Why AI initiatives stall—and how IT can fix it

Good afternoon, everyone. I'm so excited to be here with you at Wiz. I hope you all have been having a great day. My name is Melissa Matross and I'm SVP Head of Design and Research at Asana.

And I am joined by two amazing, colleagues from Amazon Web Services and Anthropic. I'll let you guys introduce yourselves.

Hi, everyone. My name is Michael Horn. I'm the head of AgenTeq AI and go to market for Amazon Quick at AWS.

Hi. I'm Dan Mason. I am a member of the Applied AI team at Anthropic. In practice, that really just means a technical partner to sales and go to market and also closely connected with product research and and the model teams.

We're here to talk about how to implement AI the right way. There's part about technology and part about people, and my hope for today is that you'll walk away with a couple of new ideas for how to implement AI within your organizations or walk away feeling more confident with what you're doing.

So, I'm gonna charge you guys with a couple things on the panel today. I would love for you to give at least two specific examples as we're talking through some of the questions so folks can get really clear on what you're seeing and and have some concrete examples to work with.

So, why don't we jump in? Sure. So, Dan, most AI projects have a clear technical blueprint and no people blueprint. Is that consistent with what you've seen?

I I I would actually say I think, you know, a lot of times what what we observe is people, you know, I mean, people having FOMO in in a in the best possible way. Right? We we believe this can work for us. We wanna find out how, you know, they're they're afraid of, you know, essentially not pushing themselves far enough, which is a really nice inversion from a lot of what you see, which is you wait for technology to mature and only then do you tiptoe in.

Right? We we very much are seeing the opposite. I I think it's...

You know, there may be a technical problem that that people wanna solve and they may have a blueprint for it. A lot of what, you know, we we work to try to help with is, you know, sort of opening minds and sort of thinking bigger about, you know, I guess, what the right kind of investment is.

Yeah.

Right? Because a lot of the stuff that, you know, people are used to doing is, you know, planning and and, you know, trying to predict sort of, you know, where we're gonna be in a couple of years and what kind of tech stack and what kind of products we're gonna need.

AI is the enemy of of that kind of planning. It it just is. And and so, you know, a lot of what we're trying to do is help people adapt, give give people the right materials to to think through some of these things, and also, you know, to to to figure out how to interact with systems like this. It's it's it's very foreign, you know, to Yeah. To most most planners.

Yeah. Yeah. Michael, from from an infrastructure and platform perspective, do you see similar patterns?

Similar patterns. Yeah. I think the major problem, like Dana mentioned, is that often companies wait for the perfect moment Yeah. To try something out or to test something.

And especially in the AI world, you can't wait Anymore. Things change on a day to day basis, week to week basis.

So what's really important is not necessarily having a blueprint from a tech perspective or even the right people to do something, but it's ensuring that you have the right motives behind what you're doing

And specifically worrying about the transparency between different teams. So, for example, if you're an IT admin, I'm a former sys admin, it has never been more important to speak to leaders of business for the things that you might be changing. Like, nobody wants to wake up and have their entire workflow changed when they log into their computer in the morning.

Yeah.

But speaking to those leaders of businesses and making sure that you're not only helping them with their workflows, but ensuring that if you're trying to change a workflow that everything is interconnected. So I think just overall in terms of the people aspect, it's less about having the right people to push something forward, but ensuring that you're talking to the right people in the thing that you're trying to change.

Yeah. So clear communication and then you mentioned the term interconnected. Let's talk about that for a second. Sure.

When an organization comes to you with a fragmented stack, you know, legacy systems, SaaS silos, data distributed.

Where do you tell them to start?

Yeah. So we unfortunately do see this all the time and it's very normal. But the thing that I want to emphasize is certainly starting with something very small. And it may seem obvious at first, but starting with something very small that you can achieve results in and be able to measure those results.

