Tom Arbuthnot: Welcome back to the Teams Insider Podcast. Really excited to have Tristan from EY on the podcast for the first time. We go through a little bit of his history in the space, a very similar journey to me and a lot of people in the community. And then we get into talking about the modern management of Microsoft Teams, and also Copilot and consumption, and how you can deal with those things in large and complex organizations. Many thanks to Tristan for jumping on the pod, and also many thanks to Landis, who are the sponsor of this podcast. Really appreciate all their support to the community. On with the show. Hey, everybody. Welcome back to the show. Really excited to have this conversation. Had a lot of good offline conversation with Tristan, most recently at a round table. And fortunately, we've managed to find a slot we can both do. So, Tristan, thanks so much for coming on the show.

Tristan Gayford: Yeah. Thanks for inviting me.

Tom Arbuthnot: Awesome. Pleasure to

Tristan Gayford: be here.

Tom Arbuthnot: Yeah. So, give us a little bit of an intro to yourself and your current role, and then we'll kind of go back through your journey.

Tristan Gayford: So, yeah, so my, my current role at the moment is I'm a Teams Product Manager within EY. I've been working in kind of the Teams, Skype, Lync, OCS space since I think 2008. I, I, I joined Deloitte, 'cause I like professional services organizations it seems. And somebody asked me to do a training course on OCS, and so I said yes, and then Googled what on earth OCS stood for, 'cause I just wasn't in that space at all. And the rest has been, an interesting history since then. For, for my IT career from, from that point at least. Prior to that, lots of Exchange, lots of NT going all the way back to when my parents got me a, a... didn't get me, actually got our family a ZX Spectrum 16K. So that was my first computing experience, and I've been in computing ever since, basically.

Tom Arbuthnot: Amazing. And I like, I'm sure most people listening know, know EY, but over 400,000, I think, employees at the last count, like one of the biggest companies in the world, and very proactive in the technology space as well.

Tristan Gayford: Yeah, absolutely. I mean, we, we as an organization, we're very diverse. We're in over 100, you know, well over 100 countries. We're supporting people who are supporting their clients. So it, we... it's a very interesting role. It's a very interesting organization where you have challenges around, you know, the countries themselves, the, the technology that's available in those countries, regulations, the, the, the business that we're in ultimately. So it makes for a very, very interesting role and a very interesting experience if, if not very challenging as well.

Tom Arbuthnot: Yeah. Yeah. Definitely on the, on, on the frontier, if I can use that word, of some of the most challenging requirements and, and people that want to use the technology as well.

Tristan Gayford: Yeah, absolutely. And tha- and that's the thing, is that when you've got a, a load of consultants and they're in technology, they, they want to be on the bleeding edge of things because that's what they want to show their clients, and their clients are demanding that. So it, it does make for, for interesting challenges where you're trying to support both people supporting their clients and also supporting the wider business that goes from absolutely everybody that's, you know, within internal services like technology all the way through to those people that are serving their clients and, and client-facing.

Tom Arbuthnot: Awesome. So you've been on the same journey that, a lot of the people in, in the community have back, back the OCS days. Take us, take us through that from what is OCS to managing Teams in the biggest environment in the world.

Tristan Gayford: What is OCS? Yes. As I found out, Office Communications Server. Yeah, so, yeah, we... I got involved in that one that there was very little IM. I mean, it, it kind of appeared within Exchange back in the LCS days. And, you know, with OCS w- we were kind of trying to understand how can we create instant messaging, and this was not something, back in those days, it wasn't something you rolled out to everybody. You know, people still used- Skype, you know, their, their consumer product as it were, was at the time, and, and even Facebook. And, you know, go through that journey of OCS and then k- starting to integrate telephone systems into there, doing federations. You know, you kinda just increase that footprint. And then into Meetings. I mean, Meetings wasn't even really, you know, had the live, live meetings at the time. So it's kind of an interesting journey of, of technology that was obviously all on-prem. You know, we were thinking about, you know, having to think about licensing and how the licenses were being consumed, who wanted a license, who needed a license all the way through and, and, you know, taking that forwards into, you know, Microsoft did like their renaming at the time into Lync, and, and a slightly more mature, slightly better model, particularly around telephony, particularly around, and this is the time that we, you know, made sure that it was available to all into Skype and then Teams, you know, suddenly appeared out of nowhere almost. And, the rest is history from, from a Teams perspective... and how Teams has kind of integrated that into kind of the, the operating system of your business day.

