Name: What It Takes to Operationalize AI in Salesforce
Uploaded: 2026-05-07T17:30:52.439Z
Duration: 1 h 44 s
Description: What It Takes to Operationalize AI in Salesforce

Transcript for "What It Takes to Operationalize AI in Salesforce": Hello, everyone. I'm Mariam, campaign assistant here at Salesforce Ben. Whilst we wait for others to join, please tell me where you are tuning in from around the world. It's always nice to know where you are tuning in from, so please let me know in the chat box where you are joining from, and and we'll just give a moment until a lot a few a few people join. If you're joining with us for the first time, welcome. And if you've been with us before, welcome back to another Salesforce webinar, this time in collaboration with Coastal. Today's session will be a deeper dive into what it takes to operationalize AI in Salesforce. During this session, we will cover where do AI initiatives actually break down, are AI initiatives falling short because technology isn't delivering or because success was never clearly defined to begin with, and what separates organizations that are scaling AI successfully from those waiting for it to click. You are probably thinking, how do they know the answer to this, and what is it based on? But luckily for you, Coastal's twenty twenty six AI operation reports look at how 800 business and IT leaders are building and running AI inside their organizations. And in this webinar, we'll be diving right into the data and what it means to up to for operating AI in Salesforce. But before we get started and we dive into the good stuff, just a couple of housekeeping bits. The session is being recorded and will be shared via email after the webinar as well as Coastal's twenty twenty six a operational operations report, so please look out for it. There will also be a short survey. So if you could please help us figure it out, we'll really appreciate it. You are also in listen only mode, and you'll be mute muted throughout the session. But please do interact using the chat box and the q I the q and a tab for any questions you might have. If you're new here, just a quick introduction to Salesforce Spend. We are on a journey towards a future where there is a clear path to success for anyone wanting to start or grow their careers in the Salesforce ecosystem. We do this by sharing practical, relevant, and accurate information across a range of our platforms. So that includes articles, newsletters, in person events, virtual sessions like this one, and YouTube and our podcast. So there is something for everyone. If you haven't explored these yet, feel free to check them out on socialspend.com. And now without further ado, I'd like to invite our lovely speakers to the stage. We have Eric Buried, CEO of Coastal. We have Sam Taylor, director of innovation, also from Coastal, and Scott Juroski, senior director, agent force and data cloud consumption from Salesforce. Hello, guys. Hi, Miriam. How are you all today, and how are you feeling? I feel great. How does. Sam. feel? How does Scott feel? be here. Hello? Excited. You guys wanna do a quick. introduction about yourselves and what you're most excited about? Why don't you start, Scott? Sure. Hi, everybody. Scott Jaworski, senior director of consumption sellers of at Salesforce. Been at Salesforce for almost fourteen years. And, just excited to talk about AI implementations and data and all that great stuff. It's our it's our bread and butter. It's a great topic and, an exciting topic to talk about. Go for Sam. Sam Taylor, senior director of innovation and emerging tech here at Coastal. I'm really excited to share some of the learnings that we've had over the last eighteen months or so of Agent Force, and what we see kind of going further in the market here. Really just also love to see the feedback that we get from our users in our in our chat and our poll later on. And I'm Eric Berridge. I'm the CEO of Coastal. Really excited to be here today. Excited to have Sam and Scott with us as we talk about the, the the topic of AI, which seems to be the topic we all talk about all day long every day. So hopefully we can get some clarity here over the next forty five minutes or so, and then open it up to the audience for some questions. So I won't drain the agenda here, but we're gonna talk about the AI implementation gap. As Miriam mentioned, Coastal has done an exhaustive study teaming up with Oxford Economics where we surveyed over 800 global organizations to really dive into what organizations are doing with AI today. It's obviously a headline that we see in our daily lives. AI is something that we're using in our personal lives. But I also think it's very safe to say that businesses are struggling to figure out how to harness these powerful technologies to drive true business value. And one of the things that we found in our report, is that there's an operations gap that organizations need to close to truly leverage the benefits of AI. So So that's what we're gonna talk about over the next forty five minutes. We'll try to keep it lively for you. I will start by talking a little bit about what we're calling the AI implementation gap. The bottom line is leaders are committed. The results are lagging. There's a gap between belief and investment and performance. 84% of our respondents to our Oxford Economics survey say that AI is making them more competitive. 74% say they're increasing their investment over the next twelve months. But almost half of our respondents has said it's not AI is not meeting their expectations, and only one in five. So 20% of our respondents strongly agree that AI has delivered measurable business value. So that is the implementation gap that we see out there. And the bottom line is the belief in AI is there. The money is obviously there. I mean, we see it driving economies today in terms of how AI is being built out from various vendors, including our friends here at Salesforce, but the results aren't yet catching up. The technology works. Coastal is a long time Salesforce partner. We were one of the first partners to adopt AgentForce. We've now deployed it across a multitude of clients. The technology works, but there's an operations gap. And that is holding organizations back from actually driving true business value using AI in their day to day b to b operations and b to c. We've learned how to launch AI, but we really haven't figured out how to operationalize the infrastructure to sustain it and really get that flywheel effect that we all know it has the capability of delivering to our businesses. There was an interesting report that came out last year from MIT that was pretty controversial, and I think it was a little over the top. MIT came out and said that 95% of AI initiatives show zero return. That isn't what we've seen. We've actually seen organizations drive return when they take the right approach and when they operationalize AI correctly