Maybe it is one just business unit, or maybe it is just one team, or even just one person. But making sure that the data that is there is not necessarily ready, because I think oftentimes, like we mentioned before, companies wait a really long time, unfortunately, and the data changes and things change. But I think what they really need to do is focus on a small team or a small problem and see if they can enhance that with AI. Right?

And so, going into a small team and changing certain aspects of it and working again with that team, the clear communication aspect is so important. So, I think overall just avoiding kind of the silo problem and trying to boil the ocean Yeah. From day one. It's good to have very lofty goals and to have sort of a north star maybe where you want your organization to go, but you have to start small and be able to find the right tools that you need especially for that business or your overall organization.

There's so many different tools now in terms of interconnectivity where finding the right tool may be different for one business unit than another one. And you might have a bigger tool for your entire organization. So starting small, I think, is really just the best place to go, and with a team that you can get proper results and critical feedback from.

That's great. That's great. I love starting with a single proof point. Yes. Builds enthusiasm and success in the organization as well. Right?

So Dan, models are only as useful as the context they're given.

How much does poor data quality or disconnected systems actually limit what AI can do for the enterprise?

Well, so I I think it it depends what we mean by by disconnected systems because think there's a couple different ways that can manifest. Right? So, on the one hand, like, think the the the shorthand that we try to use is the thing you're asking Claude or some other agent to do, could a human do this with the access you're providing? And if the answer is no, you're gonna get hallucinations, weirdness, you know, results you don't like.

And so a a lot of this is just, you know, we spend a lot of time, again, in this sort of, you know, the planning industrial complex. We spend a lot of time thinking about how to construct systems and, you know, ultimately sort of make them perfect. And again, everyone agrees that like, you know, the perfect is the enemy of the good. You can start with good.

But I think the the single biggest thing is, you know, you you need to find a way to get comfortable, and this is hard from a product development standpoint. Like, I'm a former product person. Yeah. You need to get comfortable with building something that just barely works and not necessarily calling it an MVP because an MVP implies, oh, well, this is, you know, not what we're actually eventually gonna ship.

Like, this can't possibly be the final state of it. Maybe it is. Right. And, you know, what you what you need to figure out is, like, once you can make it work at all, is it too expensive?

Is it too slow? Does it need more access to data than I didn't already give it? Right? You can always expand from there.

Right? But people spend a lot of time usually pre optimizing these things. Right? You know, the...

That that is the main thing that you cannot do with AI is, like, spend a lot of time thinking about exactly what you need so that the AI can be successful. You gotta try it. You gotta see how it works and then you gotta optimize it only afterwards.

Yeah. So once again, starting with a proof point, seeing what works and then going from there. Yep. And then you talked about building where it doesn't have to be perfect. Michael, I'm interested if there's a point where an integration approach is good enough to move forward or do you find that incomplete data can come back and bite you?

There's a couple aspects to that question, right? So, if you have bad data going in, you're gonna have good data going out, right? I think we all know that at this point.

There is a certain time where you can test some things, and like Danton mentioned, it's that if the data is good enough for someone to do a workflow today and it's part of their day, you can test it today. But you have to go into it with the aspect or even mind... The thought process of we're going to invest in this thing and it may not work and it may not work in the right way, but we are at least evaluating something new. And we may shift and we may evaluate something else. Right?

And so it's just a very large aspect of that in terms of being able to make sure that, yes, you can enhance your data, take a look at the data, and we certainly don't want a migration of that data. The the best case scenario for all of this is to have data sit where it's currently sitting and be able to utilize it Rather than migrating it to a new platform or shifting to something else. Right? So I think it shouldn't stop anyone from doing something today and being able to test something. But at the same time, in parallel, you should be able to look at the data and really criticize your organization of how it's getting data to begin with. Right? So if we're looking at, like, a CRM system, for example.

Are there fundamental things broken in terms of how people are inputting data into that CRM system? And with that, you're not going to get great results. Right? So I think it can happen on both sides, really, but it shouldn't stop anyone.

Yeah.