Tom Arbuthnot: And it's quite interesting on that journey, we went from server to cloud, and I, I did some work with Deloitte, I did some work with EY. Like, like, like these organizations at the time, going back, were very change control forward, very careful planning. E- each CU- went through this huge process of testing in the QA environments and like, like talk us through your perspective on some of that 'cause it was amazing to see th- all organizations, but those types of organizations go from a, "We patch the servers," to a cloud run model.

Tristan Gayford: Yeah, and it's very difficult because Because with the, the nature of professional services, you have a, effectively a bunch of firms that all look and sound the same. But they're managed as a partnership, and the, the challenge with that is they have their own cost centers, they have their own budgets, and they have their own regulations. You know, different countries have different regulations and, and, and different readings of those regulations as well from risks Teams and from the clients. And so you, you end up with this challenge of where where is the data? And, and back in the kind of early days of cloud, that typically meant it was in the United States and maybe Germany, and that was kind of it. And, and so people just didn't want their data to sit somewhere else to, to traverse the internet. It was very unknown and very... Yeah, there was a lot of hesitation around it. And yeah, I remember hearing comments around, "Yeah, but the clients don't want the data outside of that." And then you kind of look at it and say, "Well, the clients use the cloud, so they can't say that we can't." But there was a lot of hesitation around it because it's very new and un- and, and you've gotta be very careful in these businesses. You've gotta make sure you know where the data is, where it's residing, who has access to it. So you know, we definitely weren't at the bleeding edge when it came to cloud consumption and, and moving into the cloud. But you know, moving from that model of patching, going through cumulative updates, going through service, service packs, and, and new versions and, you know, that is not something that I think anybody misses. I'm pretty- Yeah... certain about that. A,

Tom Arbuthnot: a gi- a giant expensive project to patch the SQL Servers for no real business outcome other than we, we all spent six months patching the SQL Servers. Exactly. Exactly. It, it's...

Tristan Gayford: Yeah, definitely don't miss those days.

Tom Arbuthnot: Yeah. Yeah. It's funny to think back 'cause it all seems, cloud seems so normal now. I, like, just it's flashing back to various things there where it was... And obviously these were critical comms infrastructure, particularly when it went to Telephony. So it was very serious infrastructure to be managing and patching and upgrading and, you had to follow the SQL cycles and the Windows cycles, and obviously the, the major versions of the releases as well.

Tristan Gayford: Yeah, absolutely. And I, and I think a, yeah, a lot of people maybe just have parked that memory. But it, it was a challenge. You know, you, you had to get it right. There was no You know, whoops, got this wrong. Yeah. Let's, let's just rebuild or anything like that. You know, just fail it over to another server. You, you had to be so care- careful around and around. We, we spent a lot of time and effort making sure we had resiliency, you know, across the SQL servers, across the, the Lync and the Skype servers, across data centers. It was a lot of hard work to do that, but it was important because your business runs on this. You know, your telephony runs on this, your contact centers run on this. You, you gotta get it right and you, you, you know, there is no room for mistakes really

Tom Arbuthnot: So in, in these bigger environments that you've worked in both now and in the past, what does it look like in terms of architecture and operations? Like, obviously share what you can share, but like it's... We, we were at a, a roundtable last week, and it was really interesting. Different organizations of all sizes around the table and the very different architecture operational support models.

Tristan Gayford: Yeah. I think, I think the, the models themselves are not that dissimilar, it's just the scale of them. So rather than someone being an Architect and a Product Manager and Operations Lead, you, you have very distinct Teams. You know, there, there's a wider picture. You know, an organization like EY, for example, naturally has a lot more software, you know, going on than maybe a smaller organization. You know, it has a lot more people to be able to support that because it needs to. It needs to... You know, there, there's... 24 by 7 is, you know, is even more important in an organization like us because we literally do have people working every day of the year, you know, in different countries. So, you know, the, the, the way we structure is predominantly the same. We have the same types of roles. We have those architectural roles, we have those product management roles, the support roles, service management roles. It's just the scale of it is much bigger. And obviously those groups are very defined. And that, that helps. I mean, in some cases you could argue that that kind of delays progress, but I think it delays it for good reasons. It's, it's not just delaying it for the sake of delaying it, it's delaying it because we've gotta be careful with this stuff and we, we can't just go in there and make, you know, changes, just react to something just because a vendor has said something or a client has said something. We, we gotta make, you know, very conscious decisions to make sure we're doing this for the right reasons.

