendors, "Hey, what about this?" And certainly vendors, I think, would be well-served by spending a little more time looking at some of the nuances of what their competition is doing really well. And I know, you know, we're gonna talk about that, especially in the context of

Tom Arbuthnot: AI. Yeah. And that's kind of interesting, isn't it? Our respective, beats or communities or content, however you wanna look at it, has not just. It feels like more than doubled in the sense of you've got to keep an eye on the UCaaS space, the CCaaS space kind of creeps in there as well, and now AI is intrinsically integrated as well, and the AI piece of this puzzle is just moving so fast

Kevin Kieller: Well, and you see, and I think it's interesting, you know, I. You call out the very, you know, importantly that UCaaS, CCaaS blending, and now we also see some of the AI providers, you know, everybody always jostles to be the top of the stack. And so, you know, people are asking a valid question, which is, you know, does unified communications go away because I start and end my day in Claude or whatever my, you know, or Gemini, and when I want to do a meeting, I ask Claude to set up the meeting and launch the meeting, and I think we're seeing, you know, you know, vendors like, for example, Zoom and their MCP server, you can in Claude, you know, do exactly that. You know? So Zoom is saying, "Hey, we'll integrate with all these other platforms, but we'll also allow your favorite AI assistant through MCP to access all the goodness of Zoom." So, you know, I think the. It's an interesting world and everybody, you know, wants to be the top of the stack, and it's not clear, right? What, who will own that. Yeah.

Tom Arbuthnot: I think that's really interest— an interesting point. Like the, is the core pieces of UCaaS, is it really just infrastructure these days? And I certainly find, like, you look at where Teams, Zoom, and Webex have gone, and the other players as well, like at their heart, the fundamentals are there or thereabouts on all the major platforms now. Like, yes, we have mobile apps. Yes, it works. Works. Yes, like they do transcriptions. So like, it, some of them have bells or whistles, but the development cycles are so fast now as well, it's not usually month to month if one has something the other one doesn't

Kevin Kieller: Yeah, and, you know, and I think that this applies, and has applied for several years to, you know, the UCaaS platforms. There was a feature race and it, you know, the features are growing faster than I think the adoption of the capabilities by the typical, you know, knowledge worker. So all the bells and whistles, you know, you may list all these things on the RFP, but the truth is, you know, for the core features of holding a meeting, all of the players, you know, the top players and even the, you know, kind of the second tier and third tier players, they all do a fantastic job, and most end users don't use all those bells and whistles. You know, you know, whether it's breakout rooms or some forms of annotation or, you know, fancy displays. They're nice, you know, for people that can leverage those, but for the vast majority of people, there's, you know, whatever, it's the 80/20 rule. I think it's more like the 60/40 rule, like 60% of the features get used.

Tom Arbuthnot: And I'd be surprised if it's 60% if you look at some of the feature stacks, to be honest. But, so let's talk AI. How would you summarize how AI has impacted kind of Microsoft's UC strategy, Zoom's and Cisco's?

Kevin Kieller: Yeah, I think, I think that there's, broadly, so I'll start there, broadly there's different philosophies that I see emerge from, you know, if we look at Microsoft, Zoom and Cisco. So, from, you know, putting on my analyst hat, you know, Microsoft went very quickly and obviously told every product manager to add Copilot and this is the typical Microsoft approach over time, and this has worked well for them. It's like move fast and add a whole bunch of capabilities. The capabilities were very different when it started, and some of them, quite frankly, were quite poor. And so Copilot initially in Excel was pretty terrible. But this is what Microsoft does well, is then they iterate and they bring it together in Work IQ. Now there's common context, which wasn't the case a year ago. And so, you know, they're now consolidating and bringing everything up to par, in a sense, where it should be, right?

Tom Arbuthnot: Yeah, I think that's fair. And the marketing was way ahead of the reality, which was the biggest challenge. It was a lot of vision and, like, then people would get their hands on and it wasn't quite there. And I think there is a bit of a challenge with the Copilot branding now, where it's like there was so much promise. Actually, things like Cowork are pretty impressive—

Kevin Kieller: Yes.

Tom Arbuthnot: But you've got to re-educate people that it's a different version with the Copilot label on, if you like.