Yeah. I mean, just to build on that, mean, think I I think it's most helpful to look at the entire thing just from a pure expense context.

Right? If I build a perfect data ontology and migrate everything and make it so that anyone, human agent or otherwise, could get a perfect answer the first time, I will pay a cost for that migration and then I will pay a lot less every time I wanna use it. Right? But there's a...

There's the total other end of the spectrum where you do nothing. You just make the data barely accessible and then sometimes you have to do three turns on the agent and has to go back and check its work. You have to jump in and actually correct it on things because you know where the data is and nobody else does. That is a very different cost profile, but it might be better for your business.

You just...

You you need to actually think about it from no matter what, I'm incurring a cost. It's just where do I want to incur that cost?

Yeah. Yeah.

Shifting gears a bit, let's talk a little bit about change management in organizations. There's real anxiety today among workforces right now that AI could impact individuals' roles. How should organizations talk honestly about that without dismissing the concern or creating unnecessary panic? Michael?

I think it goes back to what I said earlier. Communication is the key.

Yeah.

You know, again, nobody wants to wake up and have their entire workflow changed, and all of a sudden there's a new platform in and they have to learn something new.

But it's more importantly working with those customers. Right?

And, like, essentially, if you're an IT admin, your customer is your end user at the end the Right?

And working with your end users to ensure that not only are they aware that something's going to be changing, but getting their insight into what you're going to be changing is so important. And that often is where organizations fail by not including the end users of that product. And so you could talk to a leader of a business, right, but they don't have the same workflow that maybe someone on their team does.

Right.

And so making sure that you have the right inputs from certain teams or certain leaders of business is so critically important for that thing to be successful. And we talk a lot about how AI implementation fails, I think that's probably one of the top things of why it fails is because you don't have the right inputs from the right team of the thing that you're changing. And all of a sudden, has to use a new platform that they're not comfortable with Yeah. And may take them longer to do their job than they did prior just because they didn't have the right training Yeah.

Or the right communication points.

So, I think ultimately that's the best way to do it is certainly over communicate as much as possible.

And again, make sure that you have the right inputs from certain teams on the things that you will be changing so that they are not only enhancing their workflow and maybe they can do better things or gain more productivity. Yeah. But ensure that they're comfortable with that change in the first place and it means something to them in their day to day.

Yeah. Yeah. Dan, have you seen other tactics that have worked about reducing anxiety in the workforce?

Yeah. Well, I mean, this this is maybe less about the workforce and more just about our our buyers, but I think there's some overlap. So you asked for anecdotes. Yeah. So, you know, I I worked a fair bit with with some of the the the systemically important banks. And when we were first talking about implementing, Claude, just the Claude, you know, ecosystem, you know, as as as they do, they send a giant security questionnaire and, you know, we had to answer a whole bunch of things with not yet.

And what what came back from that was both some empathy on our side about, you know, we understood that, you know, these guys actually were getting so much pressure internally to implement it. Right? Now, they...

Everybody wants Cloud. Everybody wants, you know, OpenAI. Yeah. Everybody wanted all these tools. And they were seen, these these internal guys, as the gatekeepers of, like, well, you're the ones telling us we can't have this.

But but at the same time, when we talked to them, because we understood that, we were both able to use the things they were asking for which were totally reasonable as our roadmap. We were able to implement the roadmap together, bring them along for the ride. And then ultimately, you know, get them to a point where they could offer this in a way that they thought was was safe. Yeah.

And so, I mean, a lot of this was we know that just with the pace of everything that's going on, we're putting people in very awkward positions. Yeah. There's no way around it. And we we...

Our product moves fast, the models move fast. We wanna make it as digestible as possible for everybody. It's a very hard thing to do well, and and we we really appreciate everyone's forgiveness.

Yeah.

Then, like a a layer down, right, you know, when when it comes to people using it. The main thing that we struggle with, right, is is that it is very hard to straddle the line between giving access that fundamentally puts people in a in a sort of a bad position to do the thing that they're they're trying to do. Right? You know, it's not enough not enough spend, not the right models, not the right tools.

