Name: Ahead of the Curve: How Top GTM Teams Are Building the Operational Foundation AI Demands
Uploaded: 2026-09-23T18:16:15.457Z
Duration: 1 h 52 s
Description: Ahead of the Curve: How Top GTM Teams Are Building the Operational Foundation AI Demands

Transcript for "Ahead of the Curve: How Top GTM Teams Are Building the Operational Foundation AI Demands": Hello. Hello, everyone. Let me see. Make sure I can start sharing my screen here. Alright, Desmond. Well, my slide here really appreciate everyone joining today. My name is Mike Mattson, and I currently lead the revenue operations team here at Indeedata. And we're really excited to start talking today about a, a report that we just recently published that's called the, AI kind of readiness report for 2026, state of AI go to market. In today's session, the structure is going to be we're going to be going through some of the findings, some of the data from this report that we recently published, going through some of the insights, the data itself. Also, we'll have two individuals here that I will quickly introduce here in a second. These two practitioners from two amazing companies also sharing their experiences as it relates to this concept of AI readiness and, you know, getting your your your AI company kind of ready for all the agents and infrastructure that we're going to be deploying here throughout the throughout the years. So I'm really excited to get into this today. And so let's jump in. Why did we do this? The thesis that we started with was we're hearing in the market from a lot of individuals in the ops community that everyone has been rolling out AI in their businesses. I think you all had a similar mandate that I actually had at the beginning of the year from your executives, which was go do AI. Cloud Code, I think came out around February, March. It hit the community by storm and Everyone started to build, everyone started to deploy, and everyone started just getting AI into the hands of their end users and just starting to deploy as fast as they can. That was great. But one of the things that we all learned as this is happening throughout the past six months is that none of our infrastructure really caught up to the deployment of our actual AI agents and these AI use cases that we deployed. Our data stayed the same. Our business processes stayed the same. The context stayed the same. And so that's the kind of gap that we're gonna be talking about today is, you know, how do we go from fastly deploying, getting agents out into the into the ecosystem to really get value into our business because there's a huge amount of value with these with these agents and these use cases, but at the same time, address the infrastructure gap that we're seeing today across all of our customers. And so we we had this thesis. We surveyed roughly around a 150 people in the community, and we asked them these kind of core questions that we'll go through some of the results here in a second. And they came back to us with some of these results here. And so the people that replied back to us, about half of them were in the rev ops kind of marketing ops practitioner space, and so we're kind of confident that a lot of the results here are actually tied to the individuals like yourselves that are listening to us today. They're actually living, breathing, and owning the GTM infrastructure, the data, the agents themselves, the tech stack. And the core finding that we'll we'll we'll kind of speak to is that everyone has deployed AI. You can see the stat here. Roughly around 80% people have deployed AI, but there's a huge gap in actually having your infrastructure ready to actually allow the AI agents to actually consume your data to run your processes and really provide ROI and outcomes to your business. And so that's what we're really excited to dive into today. Just a quick housekeeping, we have a chat here that all of you can start throwing questions in the chat. Please kinda put your comments into the chat because at the end, we'll kinda pull out some questions that we'll have time to respond to. If you do have a really interesting question, we have someone that's gonna be kind of highlighting some of them towards us, and so we can actually speak to some of them throughout the presentation here. But please make sure you use the chat because that will be the primary way that we capture which questions to address at the end. Alright. So let's jump into the heart of the material. And first, let's introduce the two amazing practitioners that I'm really honored to have join me today. We have Jessica Kao, who is the director of b to b go to market transformation at Adobe. In her current role, she actually is working with Adobe customers on their own go to market transformation. And so she has a lot of experience and background seen across different industries and different customers on how they're going through this AI journey. We also have Olga Traskova from Birdeye. She is the VP of revenue operations, and so she's living and breathing this every single day within her business and ultimately owning the go from go to market infrastructure and all the AI use cases across the go to market ops function. And so, Olga, let's start with you. Maybe let's kick off. Tell me a little bit about what your go to market org looks like at Birdeye. You know, how have you been kind of leaving this kind of journey around AI transformation and and how this kind of unfolded at BirdEye today? Thank you, Mike, and thank you for having me here. Excited to be here with you and and and Jess. I lead revenue operations at Virei. We support the full customer life cycle across marketing, sales, account management, customer success. My team covers traditional revenue operations foundation, starting from processes to systems to analytics, planning, execution, but we have also added the go to market engineering capability last year. What pushed us get there and get serious about AI and AI readiness is that we recognized there was a lot of experimentation happening everywhere. As you just mentioned, Cloud Code went live, in the beginning of the year, and suddenly, everyone started using that, but for their own personal productivity. Teams were using AI to research accounts, generate messaging, summarize, information that would eventually supposedly lead to increasing their personal productivity. When this became obvious that everyone was eager to code, eager to build, what we did, we ran a hackathon. In March of this year, we ran a go to market hackathon, where sales, marketing teams, finance, people, all of these teams, go to market teams unrelated to business applications or engineers. Right? They were building. They were coding, and, they were generating apps using all kinds of, in AI environments. Claude, Replit, you name it. That gave us solid ideas about where the pain is and what could be actually helpful at scale. Right? And then what we did is that we started centralized, on the go to market engineering team. That hackathon gave us sort of a road map of the apps and the automations and AI workflows, that we were gonna build. And we realized that, you know, this experimentation, should be sort of formalized, under one central, team. So we moved from asking where we can use AI to sort of centralizing on go to market engineering team and started to ask what has to be true for the AI to operate safely and effectively at scale. Yeah. And I'm really excited to to jump a little bit about the ownership that you you kinda talked about there a second, here, later in our discussion because I think that's gonna be a a very interesting conversation around