Tom Arbuthnot: Nice. And I wanna get into kind of Copilot and, and Frontier 'cause, it's all, all public, isn't it? EY have made a big bet on E7 across the board, so you've got a whole bunch of new capabilities there. And the Copilot stuff obviously spans all the workloads, but very heavily in the, the, the Teams and comms and collab space.

Tristan Gayford: Yeah. And, and you know, we've been moving with Copilot since pretty much day one. You know, gradually expanding it out as, as different locations, you know, get comfortable with it and, and understand what AI can deliver for them. And yeah, Teams is a, a key driver of that particularly in the meeting space and, and recaps and meeting notes. So we, we've had good experience with that and we've got a good relationship with Microsoft as well, in kind of developing and understanding where their thought processes are. And I think it's, yeah, for us it's an important aspect because, you know, as you can imagine a, a- an organisation like ours, you know, we have quite a few meetings. People are in meetings all day, every day. W- we want them to be more effective and, you know, spending time, you know, recapping the meeting, you know, verbally at the beginning of the next one, trying to remember what people said even just, you know, scribbling notes down, you know, that's not the most effective way. So yeah, having that capability, whether we do it through the kind of temporary transcription during the meeting or after the meeting with, with retained transcription, allows people to actually be a lot more effective, to give those notes, to understand who's got the actions and, and there's, there's accountability there, which is, you know, a, a great advantage to our people to be more effective when they're speaking with their clients or even internally as well.

Tom Arbuthnot: Nice. And how... Have you found your role has expanded more into the legal thing around, like you, you mentioned retained transcription versus transient, and can we record here and can we record there? Like, back in the day, like OCS, we were like, "Well, it's all just transient and maybe we keep chat history." And then it was like progressively, oh, now it's Collab. So, like w- what's that like today? 'Cause there's so many different regulations around the world with the recording, transcription, and AI.

Tristan Gayford: Yeah, exactly. And consent as well. You know, that, that... They all come together and they're... There's, there's, there's different perspectives. So there, there is a regulatory requirement in certain locations. There is, you know, a general risk appetite, so different locations have a slightly different risk appetite based on their clients and the, the business and the service line. And you also have clients themselves who, who, you know, we want there to be a documentary evidence of everything that we've spoken about or obviously the, the opposite of we don't. So trying to meet all of those is, is a thankless task. But, you know, giving people the, the tools that allow them to make those decisions, giving them the guidance, that's what's important so they, that everyone can understand, "Okay, this is what I can do or should do in these meetings. This is what I've got. You know, whenever I click on Copilot, everyone's, you know, microphones gets muted. This is why we're doing it, and we understand that." So, you know, that change management piece and the, the adoption piece is super important because otherwise it just looks like IT's just throwing in spanners into the works. So it's really critical that we get that, that piece right. But it's, it's an interesting challenge, you know. And w- you know, I have conversations, and as an organization we have lots of conversations with, with the, the, the risk folks, the privacy folks to make sure we're ma- meeting their requirements and meeting their needs.

Tom Arbuthnot: Yeah, it's interesting. I, I guess that's one of the privileges you have in a larger organization, is you've got Teams of experts in each, each areas. And, and we were saying this in the, in the roundtable, like some of the smaller organizations have the exact same pressures and, and requirements, but maybe not the SMEs to lean on in some of these scenarios.

Tristan Gayford: Exactly. We, we have to take, you know, from, from my perspective, I, I need to be, to consider that and be fortunate that I've got a lot of people that I can lean on, have those conversations. And we have very open conversations. We may not a- always agree, but we have good conversations to understand the perspectives and make sure we're delivering the right solution for our people. 'Cause ultimately, that's what we're trying to do. You know, if... We, we could be easy and just say no to everything, but no one wants to do that, so we want to make sure that we, we're doing the right thing because we know that, you know, people can grab their phone and record if they really... really wanted to. We don't want them to do that. We want to give them the right resources so they can be effective internally and externally.

Tom Arbuthnot: I've definitely seen in some of the organizations we talk to that the AI capabilities have driven the recording and transcription conversation. Like particularly in legal vertical, like it was just a default off by default a few years ago, and now it's like, "Oh, we want the power of the AI summary, therefore we are gonna look at transcription now."