Right? And so they're being told to use it. They're being told that AI is transformative. They try it in this box and it doesn't work.

And so then they're like, well, wait. Is there something wrong with me? Is there something wrong with this? It puts you in a very awkward position As an individual.

What we want, what we want to open up more of is people having, you know, essentially the the right environment to do interesting things, like things that they will look at and be like, wow, that was cool. Right? Because then you can imagine yourself orchestrating this thing as opposed to, well, it's just a matter of time until it replaces me. It's bad now.

Right. Where is it going next week? You you have to actually have the thing in your hands so that you can see what it does, then you can take some power over it. Right?

And so it's it's kind of a perverse effect where in trying to control the rollout, I think you can sometimes create these effects where people are actually Yeah.

You know, in a weird way scared and disdainful of it in ways that, you know, is not good for anybody.

Yeah.

And I think to add on to that too, so much of it also is embracing from an organizational perspective, again, to invest in your employees.

Yeah.

So having the right training available and having workshops or hackathons or whatever it may look like to those end users in your organization, but really showing them that you're still investing in them. But you're also investing in these products. And that here's the best way to use them and here's the best way it can help you. So, no one feels left behind, but they feel empowered to use something new and gain new skills out of it. And I think that part is also extremely important for organizations today to really implement in their workforce as much as they possibly can.

Yeah.

I've seen even from people on my team, they're afraid to admit they don't have exposure to AI and so just someone showing them how Yeah.

Goes a long way.

Yeah. Yeah. Yeah. It's really tough in tech companies especially if someone doesn't want to admit that or they don't know where to start. Right? So, having the right guides or videos, again, whatever it may look like, even if it's just office hours from the person implementing it Yeah. Is super important for those people to come to an open area without judgment and being able to ask questions and be able to figure out how to use the thing that they're being told they need to use every day.

Right. Right. Well, let's talk a little bit about how we embed AI into the operating model and organizations.

From an infrastructure and platform perspective, what does good look like at the point where AI tooling is truly embedded in an organization's operating model? Michael?

It's really good question. So, there's a couple layers to it But I think the the best place to start is, again, just tracking sort of how people are using it. Right? So not only is it just usage, but I'll give a really good example of this, is that in any interconnected system, so a platform that builds agents, right, or or an agentic platform, You can connect something like Outlook to it and you can get a ton of usage out of that.

But if you're not connecting the tools that people use on a day to day basis beyond that Yeah. You're not gonna get a lot of usage out of it because all it's doing is maybe writing emails for them. That's not really worth it for them to learn a brand new tool to go through their day. And so, again, ensuring, going back to my earlier point, and I'll keep overstating this, but making sure that you are implementing the right tools and the right interconnectivity with these new platforms so that people can use the same tools that they've been using every day But maybe using it in a different way is so important.

So being able to look at usage across different things like different connectors and how many people are using them and what the usage looks like. Or maybe you're implementing a connector. I've seen this before where it's a tool that the company uses every day. The IT admin implemented it.

It's available, but no one's using it.

Right.

And you have to ask yourself those questions. Right? And it's like, did we not do a good enough job communicating that this is available? How did we do that?

Did we email everybody? Did they not see that? Yeah. So there are a lot of different things that need to be looked at in terms of usage when you're embedding AI into a new environment or even an older environment.

And again, critical feedback. Right? So going back to the end users, making sure that Yeah. Just asking them a very simple question or survey or having office hours and seeing if people know the new things that are coming out and if they're comfortable with the things that are going on around them.

So, there's no, I think, really quantifiable stat at that point, but there are many that you can implement to track usage over time Yeah. To track how people are building things. And Dan brought up really good points. I mean, we really want people building.

That's the sort of democratization of AI that we have now. Right? Yeah. And so having users being able to create apps or create agents on their own.