who does own this strategy, how do you kinda build, you know, governance around it. So really really excited to dive into that. Thank thank you, Olga. Jessica, also to you, what's your role look like at Adobe, and what are customers coming to you right now as it relates to AI? Yeah. A lot. And so I you know, having been a, you know, previously an operations leader at many b to b companies, now I'm in Adobe working with a lot of different companies. I see on an average, like, this week, like, three different companies, and working with them through, you know, the people, processes, technology, change management. So what I I mean, it's a really great opportunity to basically see all of the trends across so many different comp companies and seeing the patterns of what is working, across, what are the similar challenges that different companies are facing. And a lot of it as as as we know, as in operations role, it's really the intersection of people and especially with AI, with people and technology. So we really work through some of the organization and op model changes. The things that I hear consistently is, like, everyone understands that we the way that we have to work is different, but how? You know, how do we have to work different? What do we have to pivot? And so I spend, you know, a great deal of time with many different companies, helping them through the transformation. And, you know, as operations leaders, I think we've all been and helped companies go through transformations, but now, like, every company on this planet is trying to go through transformation with AI. And there are patterns of, like, what are what are the what do what does good look like or what does successful what is common common thread between what are companies that are succeeding? What are kind of the common road bumps or speed bumps, that companies face as they transition, right, from a single you know, what I call that that single player. Like, we're all, you know, using it for productivity to transitioning it to, you know, across, like, an AI, you know, an AI infused workflow. So those are the some of the many things that I get to I get the privilege of working with many different companies, to see and witness because we're seeing, you know, very similar very similar patterns. Yeah. That's great. Yeah. I'm I'm super interested to dive into hearing more about in your customer base and in the industry, like, what are the different trends, what are the different types of models people are deploying? I think it's gonna be really interesting to to get into. Well, great. Thanks, Jess. And I'll go for the quick, intro here. So how about we jump to our first finding out of our, GTM readiness report here, which is kind of a lot of the, you know, back end of of how this kind of conversation is gonna be kind of structured around. So the the big idea here is that we've coined this term of this concept of our readiness gap. And what essentially that means is that the the headline here is that we're seeing across our community of folks that have participated in this in this report that deployment is outpacing the infrastructure readiness. And so what that means is that pretty much every single company out there has deployed an agent in their business. Right? Again, going back to the early in the year, we had this big push, this big mandate to get AI into their businesses, and everyone pretty much followed through on that. The flip side of that, though, is this gap that we're discussing, which is only about a third of, of the teams actually believe their infrastructure is ready to support these agents. And so we've gone, you know, moving fast, getting agents out the door, but as I kinda let in in at the beginning is the data, the infrastructure, the foundation, it it hasn't changed. It's still kind of in this legacy model that isn't really structured in a way that's supporting the agents today, and that is what we're calling the readiness gap. And so, Olga, let's kick it off to to you here. Do you how how does this resonate with you? Does that gap feel real to you in in in, in Birdeye? And, you know, how have you gone through kind of this transition of getting into a business where there's that you have really strong AI readiness infrastructure? Yeah. Absolutely. We started experimenting, with AI with AI with with skills, club based skills and artifacts, the outputs were valuable. Right? But we quickly learned that even great AI generated insights will not drive adoption, if they would sit outside of the sellers or of the marketers workflow. So it's one of the gaps that when every team member, even within the go to market engineering team or within our business applications or engineering team, If we create something that is not, well positioned for adoption, it's gonna die pretty pretty quickly. I believe it is very important to meet people where they are, where they already work. So instead of asking sellers and marketers and BDRs to navigate to this artifact, to go install this scale or call out this workflow, what we did is we built a go to market portal within our CRM, within Salesforce directly. So I I spoke about the go to market team, the the go to market engineering team that we have established last year. Having that team, those resources, that skill set allowed us to actually build, the playground for all kinds of AI apps and workflows directly within Salesforce. To what Jessica said, I 100% agree. It all starts with people. And as we navigate throughout this AI world, right, having the right skills, having the right expertise, having the right, people, that help you get there and understand the business pains and the business processes, the business challenges, as well as understanding the technical capabilities, is crucial. So having the solid team within revenue operations allowed us to build the go to market portal where we are now putting all kinds of, applications that are helping, that are helping our our teams, with their day to day. For example, account research, right, is just one of the one of the apps that we have there. It lives just within Salesforce account, just one of the tabs on the page layout that pulls, internal and external signals, summarizes account research, scores accounts. But what I have also realized is that this account research alone is not enough. A seller, our marketer, they don't just need more information. They need understanding of what to do next. So our next evolution of this app would be moving from insights to action. We'll start recommending next best actions. This is the best sort of type of outreach. This is who you need to reach out to. This is the next best step. That's what we're building. So another sort of insight that I wanted to share, what I learned, like, it's a continuous evolution of all of the apps and AI workflows that you're building. Like, you would always think about enhancing them if, something that you build lands well and is getting adopted, we always continue to to improve that and and build, build on top of that. So I think the lesson is that producing an AI output is is the easy part. The the real work is embedding it into your day to day workflow, connecting it to the right moment, recommending so what and what to do next, turning that insight into recommendation. Yeah. Couldn't agree more. And I love how you've kind of made the bet to build it into the Salesforce ecosystem. I know if you were to ask people that question maybe, four months ago, SaaS was SaaS was dead. So, it's really interesting that we're I think. Right? And now it depends on the swing back. Right? Now Dreamforce, I was like, you