Tristan Gayford: Yeah, I think that- that's an important piece. You know, at the end of the day, we want... It's our data, we want to retain that data whether it's a record. It's, it's not to kind of go back on it and look at, you know, Tristan said this. It's to understand and, and be able to create notes that are, you know, quite frankly, a lot more accurate that, you know, than, than most people can take notes. So having that capability, and you're right, it, it is driving, a lot of the transcription. The only during option for Copilot has been a absolute godsend for us, because retained transcription does have, you know, still retains a bit of a risk exposure that a lot of people don't really want to have. So the ability to actually use Copilot during the meeting, you know, at the end of the meeting, quickly capture those notes, copy them out so you've got them, that, that's a really big thing that we, we use and leverage- Yeah... across a lot of our locations.

Tom Arbuthnot: Yeah, it's been interesting watching the roadmap and, the features coming through. Like things that I've heard companies ask for around recording and transcription, like consent and these granular policies and transient, like those never were top of the pile previously. But because of the, the Copilot demand, we're getting a lot more granular controls there, which is great to see. Like, like even things like, you know, down to per user stuff because this, this country's going to allow it and this country's not, that kind of thing.

Tristan Gayford: Yeah, exactly. And, and that's, that's the important thing, is that we want people to be in control of what they're doing. If they don't want Copilot enabled at all in their meeting, that ability to say nope is great. You know, we, we want that. Yeah. Because there are justifiable reasons why that's gonna happen. You know, for example, if you've got a, a HR conversation or, you know, something like that, or one-to-one, you know, maybe Copilot's not suitable for that and, and you want to make sure it remains off. You know, having that option at the user level is really important. We, we just wanna stop where it gets to. So, you know, maybe we don't want the retained transcription, but we don't wanna stop people making decisions that are right for them. We want... You know, we, we got a lot of very, very, very clever people. They, they're happy to take that accountability and, and do what's, what's right.

Tom Arbuthnot: Nice. We were talking last week about Cowork and AI in general going to a consumption model, and you had some really interesting takes on this. We're definitely at a point where lots of organizations are playing with either Microsoft's Cowork or other AI tools that are consumption based, and how does that work with developers versus how does that work with knowledge workers? Give us some of your perspective on that.

Tristan Gayford: Yeah, it's, it's a challenge. You know, we've, we've, we've had a IT funding structure for many years now, which is very user-based. You know, every user has a, a, a license, you know, an E5 or whatever it may be. E7 now. E7. We'll see what we get, an E9. But yeah, we, we've had that model where we know how much it's gonna be. We can predict that. We can predict that over a couple of years almost, you know, depending on your license and your contracts. So it's very- you know, jarring almost that we, we're suddenly faced with this, okay, we're spending all this money on these licenses, and now there's this unknown cost that's sitting on top of that. And I think all organizations are, are initially gonna struggle with this. You know, if you, if you go back a number of years now with mobile phones and telephony, you know, well, the, the, the number costs this much, and then the per minutes cost that much, but who's gonna charge that? How are we gonna get that? And we kind of fought effectively for, you know, combined minutes and plans that we can just, like, consume at, at a fixed rate. And then that was the technology cost 'cause that was quite simple, maybe an office cost. Because that was quite simple. But with things like Cowork and, and some of this other consumption billing in, in AI, who is w- w- or whether where is that cost going to be? Because, yeah, ultimately it's the technology group that signs the check, but it's not the technology group that's actually driving that, that, that usage. You know, if you've got 100,000 people, you've got some people that are leveraging Copilot, you know, extensively. You've got some people that maybe are not. It all kind of balances out, but we've got a, a standard, standard rate on that. But when you add that consumption billing in, you know, you, you... Who is going to have to pay for that? Because as a technology function, you know, the, the, the costs are, you know, the costs are just astronomical potentially, and unbudgeted, and

Tom Arbuthnot: unpredictable. Like, like hyper, hyper variable. Well, aren't they? Like, telephony's a great comparison. Yes. I was at an industry event for telephony. I was like, "You're the industry that understands talking about it, it depends on consumption, but it's somewhat linear." Like, like my... I can... As a CFO, I can see my worst case exposure is all the salespeople are permanently on the phone, which from a sales org point of view, we'd love that 'cause it'd mean they're talking to customers, and this is my, like, upside risk. But with AI, it's potentially, you know... I, I did, to push Cowork as hard as I could, when we were still in the frontier, I, I did a massive single query and got up to $270 on one query. I mean, the output was amazing. Like, it was... I could justify the ROI if I had to go to a CFO. But how, how do you do that at, at any kind of scale? Because is that a great spend or is that a terrible spend? It was bud- it wasn't budgeted. So, yeah, whole new world.