One of my customers, they created 400 agents within like ten days, I think, of starting. Wow. So from a number perspective, that's really impressive. Right?

Yeah. And so we wanna see people creating things. And I think that's a really good thing to focus on to ensure that there's usage across the board in that company.

Yeah.

Yeah. I mean, just to to build on that. I I think the hardest part right now is individual governance.

Right? It...

It's one thing to say, this is the tool that everybody's allowed to use, you know, go nuts. Yeah. It's another to say, well, I don't mean go nuts. I mean, you know, you have to do this very specific narrow set of tasks and and they need to be done really well and they have to be efficient and, you know, we're gonna and we're gonna yell at you if you go over your spend limits.

And again, I understand a 100% how hard this is to budget. Right? Like, are not making this easy on anybody.

Right.

But the question of of how to do that kind of individual governance and then, you know, essentially how to then move up a level and and and sort of do more of this stuff in public is is the main suggestion, I guess, I would make. And so, like, you know, we're talking backstage just about some of how anthropic works. And the single biggest thing is just that we have very, very few silos and very, very few, you know, not not a lot of different products and platforms that are sort of really integrated with our core work stack. It makes it relatively easy to plug everything in and get the maximum amount of value out of AI because there's just not a lot of barriers.

There are some barriers, right, but there's not a lot. And so the idea of how to change the operating model of a business in part is, again, think about how a human with these same restrictions would be able to operate. Think about your existing org, right? There are probably limits on what your humans can do because of their own access restrictions.

Doesn't mean break all the walls and throw it all away, but it's like think about specifically what you could do better and what you would need to change about where those walls are right now.

Yeah. So coming in fresh, Dan, what design choices or deployment choices, make the difference between when the tool is used or forgotten?

Well, do...

I mean, the single biggest thing is just that, you know, if...

And and some of this is about the the organization telegraphing its own values. Yeah. Right? You know, if it is about this is a tool that we think can help all of you, we're gonna invest in you.

Right? You know, exactly what you said. I think that that goes a long way to people then experimenting. Yep.

Yeah.

You know, I I had a...

Just another anecdote and this is just a a a like a a distant cousin, was basically saying, you know, I I work at a company where, you know, I get a lot of value out of of AI, but they give me this thing which hides which model it is, and I just pick the middle one all the time. Right? Because I figure no one's gonna notice. If it's the cheapest one, someone's gonna yell at me.

Pick the most expensive one, someone's gonna yell at me. Yeah. Right? And he gets a ton of value out of this thing, but he was like...

I was like, is it Claude? Is it OpenAI? He's like, I'm not even sure.

I just picked the middle one.

Right? And and so that is the org saying, we we are giving this to you, but, you know, we're gonna keep our own motives kind of obscured. Right? We're not gonna tell you what we're...

How we're gonna evaluate you. And in his case, he was getting a ton of value. Right? So, like, he should have been a success story.

But the org didn't plug in sort of their own thinking kind of into the into the whole picture.

And I think one more thing to add to that too, and even going back to what I mentioned before is on a basis of success, I think really showcasing what people build internally is also extremely important. It's something that we are really good at at AWS.

Being able to show, you know, different agents that are being built or different workflows that are being built. I was talking to few people earlier about, you know, some workflows that I created with Asana and how I use it every day and sharing that with my teammates and colleagues and having an open place to share those things. Yeah. So that other people aren't necessarily reinventing the wheel or maybe they do something else that's super interesting that takes 5% of your time but a 100% of theirs. And that just changes your day a little bit but you can just take that from them. Right?

And so being able to showcase kind of the cool workflows or the things that you build or the different ways that you are critically thinking about implementing AI, not even for yourself or for your team, but as a broader company Yeah.

Is also extremely important to kind of get that message across that people are doing very interesting things with it.

Yeah.

And it can help spur more building and more experimenting around. And, you know, even before we were talking about this, but it allows for anyone to kind of be their own product manager in a way. Right? Anyone can build pretty much anything at this point.