know, Agentforce is amazing, and, you know, it's it's just fun how the the trend changes over time. But, yeah, I think it's a really interesting bet that, you know, meeting you know, users or end users, like the sales folks, marketing folks in the platforms that they're in, like Salesforce, I think is a really interesting bet that, you know, curious to dive a little bit more in as we go through this. Jessica, what about you? What are you seeing across your customer base around this kind of, like, this gap here? Does that match what you're hearing around, you know, customers have, like, really fast deployed agents, and then they're looking at their infrastructure, and. they're like, absolutely. how do of, we, I think the whole world, this infrastructure of all seeing the same thing, especially? as ops people. And as as ops post, we're like, yep. Not that we told you so, but, like, you know, our job is even more important. And I think the kind of the hypothesis I have or the thread that I'm seeing is, like, look, AI is is a different it's a it's a different beast in terms of technology. And I think I come at it from a couple of different things. Right? Kind of what Olga was saying is, like, okay. You need to meet people where they're at. Right? If you think about it, we I've spent my whole we've all spent our whole career at the edge of, like, getting sales folks to use technology, getting marketing folks to use technology, and and there's certain ways that you can get them to use it in certain ways that they're not gonna use it. Right? We have a whole many decades of that experience. We know how to get people to use it. AI is adding another layer of of that. And so where I've seen where I've seen success or the lack of success, and I think this is a this is, you know, what ops what what we're, like, so suited to do is, like, it's not just, you know, yes, we need to meet people where they're at in the interface that they're at. But I think AI really has an advantage when we understand how to point it like a spear very specifically to the places and the workflows where it is. Right? Like, one, not every problem is an AI problem. Right? Throwing AI on top of, like, just kind of like a generic layer on top of everything. Let's AI file. Let's do identify everything. I think it's being very specific. And I also think it is different in that where I've seen success is when folks who are, like like us, are actually taking the time to work with our marketing stakeholders and our sales stakeholders. It's not like, hey, what are your what are your, what are your challenges and how can we fix them? It's actually sitting down with them and watching them work. Like, literally, like like, show me the screen, show me how you're clicking through things, understanding where the handoffs are from teams to teams, and getting to that level of detail and asking the right question, which is a very, very specific skill that I think, you know, ops folks are very suited to do. And so that's why when when when I see when companies take the time to really dissect the workflows. Right? Training is not like an 8AM session and a 5PM session for around the world. It is really being a partner and being a strategic partner to our stakeholders and go to market and understanding and mapping out and asking the right questions particular workflow. And then almost being like a doctor to, like, prescribe where AI should be in building those things into their actual workflow. So I think it's that double click thoroughness of really understanding how people work makes it successful. And having the resilience that, hey, if it didn't work the first time, that doesn't mean we give up and saying, hey, that was the wrong thing. Yeah. It's really the it's really the humans. It's always the humans. Right? It's. always the humans that that is the is the variable. So I think that's one aspect of it. And I think the second thing, when we talk about that readiness gap of deployment, we're we're putting the old framework of how we think about how we think about technology. Right? If you think about what an integration is. Right? We've all done many integrations. We're like, oh, we connect the integration. It's done. It's checked. It's once in time. And then we. walk. away when we go on to the next thing. And with AI, it's different. You know, kind of what all those things, it's a continuous loop. Like, hey. Having AI spare on output is easy. Is it the right output? Have we is that that loop? Do we have that loop of constantly checking? Right? And then and then that, quote, unquote, integration is not it's a living, breathing thing. And so our you know, we have to have dedicated people, humans, to actually also continue because the amount of evolution of AI and the technology and how fast it changes, like, it's like you have to constantly maintain whatever it is that you're gonna build. Do we have the infrastructure, the people, the processes, the ways of working to support that? Right? And how do you pivot to that as an organization? And I think the third the third thing which really kind of unifies everything is we have models, we have different interfaces, you use your plead, cloud, chat, chippy tea. Like, pick your favorite thing. It's the thing that says tried and true for each one of all of our companies is context. And more. importantly, it's shared context. Who owns it? Who's allowed to update it? At what cadence are you allowed to update it? Who's gonna govern it? Context, long term memory, governance, traceability, all of those things are important. And and I think that is where, you know, when we talk about that readiness gap is addressing those. Because that's all net new. Like right? Like, Yeah. we don't have, you know, we don't have you know, companies aren't typical typically, like, oh, we have a we have a, you know, a context governance or we have a context graph and here's the owners and this is how we update. And when a new product launches, this is this. is how we update our ICP. Yep. Those things. Yeah. Those are all brand new problems, and then ultimately, you know, the the outcome being that if you just applied agent to that, you know, and you don't have the governance, you don't have the rules in place, it just compounds the problem even faster. Because if you give it to a human, they they'll flag, like, oh, this is wrong. Right? This ICP definition seems off. Right? If you give it to an agent, it doesn't really have an opinion. It just it just executes. Right? So, I think you touched on a really interesting topic that we'll actually segue to here, which is around ownership. How do we think about who owns the AI strategy and even expanding that into who owns the context layer, who owns the data layer, who owns all of these kind of new things that we're all trying to figure out and ultimately build? And this is a core question we asked into the community, and we came back and we found that there's two primary ways people are currently structuring, quote, unquote, ownership or governance around AI. The biggest bucket you can see here is that it's a committee of folks. And so I think this is the common way we all started, which was we didn't really know exactly who owned what, how would this get, you know, managed, and so committees are easy ways to kind of just form around that. So I think that's the natural place that people gravitated towards. I think the one that's super interesting of this finding was that the next biggest kind of structure of, like, ownership model was no one owned it, that it was just kind of a bit of a free for all. And I think that's maybe, you