Kevin Kieller: Yeah, so I'd say so, exactly. So Microsoft was like maximal deployment at speed, and then eventually they kind of consolidate it, right? You know, Zoom took another approach, and I think that Zoom has always, you know, taken a beat and spent a little more time simplifying. So, you know, at one point I counted in, like in a Teams chat, there was like 20 different icons that you could click at the bottom. They've, they've, they've filtered those a little bit, but I think Zoom had five or six. Because in my mind, Zoom said, "Hey, let's think about what people really need and, you know, make the interface a little more, accessible," right? So I think, you know, so Zoom and the Zoom AI companion, they also chose philosophically to include AI at no additional cost, which was different than Microsoft, where there was, you know, kind of they were trying to monetize, you know, $30 per user, per month for Copilot. - And, you know, so that was a different approach. I think, you know, people respond, you know, Zoom talks about people being, you know, Zoom happy, have happy meetings and whatever, and I think that simplicity, and the focus, you know, that does come through. So I think that's something that they do well that others should look at. And then philosophically, you know, Cisco, you know, they talk about how they've been, you know, doing AI for a very long time. But a lot of times what they talk about, what they mean by that, and it's true, but they're doing like, you know, the AI that just shows up. It's the noise cancellation. It's that when I stand up, it says, "Oh, Kevin left the frame and I'm gonna do a be right back." And then when I get back, their AI assistant knows I was away for seven minutes and it says, "Hey, do you want to be caught up on what happened in those seven minutes?" So they're really trying to do, you know, proactive AI to help drive adoption. They've got a very interesting polling agent that during, meetings it will suggest, "Hey, it sounds like you guys are debating this." you know, to the host it says, "Would you want to do a poll? Here's, here's the question and here's the options." Obviously you can override it, but once again, they're trying to do proactive thing, you know, AI W—

Tom Arbuthnot: Which is interesting, isn't it? Because that, is actually a solution to the kind of the problem of having the 50 features and the 50 buttons. If the AI could be smart enough to be like, "Hey, there's a lot going on behind you. Maybe we should blur your background," rather than it being like drop down, drop down, drop down, drop down, like that kind of stuff. Or, "Hey, you know, the, like, there's noise in your background. We're gonna auto AI filter it out, or we're gonna auto mute you." I mean, I'm still dying for, you know. Team Fireside Chat, we have 200, 250 people. Like, surely the AI can work out that person's not talking and just there's noise in their background and mute them. Like, that seems like a really achievable thing in 2026.

Kevin Kieller: Yeah, no, absolutely. And I think that's, you know, that's what we all hope for, you know, from our AI assistant or companion or, you know, whatever, is, you know. And I think we're, we're still working there, and the vendors are, you know, definitely trying to get there. I think that, you know, we're all faced with how do we use. AI does wonderful things, and yet at some times it's a big time suck because it seems like it's gonna get there. And, you know, whether it's. You know, initially image generation was like that, right? Now it does a good job. Like, it actually spells things right, and if you give it text, it includes the text. But for a long time it was so close, right? It was like— Yeah. But it, but it's 95% there, and you'd say, "Gosh, you know, this clearly it's gonna take, you know. It's gonna get there." But then eventually, you know, maybe a year ago you'd have to abandon it and just do it yourself if you really cared about the text. Now obviously, you know, that's not a problem. But I think in a whole bunch of areas, we've, you got two things. You've got, you know, with the Microsoft piece and with other People who've tried it and it didn't live up to their expectations, how do you tell a user that they should go back and try it? You know, I was at CommsvNext and was talking to a gentleman about, Copilot in Excel, and I was saying, "Well, now it's fantastic." And he's like, "Really? 'Cause I tried it". And he had tried it, you know, whether it was six, nine months, 12 months ago, and had concluded it was terrible and was never gonna go back and try it. Yeah. But I said, "Hey, you really should." Right?

Tom Arbuthnot: And I think that's, that's even more so with, like, if we kind of split people into, like, technical and business, like the business people are busy trying to do their actual job. They haven't got time to experiment— Right. For two hours to find out the outcome. So I hear the same thing a lot, like, "I gave it a shot."

Kevin Kieller: Some pe— "

Tom Arbuthnot: It wasn't good enough." And it's hard because a lot of AI S-. It'll get you 80% of the way there, and the 20%, you know, you can do yourself. But again, that's not a great pitch to business side people, 'cause they're like, "Well, just do it or don't do it." Like, it's a, it's a new mindset of kind of constantly experimenting. But I agree with you. A lot of people have been burnt by trying things early and not coming back to it.

