Tokconomics or the spend that's associated with the token consumption that comes standard as part of AI usage is kind of the new talk of the town and it's changing the way that business leaders are thinking about the value and financial ROI of the AI adoption that many of them have just begun to implement in full earnest. So is this cloud computing 2.0 know which gave rise to the modern PHOPS practice as we know it or does the spend associated with AI usage actually require fundamentally different ways of thinking about IT costs?
Shane Cronin, the head of PinOps and ITAM services at SHI joins me to talk about those topics and explain a little more about the tokconomics foundation of which SHI is a founding member. Thanks as always for tuning in. Let's get to the conversation. I am joined now by Shane. Shane, thank you so much for taking some time. >> Thanks so much for having me, Victoria. >> Absolutely. Before we dive into some of the specifics around SHI's work, I want to ask you a kind of broad catchall question.
I know the term tokconomics is getting thrown around quite a bit. There have been more conversations, I think, lately at all stages of business leadership around the costs associated with AI usage, how to measure things of that nature. from your perspective and your vantage point, how do you see that evolution of kind of AI use and experimentation into now this conversation around token consumption and and spending? >> It's pretty interesting. So, I think 8 months ago, 10 months ago, it wasn't on anyone's radar for the most part.
I think largely the models that have been rolled out have obviously been super impressive and have highlighted a number of business cases to many organizations that have driven them to adopt partnerships with anthropic or open AI all of which is consumption based. We saw the emergence of something called token maxing which was basically the idea that we wanted all of our employees to leverage these tools and we are going to create leaderboards and we're going to see how many people are using it, how often they're using it and some organizations even went as far to say you don't even get a promotion or you're not eligible unless you hit certain thresholds.
Then came the big bills. Then came the headlines of you know we blew our budget for the next two years of AI in the last month. We didn't understand exactly that it would cost that much. And then we entered a period I would say that is somewhat interesting where there is a demand to get better visibility be able to forecast understand how the bill is constructed without necessarily slowing down the adoption that most organizations are trying to leverage.
So it's been an interesting one. It's almost like everything that happened with cloud adoption whatever like 10 years ago that happened over a period of 2 years seem to have happened in a quarter. Uh so here we are. Obviously the foundation got established to help solve for this but certainly it's been a very accelerated challenge for our clients. >> Sure. Let's talk a bit about that foundation, the tokconomics foundation. I know SHI one of the founding members of that group.
Talk to me a little both about the group itself and what you're aiming to solve for and address in the market and why you and the team felt that it was such a an important and necessary group to be formed at this stage. the the foundation itself sits under what's called the Linux foundation. So that's kind of like the overarching foundation. There's already two foundations that somewhat exist that are responsible for helping organizations manage their technology spend.
You have the ITAM forum which is sort of let's say focused more on the governance and it's been historically I would say more focused on the licensing elements of technology. Then you had the emergence of PHOPS which was helping organizations understand how do you manage your cloud costs because this is a sort of consumptionbased model. It's different to paying $200 for a subscription and doing whatever you want. Everything that you do in cloud a bill will arrive a month later.
How do you manage that? I think the big debate that happened was well is tokconomics is AI spend I should say is this an extension of PHOPS? Is it the same practitioners who will be responsible for managing the spend? Can you apply the same best practices that we had in Pinpops to tokens or does it require a whole new set of capabilities and scopes and frameworks? The hypothesis is that it required its own dedicated practice due to the complexity of managing the spend and the challenges that it proposed.
There also wasn't clarity would it be the same practitioner? Are we talking to the same audience? I think the biggest differentiator that came is that FINOS has been focused on cost and I think tokconomics is more thinking about the value of the spend. So many organizations aren't thinking like can I stop spending on AI? They're thinking like is the dollar I just spent on that actually helping me reduce costs? Is it driving automation? Are we increasing revenues?
Because we'll keep spending if we're getting that ROI. So I think that's that's kind of where the foundation has settled. I would say where it is at this point is it's still in its infancy. So right now I think we've published a draft definition of what tokconomics even is. So to think about the amount of work that we have to do, it's quite extensive if we're at the definition stage. And from there, it's really going to be a process of identifying the high priority activities that we can do to help build some sort of frameworks to guide clients in being able to manage this spend more more consistently.
There's a few things within there that that I want to ask you to break down a bit further. So to your point on the ways that leaders are approaching this cost concept, right? To your point on value and ROI and not just necessarily seeing it as a budget item that should be reduced at all costs, but as something that can really help the business in the long run. What are some of the the early maybe best practices or guidelines that you're starting to see work for leaders who are trying to determine okay at what point does this spend make sense?
I know you had mentioned token maxing seems to have come and gone to some degree that doesn't seem to be what everyone's sticking with but are you starting to sense at least a forward motion towards a commonality? >> No. Um so the the challenge right now is this is probably the same story as always but we always think about it in like three categories. You have people, technology and process to manage this, right? If we take even the people aspect there's not that many people who have many years of experience managing AI spend, right?