And so being able to take that product that you have built and being able to share that with others is also really cool and can drive different things for the company itself. Right? Right. So it could be additional revenue streams at some point.

It can create its own product for the company itself at some point. So Yeah. It just overall allows for a lot more openness, I think, in in any organization. And it should be...

Really be rewarded internally for people showing the things that they built, how they built them, and how it's changing their day to day, so that other people get proper use out of it.

Yeah. And and just to double down on that, I mean, I think it is really unnatural for most organizations to work in public.

Yes.

And there's there's a bunch of reasons why that is and a lot of them are good reasons. Right? You know, and and they're historical reasons. But it is the single biggest reason why stuff like this won't catch on.

You'll have these really isolated cases of people doing something cool and maybe you have a demo day or maybe you have some, you know, showcase where people can talk about this stuff. Maybe you have Slack and you talk about it in Slack. Right? It it can still leak out.

But it's different than you being able to sort of see everything that's going on and say like, oh, I could use that. Like, I...

Let let me try that myself. Right? I mean, that's how things catch on. And there is a governance sort of version of this where things bubble up and then admins grab them and implement them top down.

Like, it But, can you know, the easier way is just work in public. It...

But again, culturally, it's so foreign for for most organizations to work that way.

Yeah. I've seen leaders that have actively experimented themselves and showcased it to their teams and had others showcase it as part of their own learning journeys.

Yeah.

And it really creates that experimentation culture that I think really helps here.

It absolutely does. Yeah. And hearing that from the top down or seeing that from the top down, I think only encourages people to try because it shows that, you know, there's some humbleness behind it.

And what everyone's doing and what they're creating. Again, it just allows for more experimentation, I think, which Yeah.

At some point, can either get something really of good quality that can be shared amongst others. And again, you might have different revenue streams or different things that come out of it that may be really interesting from a company perspective, let alone a team perspective.

So, what would you say is the most honest signal that you've seen that AI is truly embedded, not just in demos or experimentation by early adopters, but generally genuinely part of how work gets done?

Yeah. It's a really good question. So I've been in AI for probably about six years now and and worked on traditional workflows, especially in, like, document processing. Right? And so it has come, obviously, a long way in hearing people talk about it and how they're using it for their personal life Using something like Claude. But talking to a lot of my friends and colleagues, and it comes up in natural conversation in terms of like, oh, I was trying this at work. I've been using, you know, platform or some agent that I built or something like that.

That it has become so normal now to talk about that. Think it's probably the the point that I've have seen. Completely different than a couple years ago, right, where, you know, you have to explain or I have to explain to people or companies what machine learning was and what that looks like.

But now people are using it on their own. Have access to all these tools that they can use either in their personal life or work life. And I think that bridge between the two is really what helped compounded this. Right? Is that it's not something that you're just restricted to using at work, but you can use it in your own personal life and you're finding different ways to do that. And naturally, people wanna take that ability and pull it into their workplace.

I think that's the biggest shift that we've had probably in a few years, is not that you go to work and you're like, oh, I really wish I had this tool at home, but it's I've been using this at home and I really wanna use it at work.

Right? And so, I think for me, probably in the last year or so, it's really where that sort of compounded and I've seen that in in people just talking about it in their everyday lives.

I think that echoes what Dan said this morning in the keynote where he talked about people planning their trips and things like that at home and then they want it for work.

Yeah. What about you, Michael?

So I think that the best sign is is when people's jobs change. Yeah. And and, you know, not in the sense of I got reorged and, you know, now they're telling me to something different. It's like, you know, you're you're still in the same team. You're you're, you know, theoretically delivering the same product, but the things you're doing are are different. Right? Freed up enough of your time to invest in something else which isn't specifically part of your role, but is useful and helpful and something you want to do.

And it's...

I think it's better if it's organic. I think, you know, one one one overall result though that I think we do see is smaller teams. Yeah. Right?