Kevin Kieller: Right. Exactly. And I think another broad thing that is a different topic, but I'd be interested on your take on this, is I think we're seeing, you know— Pricing is starting to change because, all the vendors right now. AI is expensive. It's expensive from a training perspective, and it's expensive from an inference. So when Yeah. Create an image or have it summarize 300 documents, that inference cost is expensive to the vendor. And so all of the vendors are trying to figure out, well, how do they monetize this? And so, for example, just at the beginning of June, you know, Zoom introduced a new thing called Zoom Mate, which was focused on completion. So they have this, like, from conversation to completion. So you do a meeting, but after the meeting you gotta actually get your work done. And so the completion part, they have Zoom Mate, but Zoom Mate comes with AI credits, and so there's some consumption billing. In the Microsoft ecosystem you have that for certain agents if you start to, like, let external people use that or whatever, there's some consumption billing. And then Cisco, same thing with their agents. Some of it is, credit-based and some of it's built in. But I think all the vendors are trying to help people figure out what has value, and then if it has value to the organization, they need to figure out how to monetize it, right?

Tom Arbuthnot: Yeah. I think we're in for a big changes this. We're coming up to the end of Microsoft's FY, so it'll be next FY, so second half kinda the year. Like, there's no way they can keep going with things like Cowork and Scout just being included in a flat plan. It doesn't make any sense. And if you look at what's happening with OpenAI and Anthropic, they've, they've both gone from very big allowances to basically pure consumption for enterprise customers. So on the consumer side, you can buy your $100, your $200 plan, you can thrash it, and they basically treat those as loss leaders—

Kevin Kieller: Right.

Tom Arbuthnot: Because they're giving away 5X, 10X the consumption cost in that package, knowing that when people come to the workplace, they'll get the Work edition and the business will pay. Something has to happen inside of the Microsoft system, and it'll be the same for Zoom and Cisco. You can't forever subsidize all that inference cost.

Kevin Kieller: Yeah, no, absolutely. So I think that, you know, and obviously CFOs aren't gonna like the unpredictable cost. It was great, you know, when you would buy an E5 or a Zoom workplace or—

Tom Arbuthnot: Yeah

Kevin Kieller: You know, we haven't mentioned Google's 'cause obviously Google with Gemini and the, you know, the Google workplace.

Tom Arbuthnot: Yeah. Isn't it, isn't it funny we haven't actually, as you say that live, 'cause I'm like, it's, it's almost frustrating that they have a pretty good AI story and it all span out of DeepMind. They've got a, they've got an, an Office suite stack, they've got a meeting platform. Yep. I feel like they should be more front of mind. They've got all the component parts to do a really good job

Kevin Kieller: Yeah, no, absolutely. And I do think that they've got some super smart, you know, research people in the DeepMind and, you know, like a lot of this, all this, like, AlphaGo and AlphaZero and the AlphaFold for the proteins. I mean, they're doing some terrific stuff. And Gemini, quite frankly, is pretty fantastic. Like NanoBanana, from an image creation, it'Yeah. It was my go-to. You know, now I think ChatGPT has upped their game, and so Microsoft has inherited some of that. It runs slow. That's the one thing I'd say about Copilot. You know, I did some comparison across and, like, it's just it takes sometimes a really, what feels like too long of a time. Yeah. You know? It says it's lining things up, and then I'm not sure, but that takes a long time, so.

Tom Arbuthnot: Yeah, but this, and this is, again, interesting, isn't it? Because do you, do you. How do you scale your compute? You know, Microsoft have 450 million users on the M365. Like, they want them all using AI. If you wanna be highly responsive to that number of concurrent users, there's a cost there versus, you know, do you use all Nvidia's CPUs? Do you use your own chips in-house? Obviously, Google have their own chips. Microsoft now have their own chips as well. But yeah, I would agree with you. And experience wise, definitely there's times where, the new, Copilot Image Gen with the Image 2, GPT models, I can trust that now for text and social images. But you're right, you're like a set it off, go away, come back.

Kevin Kieller: Yeah. No, absolutely. So I think, yeah, we're gonna. You know, and also just recently in the news, it seems like this is a cycle where a whole bunch of stories are coming out with people, like, blowing their yearly AI budgets 'cause they encouraged everybody Yeah. They're like, "Use AI," you know, whether it's the developers or whatever. And, the one company, I, you know, it's a bit of a clickbaity story, but they somehow got a bill for, like, $500 million that was unexpected. And I'm like, somebody, you know, like, all these platforms allow you to—

Tom Arbuthnot: Yeah,

Kevin Kieller: Yeah. Put some caps. So somebody really, did not do their job, right?