I mean this is fairly new. If you take the process, there is no industry standard. So it's not like there is a playbook that we can all sign up, do the training, get certified and go implement it. And then third is the technology. There really isn't right now a technology provider who you can quickly plug in and it will just give you visibility into what everyone's using, tell you how often they're using it, what is the business case that they're they've built, is it generating the ROI.
So we have all these gaps. There are companies and there are clients that we work with where we've built some I would say customized processes for them. We've leveraged things like gateways to give visibility. We're working with some of the tooling providers to bring some of their next phase of product to our clients so that they can test and iterate and figure things out. But right now it very much is this sort of iterative process where we're all learning every day.
And a lot of that will probably persist because what I think people get wrong is there is the sense that there is a token and there is a cost of a token and everything I do should equate to a cost, right? But beneath the token, there are all these data points that fluctuate that actually determine the cost of that token. And right now, we actually don't really have visibility or reporting anywhere onto what those costs are that contribute to the token.
So, while a vendor may provide a customer, here's all the tokens you've consumed, they go one level down and say, well, okay, why did that token cost that or why did that token only get us this output? We can't really define. So, we're very much in that sort of figuring it out together phase, which is why we're excited to be part of the foundation because we really want to be at the sort of like cutting edge of of of how to do this. >> I can hear business leaders or at least some business leaders right in the in the back of my head as you say that go, "Okay, great.
So, we don't really have answers, right? There's no northstar to follow. I I'm much more comfortable when somebody tells me that there's, you know, a problem and a solution and I know what to do and we can go from there." So, are you starting to see business leaders really pull back on AI adoption in any way because of costs or do you think that's also maybe a bit over inflated in terms of this kind of doom cycle of of AI? >> Zero uh it seems zero pullback.
So, I should clarify what we don't see is a consistent framework that can be adopted to solve every customer's challenge. um it varies by their use cases, what the definition of value they've associated with that investment. So then that also ties into like how are you going to track that value. So because that's so different in every organization, I think we're we're very much bringing our best practices to the forefront, understanding how to effectively leverage AI to get that concern, build optionality into your program, and then also build some visibility.
So I just want to clarify there is no there it is nothing exists. It's very much it's just very much not a consistently adopted industry best practice that exists to your question about pulling back. No, I think the the biggest thing that we see is just a want to eliminate the uncertainty. People want to know can we predict what the spend is going to be? Can we forecast it? Can we understand who is leveraging it? What those business units are doing with it?
Can we even go as far as to show back those business units? How much they are spending? What can we say to marketing? We've empowered you to use this technology. Here is how much you are costing the business and drive that empowerment but also supplement it with some accountability. That's kind of what we're seeing the demand to be. I think the pressure from boards and CEOs to be AI first company just too strong right now that nobody is sitting there thinking like I'm going to pull back on this.
I think people see that as being a competitive disadvantage if they did that. So the demand is is definitely more on building better governance rather than pulling back. >> I know too earlier in our in our conversation just now you kind of referenced cloud which I think is the comparison point that many have kind of solidified around in terms of AI adoption and and what that cloud cycle looked like almost a decade ago at this point. We saw cloud to your point also give rise to this phinops specialty practice discipline within many organizations I think at virtually every stage of growth.
Do you see maybe over the next few years AI taking on a similar role not with PHOPS exactly but are we looking at some kind of AI cost ops practice down the line? >> Oh it's a super great question. I think it's the major thing that we don't know. The interesting thing about AI right now is that what we are seeing to an extent is some organizations trying to set up these sort of AI centers of excellence to manage and govern it. What we're seeing though is a disconnect with how they operate with other business units.
So if you look at AI, it is integrated into SAS products. If you look at Microsoft right now, yeah, most people have co-pilot, but as they go to release co-work that is token based and therefore that very much sits in the tokconomics category. But Microsoft today might be partially managed by your PHOPS team with Azure. Then you might have the rest of it managed by your ITAM org. So like these three things operating independently is not going to work.
It's it's going to cause too much friction. There's going to be a lack of speed to which you build best practices because who's the who's the autonomous like the the authority, right, that can say no, this is what we need to do. Now that is what we've seen start to happen. But what we've also started to seem to happen is that for a long time the issue with thin has been and ITAM has been there's a lot of work done to provide visibility to provide opportunity for optimization to provide better usage of cloud but you have to get all these other people to agree to take action.
So if I'm a Finance professional and I create a pretty dashboard that tells us all the ways we could reduce our cost and I go to the database team, they're unwilling to consolidate, I just build a pretty dashboard like I haven't delivered any value. But what's interesting is that because of the dynamics of how you know the inputs that go into AI as we think about you know data readiness retrieval qualities prompt disciplines all of these things everybody cares about this because if we go back to what I was saying about driving accountability to other business units marketing should care about leveraging AI more effectively because if they're being held accountable to this bill then they have a problem right?