Tristan Gayford: That's the thing is that... 'Cause you, you've got this situation where you have, you know, even b- Even with Copilot you, you have it. But with Cowork and, and obviously the, the consumption billing is, is more prevalent. Yeah, I can sit here and I can do a load of stuff in Cowork and it'll cost me $100 a month. Now, whether that output is any good or not, that's a completely different matter. So you know, we're kind of almost getting to the point of, you know, justifying that, that output. And you know, we kind of do that in a day-to-day job perspective. We justify what we're doing, and they pay us, and everyone's happy. But now we've got this kind of additional cost- But there's,

Tom Arbuthnot: there's a whole structure there, isn't there? Where there's a manager and a manager and a manager, and, like, ultimately you're just- you're, you're justifying your existence slash you're doing what your manager says, and then your manager's manager. And, like, that's been... that's now very well established. Like, there's some C- CVP of marketing who owns the marketing division, who's ultimately accountable for that spend and the outcome. I feel like we end up in a similar way with the consumption cost. It's almost like having a- another employee cost. Like, I'm, I'm spending this much money on tokens, and I'm... as the, as the CVP of marketing division, I'm happy that that 200K a year, half a mil a year, 100K a year, whatever it is, I'm happy that number is driving an outcome. But then you work that back down the chain, how do I know whether Sarah in marketing spending £1,000 a month on tokens is amazing 'cause she's basically doing three people's jobs, or is terrible- Yeah 'cause she's just generating GIFs all day that we don't even use?

Tristan Gayford: Yeah. Exactly. And that... I think you, you summed it up. You know, Cowork is an employee. You know, my Cowork is my employee. I have to justify that cost. I have to justify what I'm getting out of it to my manager, who then justifies it further up the chain. And I think, you know, that's ultimately where we have to be. It's a business cost. You know, just like IT doesn't consume the, the people cost, for, for across the board, or HR doesn't consume those people costs across the board. I think that's what we have to do, is the, the business has to look at it and say, "This is valuable to our part of the business, the work that we're doing." Therefore, we're okay with spending this much money on credits and, and kinda keep that into the business. You know, it's a different model to, to what we've experienced within technology for, like, say, since, since kind of the earlier days of telephony. And even then, I think, it's not necessarily something that even the, the functions were using. So this gets really granular really, really quickly.

Tom Arbuthnot: Yeah. Super interesting. And, and, like, you can make comparisons to the developer platform side when we went from all running our own service to kind of Azure and Cloud, where, like, now it's a variable cost for my SQL service. But it was... Those costs were all insular to IT, so you could at least make the case that, like, well, the servers used to cost me this, and I used to pay X, Y, Z outsourcer to patch them, and now my upside and downside is this. But again, you could, you could, you could broadly predict. It's, it's hard/impossible to predict what the business users will be doing with this, and not even today, but in the future as well, 'cause presumably it gets more adopted and more powerful. So is there more use or is there... Do the models get better and there's less use? We just don't know at the moment.

Tristan Gayford: Yeah. And I think that's one of the challenges, is that we... You know, for example, we don't know how much this prompt that I've carefully developed is gonna cost. I don't know what model it's gonna decide to use. I have no control over that. So it's not even the case of I know what it's gonna do. It's a case of, well, after the event, I've now found out that, you know, in your example, this prompt has just cost me $250 or whatever. You know, it's, it's, it's after the event that you find that out, and, and that's really gonna be d- very difficult. And that's... I think, personally, I think that's the answer that we need to kinda get to, is how do we still... We want to predict those costs. Yeah. I know how much someone's gonna cost me in, in terms of their, their output and their efforts and their work, you know, when we're talking about a kinda human. But when we're talking about, an AI that- You know, a, a misplaced word or a, a, you know, a slightly incorrect phrasing of the, the prompt could cost you hundreds of dollars because it's going off in a complete tangent. That's, that's a problem. And, you know, we, we need to get to a point. You know, we can understand that we can't predict because these are consumption models. We, we understand that, but we do need to get to a point where there's a little bit more transparency up front in terms of how we, how we're gonna be charged for these these prompts that we're creating.