Tom Arbuthnot: Yeah. It doesn't go from zero to millions overnight. Like, there's a, there's a dashboard.

Kevin Kieller: Yeah. No, exact— And most of these allow you, yeah, you can just, like, cap it out. But the thing is what I worry about is, and I think that there's an opportunity, if a vendor could come out with a fixed plan— And, you know, fixed cost plan, and then make that work. You know, it's kind of like how, you know, Microsoft calling plans. You know, you could pool the minutes and-. You know, when they started there was, like, that 3,000-minute plan, and, like, no one was gonna use 3,000 minutes. And so you kind of knew you're paying this, and it's effectively unlimited, right? I think that if somebody can figure that out, and then, you know, what you said, Tom, like, use their own chips or reduce their cost of goods sold, you know, that would be. That cost predictability for, you know, an enterprise, I think, that would be a winner, as opposed to consumption billing, where on one hand you're saying, "Hey, everybody, AI is gonna make our organization, you know, more effective, more competitive, so please go use AI." And then, but then you have to say with a little asterisk, "Yeah, but don't use it too much." Yeah,

Tom Arbuthnot: Yeah. Use it wisely. Yeah, Don't experiment, but don't experiment too much.

Kevin Kieller: Right. And that wisely part, when you try to go explain that, you know, 'cause I saw in one of the support pages for Zoom 8, you know, the Zoom thing which has consumption ability. You know, it's saying, "Well, write detailed prompts," and, avoid asking it to regenerate the entire document or. You know, so yeah, that's good, but, you know, I think we're still trying to get end users to even understand the basic part, you know? Not go—

Tom Arbuthnot: Yeah, don't give, don't give them a new job to learn, like, specifically how to have to drive it. I think the other thing is model selection is interesting. Like, I'll always go for, you know, the latest Opus model or the GPT Right. Where I can. But, like, I'm on fixed plans. Like, it, that isn't the most efficient use case. But I don't think real people want to have to understand which model to pick when, but if the organizations, if the software providers can auto-select the right efficient model dynamically, that should make it more efficient as well.

Kevin Kieller: Yeah. No, absolutely. Like, for a long time, you know, Zoom and Cisco talked about their federated AI approach. I mean, Microsoft is now bringing in the Anthropic, you know, model. They— they've also kind of started doing that. But it's like, wouldn't you think that, you know, an artificial intelligence, it should be able to decide? Like, is this a simple thing? I can use a small language model. I could. You know, I know there's even research being done to say, "Hey, what could I do locally Yeah. My user's machine?" And I think we're gonna see more of that, both for, you know, cost containment, for latency, and also just for privacy, right? There's a lot of, you know, in the rooms, you know, Cisco makes a big point of with some of the NVIDIA chips in their room devices, and it's not just Cisco, but, you know, doing a bunch of things on the edge, that's both fast, efficient, and it also means that things aren't going to the cloud that maybe— Yeah you don't wanna go to the cloud.

Tom Arbuthnot: Rooms are a great case for edge because you think about we're all going towards multi-camera and AI direction and things like that. Like, it's just not plausible to scale that stuff up to the cloud. Like, not for the individual company, but for the providers. Like, if Microsoft want everybody with Microsoft Teams Room, and I'm sure Zoom want everybody with a Zoom Room, you can't have four or five concurrent cameras streaming up to the cloud and it doing AI direction. And not just the bandwidth, but the AI direction means I need to be AI processing all those frames all the time. So I think we'll see a big push to AI on the edge in rooms because that's where it really ROIs.

Kevin Kieller: Yeah, and I think, and I think we get back to, you know, kind of the whole client server model. Like, there'll be some things that you're doing in the cloud and some things you're doing on the edge, and whoever figures out the most effective way to do that, I think that they win because it's a cost effective and it's a great user experience, right?

Tom Arbuthnot: Well, what's your take on the data conversation? 'Cause I think that's super interesting, is you've got obviously one of Microsoft's advantages with a Microsoft customer is they've already got you the SharePoint email, like a Teams with chats and meeting transcriptions. Zoom seem to be pushing more towards a suite story where they would have you mainly working in the Zoom ecosystem. Cisco are a little bit different, I think. You know, they seem to mainly be, we wanna be really great UCaaS, CCaaS, air gap devices. But it seems like who can get to the data, who controls the data is a big part of the value prop here.