If they're using it very inefficiently, then that's an issue. With cloud, you didn't have that. Like, did marketing really care if Adobe was like sitting on prem or sitting in the cloud? Couldn't care less. Now, they're very much going to care about good quality adoption of cloud and that will extend to legal, finance, it fops. So, I think what the opportunity more is that AI will become a driving force because tokens will become more relevant across all the rest of the technology stack.
So how do you not think about having a new team just for AI but you actually build a technology cost governance organization that's responsible for everything and I think if organizations go that direction those will be the ones that that win ultimately those who set up AI specific I think will find it a very difficult path because they won't be leveraging all the capabilities that already exist. So yes long way of answering your questions but hopefully hopefully somewhat contextual uh why I did that.
Yeah. No, absolutely. I think too that also spurs a question that I'm hearing from business leaders, from partners like SHI, also just from peers generally in in the workforce who go, "Right, my boss 6 months ago said, "Use AI at all cost. I want everybody using AI. We're going to be an AI first company." To your earlier point, and now they're seeing those bosses and those leaders go, "Okay, actually, maybe not using AI for for every task. Maybe don't do that." Right?
And I think it's creating a bit of maybe a sense of mental whiplash for some who are going okay so what is efficient AI use? Have you seen businesses are you working with leaders to start to identify then what those use cases look like in terms of how marketing to use your example should actually be be using the technology >> 100%. So a lot of what we will look at is what are the various use cases that actually exist and then we try to determine what are the options right?
So what are the different models that can be leveraged to to drive that outcome? You know, if you're talking about basic data analysis, you don't need to use a tokenbased model. You could use something like co-pilot that has a subscription. You can use it as much as you want, but then there is those more complex tasks that might require you to use Fable or something, right? And ultimately that's going to drive token usage. You know, right now we even see scenarios whereby if you, you know, are a Microsoft co-pilot user, you're able to select the model that co-pilot uses.
You can toggle between claw, open AAI, and traditional co-pilot. If we were to forecast what's going to happen, like most models will likely move to token based. So, companies really need to prepare for that effective use because if that switch happens in a relatively short period of time and your users are still writing recipes with Cold Pilot, then you've got a problem. We do work on that a lot and that's kind of the big pivot that we've seen with tokconomics so far is that instead of trying to focus on the technology, we're actually more focused on the accountability.
So who owns it? What is their guard rails for usage and then how do we build the right governance inclusive of enablement to ensure that they are educated enough to build the right usage of AI for themselves and their teams. So we're absolutely seeing it. It's probably one of the lowest hanging fruit activities that we can do today just to alleviate some of that pressure and build confidence that okay, we got a surprise bill, but we're more confident that that surprise bill was used for the good for the right reasons. >> Shane, as we start to wrap our our time together today, I do want to ask the crystal ball question that nobody has an answer to that in many ways inspired this whole conversation and the the tokconomics foundation, right?
But where do you see this continuing to evolve over time? I know you just said that tokconomics and token costs generally speaking are probably going to be more present in the future, not less, right? Businesses are starting to see these bills and taking action. How do you want to see the broader tech community, business leaders, etc. >> start to get a handle around what they want to see maybe by the end of this year to build that for next year? >> In an ideal scenario, I would go back to the previous discussion about how we build the teams who manage this.
I think it's very important right now that organizations look at technology spend holistically and have a clear idea of what is the technology stacks that they're invested in understanding what is the roadmap of that that technology stack and then bringing if they have those disciplines together. So this will take strong governance to really integrate the thinking of tokconomics in the planning stage. Tokconomics cannot be something that you do once you've deployed AI and you're like oh let's now make the usage effective like day one you need to start thinking about this but like I said it cannot be done in isolation you need engineering you need the AI team you need sourcing and procurement we need legal we need ITAM we need PHOPS everybody working together so whatever way organizations can achieve that sort of northstar of having a single group responsible for all technology spend that works with the various stakeholders I I think those will be the ones that win.
There will be a tremendous amount of data that already exists that can be leveraged in order to build those best practices. I think in an ideal world, this doesn't become a discipline that just stands alone because fundamentally, like I said, it's going to be integrated into almost every technology stack and that's that's the direction we need to go. So, I guess it's two things. One is this idea of shifting left, building these protocols and frameworks and processes early and being part of the planning stage so you are best equipped before you go live.
And then second is bringing all of those groups together to build a centralized view of all technology spend. Easier said than done, but I think most people could do it in the next 6 months. >> Well, I'm sure six months from now, we could have this conversation and many things will be true and some things will have changed because that seems to be the AI story in a nutshell. But for now, Shane, thank you again so much for taking some time and joining me today. >> Thank you so much, Victoria.
This transcript was generated automatically from the
video's captions and may contain errors.