know, anecdotally, like, background is, like, move fast. We gotta just keep going, and, like, let's not, like, let governance or rules can didn't get in the way of, like, innovation and and just quickly moving, in speed. And so it's just really interesting to see that these are the two structures that we're we're we're seeing in the results here. And, Jessica, since you've chatted with a lot of customers, I'm sure you've seen many different models around ownership. So maybe can you speak to what are you seeing as common trends around how your customer base is structuring their ownership around the AI strategy and all this agent management? Yeah. And, you know, we were actually you know, ran our own research and report and and things and it, of course, I mean, lines exactly to what yours are saying. We're seeing very similar similar things and more of, like, more specifically around when companies what does what does good companies are successful look like when they have a dedicated single AI owner. Right? Doesn't mean that, you know, COEs don't exist or steering committees don't exist, but that needs to necessary but not sufficient is that single owner. When we see the maturity where the where the people are the companies are ahead are the ones that have an AI owner. Because think about it. There's no captain of the ship. I mean, we're. just gonna be drifting. Right? And, like, especially, you know, we live all of us live, eat, and breathe, like, governance and cleanliness and data and goodness and data infrastructure. Right? We understand how important that is. You know, how you have to balance that and moving fast, but that's I mean, it aligns. Right? Like, having a single owner is directly related and correlated to, you know, how quickly or how how quickly they're able to achieve the the business outcomes and the success with. AI and go to. market. I couldn't agree more. I think most I mean, if you take a step back and look about the typical ownership models around, like, stack or data, like, there's dedicated teams and there's dedicated kind of owners that, like, own that. But within AI, it's like everyone has a seat at the table because I'd I think there's a lot of this conversation, like, where does it live? Like, does it live here? Does it live over there? Like, how do you kinda structure that? So, I couldn't agree more. I think as I've seen in my own kinda practice and kinda my relationships out there that shifting towards more of a single base owner seems to just have way more just, like, ability to move and just have kind of an ability to actually build quickly, instead of having kind of, like, committees that are essentially kind of, in my opinion, in some cases, like, slowing down. Like, what's the path forward? And, Olga, I know you at at Verdi went through a bit of a a transformation where you kinda started early around kinda ownership where you landed. Maybe talk a little bit about how you've kinda gone through that transition at Verdi. Yeah. And that's a that's a very compounded question. So let me just start with, initially, ownership for us was distributed. Right? As I said, different teams were experimenting based on their own needs, which was, I think, healthy. It drove a lot of energy and, help us generate a lot of use cases. But this distributed experimentation creates a lot of duplicated tools, duplicated AI workflows, credits consumption, then inconsistent data practices. And it's it's again about just optimizing your own efficiency versus creating it at scale. So we formalized the go to market engineering under revenue operations, because I truly believe that the ones under revenue operations, we are able to connect business priorities, data, technology, frontline execution, drive adoption, revenue operations is very well positioned for that. We can see the entire customer life cycle. We understand the dependencies between the teams. We understand the customer journey. Having said that, I also think, we should consider the build versus buy. If you can build, doesn't mean you should. Now the, the the the the threshold of entry into AI build has been is is extremely low. Everyone can build. But think about the ownership there. Right? So I I agree it should be this one owner. When you build AI product, what I learned, you become a product manager. Yep. Now you're responsible for administrating, maintaining, optimizing, working on the feedback, any glitches. So now you're all in one. So if you built an app, and I built many apps, now I have sales leaders, marketing leaders coming to me with requests and with with with with with things that are not working and need to be fixed. Right? So do you really have the capacity to own it within your team? So if you're making the decision to build an AI application, AI workflow versus buying something that is right away available, are you ready to maintain it? Who's gonna own it in three years from now? Who's gonna own it in eighteen months from now? Right? Because if if that's a single owner and that that person changes a role or leaves the company, is there capacity, within the company, within the team to take this on and continue, maintaining that? So that's just one of the key considerations, what I'm thinking about, the ownership. And I agree that every use case that we build, it needs a very clear named business owner, the technical owner, the measurable outcome, and someone who's gonna be responsible for adoption, and then someone who continues to own it after it goes live, right, with all of the glitches. and feedback and improvements management. Yeah. I love how you touched on the build versus buy because I think first half of the year, everyone was in build mode, right? Cloud cloud code. Everyone's like, we can build this. No one needs a CRM anymore. Let's just build one. And I think the pendulum again is starting to to shift back to your point where I built a lot of apps too and a lot of different types of agents, and it just becomes a huge maintenance lift that you have to continue to maintain. And then it kind of, like, expands even further as you, like, you kinda own the agent. You then have to own all the data that's feeding the agent. You gotta own the context that, you know, Jessica was mentioning that's kind of, like, managing this agent. So it's like your role just just compounds significantly larger, and if you don't have the resource and capacity to do that, it isn't a scalable solution. That brings us to our next topic that I wanted to jump into here, which is our, number one thing that we're hearing right now in in the business related to, what's essentially blocking the ability to really get strong ROI or start scaling your AI use cases into the business, and it's it's data. As you all know and as everyone's kind of mentioned this many times is your agents are only as good as the data, and and now more recently, you hear the term context. They're only as good as the data in the context that you're able to feed it. And so if you have a poor foundation, if we have poor data, you're just passing over bad outcomes to an agent that can essentially just replicate that significantly faster than a human. The the example I always like to use and I kinda mentioned a second ago is if you pass a bad ICP or you pass a bad, you know, data point to a human, they're gonna flag it and say, this is wrong. I'm an AE. I know better. I'm not gonna action on this bad data point. If you pass it to an agent, they're going to just execute. I think that's a huge challenge that a lot of people are experiencing in their businesses is the data quality. The next biggest one that we heard from the survey is