You will you will take a team that used to have all this scaffolding and all this need for sort of coordination, and you admit, well, we don't necessarily need the coordination, but you guys all know what we're doing here, and you're you're good people, and we wanna do more. Let's let's split this up, and let's have, you know, two teams which can now go chase our our instead of just one. Signs like that are really good, but in general, just the idea of, if all you're getting is, well, I do the exact same thing but AI does half of it for me, I I don't think that's a real win. I think it it has to be some kind of transformation the way that you think about the work.

Yeah. Yeah. Yeah. So as we get to our last couple questions here, I'd love to ask if an IT leader in this room could only do one thing in the next ninety days to meaningfully advance their AI readiness. What would you tell them?

I I think the...

Especially if you're asking a lot of these same questions that we're talking about here, like, what are my people doing? Is it is it helpful? Right? You know, like, why are we making this investment?

Understanding that in, you know, a a sort of a controlled environment, I think is is the number one thing I would do. So an example could be, you know, take people and give them, you know, a a generous budget to try things. And then, and we have a product that does this called Smart Reports. There's other versions that would sort of allow you the same kind of insight.

Then look deeply at what they did and figure out, are these things that I want my people doing more of? How much, you know, would it have cost if you didn't have this tool? Right? You know, just try to actually do apples to apples on kind of a task basis.

You know, how much are people succeeding? How much are they failing? Why are they failing? Right? That gives you a leg to stand on when you then wanna go bigger because then you can say, well, okay. I learned this much and now, you know, I have a sense of this thing that that we really can get ROI from. Let's go find the next thing.

Right? You know, that kind of snowball, you know, either comes in a big bang, you just give it to everybody and see what happens, which is also great if people are willing to do that. But if you're not, you know, then it is sort of this more controlled environment to try to get a handle on it.

Yeah. What about you?

I think I'll go back to the same thing before, which is like IT admins, please go talk to your leaders of business, make friends with them, chat with them. My favorite question to ask, really any customer at this point is, you know, what keeps you up at night? So Yeah. Going to these leaders of business and focusing on, you know, what really bothers them about their workflows, their tools, how it could be better, and having that real transparent talk between each other will allow for the creation of good tooling and things that can actually be utilized.

And going back to, you know, are people actually using it? Is it being embedded in the company? It's the best way to start with the problem statement. Right?

And so rather than looking at something and saying, oh, we need AI in our company, which we know is coming down from a lot of boards and a lot of CEOs. Right? But we don't know what to do with it. Yeah.

Speak to the leaders of business. There are problems that need to be solved. There are workflows that can be fixed. There are things that can be enhanced.

But without having those deep conversations across IT and across the business units and even, like, finance, for example, having the right budgets for all of those things, you're setting yourself up for failure without having all of that transparency between all those different units. So that would be my biggest kind of statement to make and to, again, overstate that as, like, transparency, transparency, communication all across the board. Yeah. Make sure that everyone knows what's going on, and make sure that you're getting the right inputs.

Again Yeah.

Make sure that they're using the product that you built because so much time goes into it, so much investment, resources, and money, it should be used.

Yeah. Right? So, last question. One word or one sentence each or for each of you. What's the most underestimated factor in a successful AI adoption? I think you might have just said it.

I will say transparency.

Yes. Always. Yes.

But, you know, just to add a little bit more on top of that too, I think allowing users to be creative Yeah.

Is another really good way to do that. Yeah.

No. I I think it is that that's...

It's transparency, it's openness, it's, you know, it's it's willingness to rethink things in in a way that that can be human positive, that can be good for your business, good for your teams. Right? Like, you know, finding the upside in all of this, you know, is is hard when, you know, you you you sort of feel like you're struggling against the structure. Right?

And and as much as you can get, you know, just to a place where you can see those positives. Right? And, you know, really kinda embrace them. I think that's that's the main thing I would do.

Thank you both for your time today and for your insight for for all of our guests here.