Kevin Kieller: Yeah, 100%. That's a great point to bring up because, I think I wrote an article on No Jitter saying, I it was when ChatGPT-5 came out, I said, "Well, context is more important than having the better model," right? Because, you know, the models are trained on general knowledge, but I mean, in a business context, it's not just often general knowledge. It's like we want it to have access to our knowledge. As you said, for organizations in the Microsoft ecosystem, that's a big advantage. But one of the things that, you know, I think all the vendors are doing, Zoom certainly is like it has all these connectors, so you can connect it to Outlook, Exchange, both your email and your calendar, also to Gmail. You know, my Zoom is connected to Gmail and Outlook and OneDrive and—

Tom Arbuthnot: Yeah.

Kevin Kieller: Google Drive, so I use Google, I've got our family, so some more personal stuff, but it's all connected, and then they can connect to, you know, the Jiras, the Salesforces, et cetera. And you know, and Cisco as well uses, they use Glean, and they have other connectors as well, and now they've got a WebEx agent that's kind of runs inside of Teams, so it can access your WebEx meetings. But I think the challenge is, and I did a session at, CommsvNext, you know, How Copilot Lies To You, and I'm gonna talk about that at Commsverse coming up as well. The, and that's kind of tongue in cheek— Love that title. It's brilliant. But the, but the thing is when you don't understand the context that's being accessed, so for example, you know, Copilot makes, I mean, all the AI assistants, when you say summarize this chat, if you don't provide any guidance, it picks an arbitrary. Well, it's not arbitrary. It's kind of as given, but, I think it's still 30 days back Right? So similarly, when you say summarize all the emails that Tom sent me, it's gonna pick a period of time back, and I think one of the challenges with context, even if you have access to all this wonderful stuff, and even if you've built a semantic index so you can access it quickly, if you don't know how far back or what it's looking at or not looking at, like on SharePoint site there's a 300-document limit, you know, how far do you go?

Tom Arbuthnot: Yeah. I feel like it. There should be a rule where the UI has to surface, like, these are my plausible parameters, 'cause there was one point in Teams where it was different backless for meeting chat versus channel chat versus chat, and it was very confusing. I think it. But you're entirely right. Like, whatever the limit is, there is some limit. It's not going every conversation you and I have had over the last 10 years and summarizing it in that chat.

Kevin Kieller: No, and I think this is a place where, "Hey, all the vendors, take a look at what Zoom's doing," because when you access those other sources, there is a UI and it's kinda, like, it pops up and says, "Hey, for these meetings, is it this week, this month, this quarter?"

Tom Arbuthnot: Clever.

Kevin Kieller: And the other cool thing is because you can run into trouble when it accesses things you don't. Meetings you don't want it to, it kinda gives you a list and you can select all, but you can just say, "Well, wait a second. Oh, no, it's these last three meetings where Tom and I talked about this particular topic. Yeah, that other one we got off on a tangent, so-" Yeah."don't, don't include that." So I do think that AI, to your point, has to do a better job of saying, "Hey, okay, I'm summarizing this, but by the way, it's. I've limited it to 30 days," and if you want you could ask it or to do more, but it. We all, you know, being explicit, like, so context is really important, but it's, it can hurt you know, if it pulls in something out of date, right?

Tom Arbuthnot: Yeah, and this an interesting kind of, commercial challenge there. We talked about the consumption of the AI models, but, like, you look at what Salesforce are doing and ServiceNow and others, they're starting to think about are there API charges— Yes. And just to build Microsoft now Work IQ's gonna have an API charge. So yes, they'll build that semantic layer, but you're paying consumption to access it. Not if you use the first party AI. Like, if you use Copilot you're cool—

Kevin Kieller: Right.

Tom Arbuthnot: But actually if you're building your own agents because you're building ServiceNow agents or you're building ChatGPT agents or Claude agents, then if you want Work IQ you're gonna be charged, and it's gonna be the same with Zoom and Cisco. They can use Graph—

Kevin Kieller: Right.

Tom Arbuthnot: But does Graph stay free forever? I mean, it's gonna get absolutely thrashed if people are constantly querying. It was never really built to deal with that kind of use case, although it's, it's possible. It wasn't just envisioned, was it?