integrations of the stack. How do we go from this legacy model of having these traditional point solutions or in some cases maybe more consolidated solution and then start adding AI into those? As Jess mentioned, it can't just be throw AI on top of everything. There has to be a more integrated type of strategy here. And so, Jessica, maybe let's kick it to you since you've, again, had probably a lot of different, you know, customers and points of view of how they're thinking about data and how it's gonna be, you know, executed against their ages. Are you are you hearing a lot about clean data? Is it more about context now? Is it a blend of both? Talk us a little bit about what are you hearing in your customer base related data. I think it's absolutely a a a bit of, both both. Like, the the cleanliness of data and then and then also the context. So we'll start out with, like if you think about in the pre AI world. Right? We had data. It was I mean, it's it's never it's never gonna be squeaky clean, but we operate in the world of, like, imperfect data. And if you think about it, like, where where now we're talking about humans, taste, judgment, wisdom, all of those, like, very squishy things. If you think about in the pre AI world, it was, you know, like, my team would I always I always wanna when I when I talk about operations, I always talk about, like, data and and one of the best practices is, like, I I like to go on a date with my data. I know my data really well. I'm best friends with my database. Right? Because you know how to interpret. All of us knows, like, hey. If we wanna cut a segment of customers, all of us know exactly we immediately go, we translate in our head. We know exactly how to do that. That's the wisdom and the intent. Right? So in pre AI, we had us, my team, your team, who would basically translate intent of the actions of sales and marketing and go to market, would go through our wisdom and our experience, and that's how we operate. I mean, that's how I think about it, how we operate to get the things to happen from a go to marketplace. Now when we have AI, we're, like, going directly into the data and there's not that human intermediary. What is the thing that is the gap? Right? Are we gonna get our data perfect? No. Can it be better? Absolutely. But that doesn't mean that we have to what I see a lot of times is people swing the other way. Like, oh, well, then we had to wait until our data is clean. That data. is never gonna be seen. How do we operate in a way? We've always done it. Right? In a way. of, like, we've operated in imperfect data and we it's just something that we live with. But with AI, it's like, how do we have AI? How do we have AI enhance what we do in our workflows in a world of imperfect data? And I think there's two aspects of it. Right? Which is one, again, it comes back to that context. How do we take the human, you know, the two decades worth of experience of myself, my team, operations, and give it to the AI in a way that it can operate in a world of imperfect data? So, right, that shared context, business context, all of those things. That's why I'm thinking that's why I think, like, the gap and the and the solution to the gap is that glue. The second thing the second thing is we can utilize AI to do things and speed up the way that we improve our data quality in a way that, you know, we we just never had the ability to do it from. a resource perspective. So I think it's it's two angles. And I think the thing that the thing I see a lot of companies get stuck with is the process of, like, oh, I have to wait until my data is perfect before I do something. And to me, the companies that I see that are successful are the ones where, hey, you don't know what to fix until you get in there. Right? So it's. like the chicken and egg. Like, if you get in there and you understand, like, hey, we gotta operate. We've always operated. We've always operated with the data that where it is. It's not like business has stopped. Right? But it's really around if we get in there, oh, and now I know where to fix. Right? It's. kind of like if you're learning how to cook and you're learning it from a YouTube video, you're never gonna really learn until you get your hands in it. Right? And that's when that that that you gotta start, but have that. in mind that you have to operate in the world of imperfect. We've always done it. Well, all of us in operations have always done it. We've always operated in imperfect data. How do we take our brains and give the and continually feed it the context? And that's where I think humans, that's the role of humans. It's contact you know, we talk about context engineering, but, like, every role, marketer, sales, like, that's that is what a continuous it's not one and done. That is going to be our role, is to continually give it the wisdom and context. Yeah. I couldn't agree more. I think how I think about context more recently also is, like, the lived experiences that that we're having that you're kind of alluding to. Like, I know when someone asks this kind of question, like, what are the I I know the fields I need to go look at in Salesforce. Right? And I I know what kind of business context of why they care about this. And how do you then build that context into the agent is gonna be the huge unlock that I think we're all shifting towards and trying to figure out how to best build this out. And it's gonna be really interesting to hear in the next, I would say, probably six months on how this kinda translates over time because that's gonna be the the huge unlock is is context. Olga, how how about we comment on this? What are you seeing in in your business? Maybe can you speak a little bit, at Birdeye, your kinda journey around how you've gone through this transformation and and where you started for rolling on agents in the early days and and data and and then trying to optimize data cleanliness and maybe what what didn't work, what did work. Talk to us a little bit about your journey there at at BirdEye. I agree with everything you guys are saying around data quality and the context and structured and structured data. Right? It's just like, as you're peeling the onion or building the foundational layers, it's just it's the it's the very first bottom foundational layer of having the data in there. I would talk about the data availability. Right? So Jess talked about integrations and all of the tech stack and how historically we've been working with a bunch of tools and making sure the data flows. What I realized in the beginning of the year as I'm building, for example, pipeline inspection agent or forecasting accuracy agent or highlighting risks on deals. If my calls are not recorded, then there's no point in building any AI solution. Right? So we're still going back to one zero one of revenue operations of, having all of the proper tools, integrations, and data flows configured, and and flowing. Right? Because what I what what we realized, in the beginning of the year that 30% of our calls were not being recorded for one reason or another. It was a user error, system issue, integration issue, and, our pipeline inspection agent just did not yield, any results. So to me, there's data quality, and there's also data availability and building that layer and ensuring that 100% of your data points that you're using for any AI apps that you're looking to build. Right? Having data on the point of decision is is is crucial and having trustworthy data. Right? So we spend time, fixing that layer. We spend time fixing all of the integrations, ensuring that, you know, SalesLoft to Salesforce is connected of