Tom Arbuthnot: Yeah, I think that's a really good point. And, and we're in a weird spot at the moment in the adoption curve where for a while we were like, "Go crazy. Experiment. Just do stuff. See what happens." And now we're like, "Experiment, but don't experiment too hard," and like, like, like, try... And, and I think it's crazy that we're in a phase where we get, like, we're trying to teach business workers to pick models. It's like, like, like, like you should understand what Opus is versus Sonnet is versus Fable is and pick the right one for your use case. It's like, well, I mean, two things come to mind for me there. Like, if I'm a business user and I don't own the budget, I'm gonna pick whatever you said was the best one every time because why would I not? And then it's like how would I know if, if querying a CSV is a hard job or an easy job? Like, like pick, picking my model there.

Tristan Gayford: Yeah. I know. I, I have that problem. You know, you kind of get that, "Oh, well, let's slide it up to ultra high. That must be better." You know, it, it... That's the thing is it's very, very difficult, and there is no guidance on it. And I know we're all told, you know, "Just choose auto. It'll figure it out." Again, I still concede that, well, if you can figure it out from auto, then you should be able to figure out roughly how much this is gonna cost me. But yeah. Well, I don't know which one to choose. I don't... You know what I mean? I- Do I want something that makes a nice pretty deck, or do I want something that analyzes that Excel really well? I, I don't know what one to choose and, and I'm just gonna choose what I think is the best one. And, you know, maybe that'll cost me more, maybe it won't. I... Again, I have no idea. There, there's, there's that opaqueness around it- Yeah... that, that's challenging us all now.

Tom Arbuthnot: Yeah. It's fascinating times. And i- it's, it's both exciting and kind of challenging in equal measure 'cause you see what... Like we've just enabled the, the, the Fable models in our environment, and it's like th- th- it is a, a, a step forward again in some of the stuff it produces, but we've got nine people, so like I can cope with understanding where the costs are going. It's a little bit different when you multiply up.

Tristan Gayford: Yeah. I mean, even during our pilot phases, it, it's been an interesting journey getting that data. Even getting the data so you can say, "Well, where am I with my credits?" I can do /cost, obviously. But that's that prompt and, and that, that conversation. You know, what, what, how, how far am I along-

Tom Arbuthnot: Yeah...

Tristan Gayford: this year? We, you know, we, we're getting that information now, thankfully even through to /cost. But yeah, it's, it's still... I mean, I, I do it myself. I do /cost. It's like, oh, okay, I'm over 50% of the way through the month, you know, already, and I don't know what I've done that's, that's really cost that much. But- Yeah... it, that's, that's what it's saying. So, you know, do, what do I need to change in order to bring those costs down, if I want to bring those costs down? That's, that's the challenge. I, you know, I, I have a number of tasks that are automated. They do some great stuff for me. Are they worth that? I d- Yeah. Will someone look at that and go, "Well- Yeah... come on, seriously, do you need to spend all that money on that? Can I change that to make it cheaper or less?" You know, "Do I... Should I change the model? Should I do this? Should I do that?" These are all very difficult questions to answer at the moment. And, and but they are... You know, if you multiply that out by, you know, thousands, tens, hundreds of thousands of people, they are very serious questions that we need to answer.

Tom Arbuthnot: Yeah. And you, you mentioned about like trying to think more about where you can use Copilot, 'cause that's, you know, f- fixed cost. And it's interesting, when Copilot first came along, we were like, "Oh, $30 a user, that feels like a lot." And now suddenly it looks like a great bargain, 'cause you're like, "Oh, actually I can sensibly... Like, I can thrash this because there's no consumption on it, and I can tell my users, try as much as you can." It's funny how our perspective has changed over the last three years.

Tristan Gayford: Oh, definitely. I mean, even at home when I, I use ChatGPT, and, and every so often it says, "Oh, we should try this in Microsoft Word." It's like, "No, I don't want to," 'cause there's, there's credits in there. You'll run out and- Yeah... you know? Yeah, it's, it's some becomes this thing of you're, you're very conscious about it. But maybe that's a good thing. You know, maybe we should be conscious about it. But yeah, I definitely found myself, okay, really pushing Copilot as much as I can to, you know, keep the costs into the kind of fixed fixed license part of side of the equation.

Tom Arbuthnot: Awesome. Well, Tristan, thanks for jumping on the show and sharing so much of your perspective. It's really interesting. Well, definitely as this journey evolves, we'll have to have you c- back. But yeah, thanks so much for being so engaged with the community as well. Some, some great conversations.

Tristan Gayford: You're very welcome. Enjoy it. Thank you.