Kevin Kieller: No, and I think, you know, everybody. Yeah, it's, it's interesting 'cause everybody is trying build their own semantic index, right? Because without that, you know, so the semantic index is what, like, allows you to access things quickly because, you know, and it's not doing exact match. So it is conceptually. So, but if you don't have that, and if you're just saying, "Hey, SharePoint, you know, give me all these documents," and then you're trying to do summarization, you're gonna just, like, blow your brains out and your consumption— Yeah and it's gonna be slow. And so everybody wants to build and maintain a semantic index kind of inside their ecosystem, even for external sources, right? But that also takes storage and compute and. So it's unclear. Yeah, it's, it's unclear who's gonna win that battle, and I think, you know, no one wants multiple agents, that they need to query for different things and gives different answers and, but I think we're, you know, we're in for a little bit of a period of that.

Tom Arbuthnot: Yeah, I mean, the current, you know, the current hotness is kind of the OpenClaw, Hermes. Microsoft now released their own variant with Scout and Autopilot— Right. And I'm sure, sounds like Zoom are doing similar things where the agent can go off and do compute use and get to multiple systems and bring multiple things together. And it's a really good story, but if you're Salesforce and you used to charge X per user per seat, and suddenly an agent in the business is doing all this work— Yes. On the Salesforce platform, and you're not, you know, it's one seat, but, like, it's getting 50 agents thrashing it, that's a, that's a different model than what they envisaged, and it's the same thing with all the platforms. I think it'll be interesting to see how that works commercially, 'cause I think we're all gonna be using some form of proactive agent at some point in time. It just feels like they're getting useful enough that'll be a reality. But that model doesn't quite align to multi-vendors, multiple data sources.

Kevin Kieller: Yeah, no, absolutely. Let's just pivot a little bit because, to this other thing that I was just thinking about, which is, like, you know, I think vendors have over-rotated on AI, and now it seems like they're did a search or replace and say agentic. You know, I think everybody, and if you're an IT pro, if you're a vendor, if you're an end customer, you know, everybody really should focus on what outcomes, are gonna be important to your business or your personal brand or whatever. And then figure out if AI or an agentic agent, if that is something that you think is gonna get you there, because I think it's like in the early days of unified communications. People started being over-rotated about like, "Well, let's put everything together." And it's like, okay, but why are we doing that? Is it making our lives any easier, or is this just like some technical exploration? And I, a little bit with AI in context and MCP servers, and it's super cool, but everybody just needs to make sure that it's actually making your life less stressful and more productive. And in a business context, especially if you're paying for this, which you will-, you know, is it actually, you know, driving more sales or reducing my costs? 'Cause if it's not, I would argue that as agentic as it is, and as, like, cool as it is, it's, it's kind of distracting you from being successful.

Tom Arbuthnot: No, I think that's really valid. And it's really hard, isn't it? Because, I think it's pretty clear AI on some timeline will be impactful/of value. But with say UCaaS, to use your example—

Tom Arbuthnot: I've got a couple of, you know, a few thousand users. I've got Cisco Call Manager, ISS, Webex, and Lync or Skype at the time. I'm either ready to jump or I'm not. Nothing really changes for the next three years. But in AI, like, I can't assess it and then wait for another three years. Like, I kind of have to go, "Well, is it ready now? Is it ready just for the marketing team? Is it ready for the salespeople?" Like, it's, it's moving so fast that where's the business value conversation is not a one-time activity. It's almost like a constant assessment.

Kevin Kieller: Yeah, and I think, you know, you bring up a good point and, you know, to lean into that. I think, organizations should really make sure that whoever's experimenting is doing a good job of sharing with other people in the role. Because one of the things that I love of in terms of talking to other technology people or even, you know, just other end users, is everybody's thought of something that I haven't thought of in terms of how to use AI. Now, it may or may not be applicable, but sometimes people go like, "Oh yeah, I used AI to do this." And I'm like, "Wow, that's a brilliant thing. It really could help me be effective." -. And I would've never thought of it on my own, right? And so, you know, marketing professionals find out, you know, figure out a way in your organization so that they talk and share amongst themselves, and sales professionals and HR professionals, because it's really the specific use cases, not really kind of the horizontal stuff that I think really, - Yeah drives the measurable business benefits.