course, to Salesforce is properly connected, working with those vendors to, to to bridge the gap so that then we can, build in, build build build some AI, on on top of that. Yeah. And, Olga, appreciate that that background. You just made me think of of a question or a comment that maybe each of you, Jess or Olga, can speak to that I found is, have you seen times where even if we have good data, so to speak, there's only so much data or context we can give an agent before it gets too overwhelmed? I've seen this in some cases of, like, if you just throw everything at an agent, it can't, you know, understand and consume all of it. And in some cases, it is actually worse than being more narrow in what you're passing to it. Curious if you guys personally have seen that or or, Jess, you've seen this in the customer base of yours. Like, how do you think about, like, sure. We get good data. We know some of it's bad. Like, what's the right stuff to pass to an agent? How do you make sure we're not essentially just throwing everything at it that ultimately is not valuable? That's what I've been for the past year, I've been, you know, on my top circuit and working with customers around context is the new data. Context is the new data. Like, how we have been focused on, you know, in the past around data cleanliness and activation of the data, enriching the data, that data strategy. Right? Now context is a strategy to exactly your point. Right? Like, I mean, just even in your in your personal use of, you know, your personal professional use of AI individually for productivity, you know, for for for for you know, because we're we're ops people. Right? Like, have you ever tried, like, running the same prompt on different models? I'm sure you have. Or different effort levels. Right? Like, sometimes more effort on more effort is not better. Actually, That's worse. worse. Yeah. It's actually worse. Sometimes I over I put like, I'm like, oops. I over I gave it too much context, and it, like, went off of left field. I'm sure we've all experienced that just on an individual level. So think about it at a at scale. So it's a combination of, you know, your ask, the model, the reasoning, the context. And so that's why, like, when everyone's focused on, hey, let's build AI agents. Let's build this. Let's build that. What is the context strategy around? Right? So now we have data strat data cleanliness, data hygiene, you know, data quality, but then there's also strategically giving it what context it needs. Because that can, you know, that is can be the difference between getting the out the outcome the output, the outcome that you are looking for and, like, veering off into left field. Right? And and I think. that's where when we're we'll talk about coming back to building. What's the infrastructure? What's the how are the ways of working? The ownership. It's it's these new human roles that have to really think about, well, when your UI has come to natural language, right, how do we translate probabilistic to deterministic actions and go to market for for everything that we do? Right? Like, if I'm saying, like, hey. I want people I want people to be able to query their data and and and and decrease the insights to actions faster. If if I want to report out on a campaign performance, you know, human a a person can ask it 20 different ways, but I want that to trigger and invoke a very deterministic thing. How do I. the guardrails. What are the guardrails that I have to put in to make sure I think that happened? That's that's not at someone's part time. That's not someone's side gig. Those are people in what I think people in teams, like, full time job. The adaptive layer of making all of that happen. That's. gonna I think that's net new and and net new focus area, and that's with more and more AI. That's that's a that's a different type of maintenance. Yeah. And you're you're speaking my language on the deterministic and probabilistic scenarios here. Because, yeah, I I personally feel and, you know, we as a business feel too, there's definitely a hybrid model here where you're not gonna wanna run probabilistic across everything. Right? Like, Deterministic rule base is still 100% significantly better for a lot of tasks compared to just throwing an LLM at a problem or at a process step. All of that goes back to what you're saying, which is how do you establish the business context of the rules of engagement even around, like, what agents do versus what's this, you know, deterministic workflow? How do we build business context? I talk about process context a lot. Like, you know, you have a lot of process workflow across your business from lead to cash, essentially. There's all these process steps. So how do you know what the process context is to essentially be able to pass over to an agent where, in most cases, those processes are deterministic? How do you then blend that into those probabilistic models is, I think, a very interesting place that we're shifting towards as a business and me personally are really interested in exploring is having process contexts that you can feed into these models. Olga, is this anything that you're experiencing that's a bird eye around context and how you're thinking about context in your business? Yeah. And I would just say, you know, in a simple language, to me, it's just about nailing the prompt. You need to be very precise. Right? Any AI app is powered by the prompt on the back end. You're essentially asking questions. You're configuring those prompts. You need to be very clear within your prompts exactly how you want the data to be returned and what format and what do you want it to look at. So from that perspective, I'm I'm believing in human in the lead, not the human in the loop. So we're we're still configuring that, you know, very precisely. We're spending a lot of time. We're spending a lot of time testing back and forth. Right? So now we have more than 80% of fields in Salesforce are filled in on the opportunity object filled in via AI. Some of the. fields, as you can imagine, are pick list values or. multi select pick list values. So you need to be very precise when you're configuring those prompts. And there's, there's some back back and forth and testing, that is that is required. So to me, it all, you know, just boils down to, to to being very explicit in in your prompt, connecting to the right data points. So I think this is what you guys are calling deterministic. I agree with you. Yeah. Well, great. Well, I I could geek out on context all day long. But I know we want to jump to, our final finding here, our fourth finding, which is related to trust. I think this is predicated on the idea where if you have agents running and you have, let's say, mediocre or bad data and you have a lack of context, we mentioned, like, it's gonna push out outputs regardless. In those outputs, they're gonna be potentially not very good or end users then, whether they're salespeople consuming, like, a recommended next best action or email or something, don't think it's very good. Or as opt people, we're not really confident the agent's actually updating, although to your example, the fields correctly. It's actually overriding the wrong values. And so a lot of that is gonna stem into this trust, issue that we are seeing and what the community has kinda said back to us is that, you know, the vast majority of folks you know, 60% of folks who kinda responded here say that their top fear in AI is acting on the wrong record. So user. example, Olga, it's updating the wrong opportunity. It's inputting the wrong ARR amount. It's inputting the wrong, you know, next steps, whatever that, you know, scenario is. Like, that's the top fear that people have. And what people are asking for is is more visibility or more audit trails, if you will, around what the agent is doing. And that today, I think, is a a bit of a gap in in in the in the market and in these kind of models is there's not a really good way to track exactly why did an agent do these tasks. Like, why did it take this action? What was its rationale? What did it change? And how do I track that and ultimately validate and confirm and QA it essentially before I unleash it too too far and it, you know, ultimately goes down a path that I don't want it to. So, Olga, maybe over to you because you've talked a little bit about how you're you have agents kind of updating 80% of opportunities, which is amazing. Talk to me a bit about, like, how do you think about audibility, governance, and those kind of rules to make sure the agent isn't going rogue or going in the direction you don't want it to? Yeah. Agents are updating 100% of opportunities, 80% of fields within those opportunities. And that is, on purpose. yeah. We're reserving some fields for, humans to to update. So we wanna be able to distinguish, what AI is filling in versus what humans are filling in. I think it all comes down to what we just discussed, right, the underlying data, the availability, and the quality of data. So when I look into building an AI workflow, we always start with data. Do we have the right data? Do we trust the data? Then if we're building it, what's gonna be the impact? What's the use case? And how reversible the decision is and who's gonna be the owner? So, essentially, what we everything that we just, everything that we just discussed. Right? Are we able to maintain it on a long term bay basis? So what's gonna be the level of effort for us to maintain it? What's gonna be the impact? Should we build? Should we buy? Who is the owner from the business side? Who is the owner from, from from the technical side. Some of the polls can be 100% AI led. Right? Some of the low level, impact, flows. But when it comes to the pipeline inspection, when it comes to the forecasting accuracy, when it comes to something that then is being brought up to c level, board level, execution level, this is where I'm having a human in in in the lead, and we inspect what AI is, AI is producing. Yeah. Trust requires it sound. So, yeah, I was saying trust requires visibility. So I'm ensuring that we can get into the, back end of whatever AI is working on to just spot check that as well. Yeah. So it sounds like you have some type of a matrix that says maybe low risk agent can maybe do that task. It's higher risk to say pipeline or something that's maybe a human in the roof or human led. So you have some type type of matrix, I would assume, that's determining what are the rules. That exactly it. Low risk could be account research, summarizing information, drafting some emails, flagging some missing information. Right? Medium risk would be, like, prioritization of accounts, deal scoring. Then high risk is where we have humans involved and deterministic control is forecasting commits, territory changes, customer commitments, deal deal quality, Yeah. etcetera. I love that. And, just maybe speak quickly here because I know we're pacing a little behind, but, what are you seeing in your in your customers and in industries you're working in around auditing and visibility and this idea that people are really afraid that AI is going to go rogue and do something wrong? I mean, this is I mean, at Adobe, I mean, our philosophy is, like, with we're we're very big on, like, governance. We're very big on traceability. I mean, I think it should be a 100%. Like, you you wanna know. You wanna be who the audit I mean, if you think about, you know, in in our in in the in the stack that, you know, as an operations person, you typically manage. Right? You have users, you have roles, you have permissions, Mhmm. things are locked down, like, what you can see, what object can I see in my CRM? Right? That's just table stakes, right, around who can who can read write fields. I mean, that's just in the that's that's kind of right? Like, we we wanna. make sure we call that. zero one. Yeah. Exactly. One zero one. Why would that be any different when we have AI or any anything, any person changing data view? We wanna be able to have that all that trail and that traceability. And so I think that is governance. Governance, governance, governance should be, you know, it's not the sexy thing. Right? But it's so so so important. And I'll I'll give you an example of just, you know, when I was at a conference and chatting with, you know, peer peer group, like minded folks, and and people get very excited about AI. And they're like, great. We wanna, you know, we wanna put AI on top of all of our data. Everyone can be able to, ask it questions. And what was fascinating and interesting, was that, you know, these were, you know because I I used to own, like, analytics for QBRs for marketing. Right? My team used to own all the dashboards. These are fixed. These are locked down. You will only use these. In a world where anyone can query the data, guess what happened. right, at QBR. dashboards. And they and basically, all of a sudden, they spent the entire time showing data that conflicted and talking about, well, who has what is the right number instead of actually going through and making and and and and leading to insights. I don't even talk about action. Can we get any insights? No. Because the problem was everyone came with their own. Right? Because. they had their own they asked their own questions which came with their own dashboards. And so I think I think that goes more to all the things that are tried and true around governance, around, you know, having a strategy is so important in order to make what all of us, right, like the do AI and to make AI successful is that you have to have a strategy around all the things that we've talked about today. Yeah. I love it. And I think that's a great segue to our kind of final commentary around, you know, where do we think that businesses should be kinda making their bets to kind of, to your point, have this strategy. And we see it kinda falling into three pillars here. So our point of view is that in order to have really successful AI strategy, you're gonna have your you need to have all three of these types of things that we we've discussed throughout the the conversation today, which, by the way, has been great. So I really appreciate you guys in in this dialogue. The first one, as we talked about, is is data foundation, having some type of structured clean or as clean as possible type of data that you have governance models in place, you have visibility into, you have right, you know, tables, all the all the kind of unsexy stuff of having strong data foundation and and more importantly, having the context behind a lot of that data foundation is another big piece. Business process truth. So we talked a lot about, like, what is the deterministic workflows, what is the documentation, What are the rules based that Olga was talking about? Like, the agent can do these types of tasks against these types of fields because that's the rules we put in place. So having a really strong business process truth that are really deterministic in nature that you can control and really kinda manage. And the third one, which is I think the huge unlock here that we're gonna see in the in the short term is around this shared customer context. So marketing, sales, CX, all working off of the same business context of