Tom Arbuthnot: I think that's really good advice. We do a lot of roundtables, and we've got a legal AI roundtable, and that's been-. Really interesting because if you're, you know, obviously competing businesses in some senses— Yeah. But if you can talk to your peers that Yeah the, you know, departmentally aligned but across different companies, that's really helpful. And also, all the vendors are pitching everything is the next coming, right? So it's a good to get a reality check on, "Oh yeah, we tried that and this worked and this didn't," but, and as you say, like new ideas of how to make it work as well.

Kevin Kieller: Yeah. No, absolutely. So all told, I mean, we've covered a lot of topics. Like, obviously this is an exciting time, for, you know, for technology people. It's also, you know, I think it's a stressful or challenging time for IT pros. You know, end users have to keep up, but then the IT pros as well are struggling to keep up so that they can both advise and answer questions from their end users. And I think things are changing at least faster than I've ever seen it. How about same for you?

Tom Arbuthnot: Yeah. Like, for sure faster, and particularly following Microsoft. I've never seen Microsoft internally moving this fast. Like, you know, the Autopilots is a great example. Like, OpenClaw was, I think, a November that was an open source project, and now we're in June and Microsoft have their own variant of it, and that's like insane in Microsoft terms to have something— Right even in preview that speed. And it speaks to where the market is, and I think you're right, it's a super exciting time for IT because we're/AI are relevant to the business. So it's important, it's impactful, it's gonna change things, but also that means you're moving into advisory, security, risk, choice. Like, you're being looked to have an opinion on where things are going and what the business should pick, which is—

Kevin Kieller: Of course.

Tom Arbuthnot: To your point, yeah, quite hard to keep up. But, I'll, I'll take the, I'll take the stress with the fun. At least we're, we're super relevant at the moment, which is a good thing.

Kevin Kieller: Yeah. No, and it's interesting, you know, and even IT's, role is changing 'cause a Commsverse session that I'm co-delivering with, Tristan Gayford from, EY, that's gonna Oh, great. Oh, that'll be

Tom Arbuthnot: Awesome. Love Tristan.

Kevin Kieller: Well, yeah. So we're, we're looking at, okay, how is AI changing the role of an IT admin? And increasingly, what happens when your end users can, you know, use, you know, Copilot or Claude or whatever to do things that you used to do using your— Yeah, does

Tom Arbuthnot: The, does the support ticket come with the answer steps? They just can't do it 'cause of RBAC, but they're like, "You need to click this, click this, click this, click this, please."

Kevin Kieller: Yeah. Well, and all these PowerShell skills that, you know, people have honed and, you know, now you can kind of ask Claude to do it, but then eventually you just ask it to do it, and it does it, you know, under the under the covers. So, you know, the role of an IT admin or a Teams admin, - Yeah. That's good. That's a great topic. Is really is likely to change. So we're gonna explore, you know, some concrete examples of, you know, and, what it can do. There's a Teams agent, admin agent, and it can do some things. And so, I think that's, yeah, it's, it's interesting. But, yeah. Fun. Yeah. Well, I,

Tom Arbuthnot: I look forward to seeing you at Commsverse, and that sounds like a, yeah, definitely a talk I'll be at 'cause that is. You're right. That role is changing so radically. And, - It's-. There's a lot of, a lot of benefits to it, but also— It's-. There's definitely probably some fear there of like, all my skills that I've honed over the last five, six, seven years are now inside of, Copilot/Claude/ChatGPT.

Kevin Kieller: Yeah. No, absolutely. Hey, as we wrap up, Tom, I just do wanna kind of, you know, thank you and just, like, call you out for just, like, you've built this great community of bringing people together. And you know, I think that's, really important in these times with things changing. I mean, you hold a lot of sessions where I know they're not, you know, some of them aren't recorded, but you get hundreds of people, you know, who can talk freely about things that are working. You know, bring some great vendor representatives from Microsoft into those. And anyway, I find that super useful, and I just wanted to publicly anal— acknowledge that.

Tom Arbuthnot: Thank you, mate. Yeah, I appreciate that, and appreciate likewise everything you do for the community, both, you know, public and behind the scenes. But yeah, it's, it's great to bring people together, and it's, a privilege, although sometimes a stressful privilege, to be part of bringing everybody together.

Kevin Kieller: Yeah. No, well, so thanks for what you do, and, yeah, I've enjoyed the discussion. Thanks for having me on.

Tom Arbuthnot: Awesome. See you again soon, Kevin. Thanks a lot.