their processes, of their data, of their agents. Whereas before, I think people have been kinda just going purely off of data. Now this idea that, Jessa, I really like how you've kinda framed it. This is, like, lived experience of how we pass over that information to an agent to really have, for lack of a better term, a brain to actually do the right thing, that is all the context that's gonna be required to pass over to those agents who really unlock these things. So we see this as as a core takeaway. These are the three pillars to have a really strong AI strategy going forward. So before we jump into q and a here, I'll just kinda pass it to to Jess, you, and and Olga here and have any kind of final words. Like, we talked a lot about different topics, a lot of different concepts. Where do you think people should start, just to kind of get this ball rolling here? Maybe, I, mean, I think we're us. Yeah. Go ahead. think we're right. We're all being asked to show success, and so what what does that look like? It's like, what do you when there's infinite things that you could do, where where do you start? Right? And that's you know, I'm a big believer in in in in choosing something, showing success, and building the momentum, right, not try to boil the ocean. So basically triangulating what do you control and own, what are the skill sets that you have, what is the technology that you have available to you, what what is the data that you have available to you. Right? There's lots of things that we could do, but a lot of times, like, hey. I can't do that because because, you know, to get access to that data would take a, you know, you know, a lot more additional alignment and work with the different teams. Right? So what do I have control over in my domain? What can I do? What can I show? And what directly ties to an ROI or an impact? What drives revenue? And everything that we do in operations should be that we're doing ties directly to revenue. And so really about translating that and then showing success. Right? And, like, again, starting with the strategic, like, looking at the workflow, dissecting, pointing AI very precisely to where it can augment, and then executing that and showing success and momentum, and then building that found and then moving forward. Love it. Olga, any final words here for the group? Yeah. Exactly to what you guys said. You summarized it very neatly. Don't start with the tool. Start with the business problem, with the outcome that that matters. Map the full workflow. Check your data quality, data availability. Do you believe the data? Do you have the data and the processes that are strong enough to support, your your workflow that you're trying to build? So start now. Build the foundation, that your workflow is gonna, rely on, and then earn the right to automate more. Amazing. Well, great. Well, appreciate again, Jessica and Olga for having this great conversation with us, and with me. This has been amazing. I really appreciate the dialogue here. So we'll open up to q and a for folks. If you have any questions, please throw it in the chat. We'll give about a minute or two here just to get that going before we close it out. Let me see questions I have here that people are discussing in the Q and A. Here. We have one here. I'll just paraphrase it real quick. We have a question that says, we've mentioned that sometimes you need to sit with end users to walk them through AI tooling. I think this is Jessica, maybe something you were talking about, like sitting down, understanding their workflow. How do you approach deploying AI into very large teams where maybe, you can't maybe get visibility to every single person's workflow? So how do you kinda work through, like, understanding the the workflows of our team members at a kind of a large company, and really understand the needs that they have before we start deploying agents there. It starts with the use case or the scenario. Right? Like, instead of you're not trying to map, like, everyone what everyone does all at once with the with the with, you know, with their with their roles and their remits. It's picking a a particular business process. Or it's like for example, it's like, okay. We're we're gonna map out and look at, you know, the production of, you know, paid media campaigns and how it needs to go across all these different teams. Right? We start with, you can either call it a use case or a scenario, and then it only touches very specific people. And then you focus on the a representative or a core group of representatives from each of those processes. You interview them, and then and then you ask the right questions, and then you map that out. Right? Or it's, or it's something maybe in in sales and maybe we'll go get, you know, talk about something that she's mapped. But, like, you start with you start with that business process, right, Mhmm. or a or a scenario. We're not trying to do again, we're not trying to boil the whole ocean. We're not trying to say, hey. How are we gonna how are we gonna improve AI for a particular team? It's a it's AI has the greatest impact when you're looking at things across teams when it's the hand off. Yeah. Olga, anything to add to that around getting end users processes and helping deploy AI to to solve their problems? We have very much we have very prescribed sales process. Right? It's it's fully configured. Right? There's there's very little autonomy within our sales process. Right? So we don't really ask sellers how they like to do things. We tell them how they should be doing things. Right? So that solves it. And then back to meeting your sellers where they are. Just build a a tool, a workflow for them wherever they are, Salesforce, HubSpot, any other platform that you guys are using. It can be some guidance. It can be automation. It can be recommended as best actions. It can be, deal scoring, lead scoring, account scoring, whatever that is, you know, with with with some insights. That is mapped out to to their process, either following up on leads or creating opportunities or your metric or any other sales process or methodology that you guys are, that you guys are using to what Jessica said, map to the outcomes, to the revenue, like, what actually drives the the bottom line and then start from there. Amazing. Alright. So we're coming up to time here. And just to close this out, we wanna make sure everyone knows if you like this conversation we had today with with Jessica, and thank you, and Olga, thank you. We're having more of these in a couple weeks from now on October fifth and sixth. LeanData is hosting their ops star event that's at the, San Francisco, InterContinental downtown. It's our tenth year, so we're super excited to kinda kick this off. Jessica here is actually gonna be speaking at the event. She's gonna be talking about the future of ops, turning AI pilots into business outcomes. She'll talk about why pilots never make it into production and how ops people can change that. So really excited. If you wanna join us, please sign up. Here's a QR code, and we can get that kinda going for you. The last thing is just to thank you for everyone for joining. If you wanna learn more about the report, you have a QR code here. You can digest the whole thing, and you can kinda read it offline. And if you wanna reach out to all of us, feel free to reach out to us directly on LinkedIn. We're we're always available to anyone out there who wants to learn more. But, again, Jessica, Olga, really appreciate the time. Appreciate the conversation, and looking forward to having more of these discussions with you both. Cheers. Thank you, Mike.