in their organizations, which again is what we're gonna talk about. But it is a fact that organizations are struggle struggling with three things. There's they continue to struggle with their data, and we're gonna talk about that. They struggle with the adoption of AI, and they struggle to truly drive measurable business value, which is ultimately what leadership is looking for. If we're gonna spend all of this capital on a new model that drives future automation, we need to be able to define the business value. So those are while the MIT study, I think, was a little over the top, these are facts that are coming out of our survey, that really help to guide future decision making for organizations. Let's bring in our panelists, and I'll start with Scott. You've got an incredible background at Salesforce, over the better part of two decades. You've seen Salesforce grow up as an enterprise application that organizations use to drive customer experience and CRM. You've now been at the forefront of how Salesforce is deploying their own AI capabilities. I know you talk to a lot of customers. So I'll start with, you know, a a a question that I think our audience is gonna appreciate. Like, what are leaders telling you privately that they're not putting in status reports, emails, on their own webinars? Like, what are you actually hearing from business leaders as they're going down the path of deploying AI in their businesses? Yeah. Absolutely. Thank you for the question, Eric. Yeah. What we're seeing is really around the idea that there was this kinda hype cycle of AI and what it could do and and all these great things it's gonna do for our business. And, you know, even as a a salesperson, you know, we kinda over hype it as well. We wanna get excited about it. And now I think we're kinda coming back to reality and I think that's where the, you know, our leaders that I'm talking to, the customers, it's, you know, they are cautiously optimistic. Right? It's they see the value in it, they understand what it can do, now it's get out of the hype cycle and come back down to reality of what it can actually do. And listening to what you're talking about adoption and KPIs and value, I think it's one of these things where we approach it as, hey. You have to understand what it can do from a, you know, you know, what do I really want it to do? Right? What am I really trying to fix? What are the problems I'm trying to fix? I think we were just throwing AI out there for AI sake, and this is where the leaders are coming back to us and saying, okay. Now that I've gone through the the excitement, what can I actually implement? And we're coaching a lot of people on, hey. Start off small, predict risk taking. You know, really try to set your KPIs ahead of time, so that you can see what the value is because it's not always gonna be an ROI. It It could be a process efficiency, especially if it's an internal use case. You know, these are just ways to enhance our business, make better customer decisions, you know, make it a better experience for our customers. But I really think the leaders are seeing that, yes, there is value. It's now we have to kinda go back a little bit and say, okay. What do we really wanna get out of this? And what is a reasonable implementation plan? What's a reasonable, scaling up of these initiatives so that we can be successful? Is there, like, one use case, Scott, that you can point to where someone's had a tremendous amount of success or a use case where. it it it just it was you know, you had to push the abort button? So success, I will say this in the airline industry, believe it or not, we have seen, like, these, FAQ agents. Right? Automating, like, check luggage, find my luggage, like, these small little nuances that, some of the airlines have put together, and we've seen those just, like, take off, where they're think about it. They're very easy to implement, easy to track. From a data quality perspective, you know, it's pretty easy to manage. And I think those are the kind of use cases where we see that, it builds trust and that AI can actually deliver what it's doing. So, you know, going back to what I was saying before, it's like implement something small, something that's, easy to manage like an FAQ, like, where's my luggage? And put some more intelligence around it, right, in terms of, like, opening up cases and things like that. So we've seen things like those small FAQ agents, especially in the airline industry, we've just seen take off, like crazy. In terms of winning a board win, that's a good question. I I'll say this. I think it's when we have to change the mindset of, you know, I spent fourteen years in corporate IT, so I know how it go before I came to Salesforce. So I kinda know how corporate IT environments work, and it's this idea of, like, you don't have to boil the ocean anymore. Right? It's it's okay to go in small. It's okay to pick that one internal use case. Right? Whether it be like a, you know, an HR use case or something like that where you're asking questions to HR benefits or something like that. Because what you're trying to do is really build up the trust. Right? You don't want that trust gap of what I'm asking my AI agents, you know, certain questions whether it be internal or external. I'm getting back bad information because of the data problem we're gonna talk about. So it's again, start off small, get those kind of early wins, get people, you know, believing in the AI technology and trusting the results. So that's why I would say don't boil the ocean. Start off small, something very achievable, and then build up that trust factor with your employees. Great. Thank you. Let's turn it over to Sam for a second. Sam, I know you've been involved with a lot of our customers at Coastal as they have gone down the AI path over the past couple of years. It started as like a pilot process for a lot of these organizations. And some customers run one pilot install, others keep building. What's your perspective on what's different between these two groups of of companies? I think it's it's actually what you just said right there. A lot of companies that just start with one pilot and put all their eggs in that one basket, If something doesn't go their way because of a data quality issue or adoption challenge or something else, then you're not really open to exploring, okay, maybe we weren't ready for AI in that particular use case, but we know there's value to be found and we're ready elsewhere. I think the organizations and and we have data in this that's in our report as well. Organizations that run more pilots are generally far more successful than those that just focus on one or two. Now that doesn't mean they always have to have a giant team, and that that helps. Right? Budget always helps to a certain extent. But a lot of it, I think, is a willingness to say, like, look and and to I think to Scott's point as well. This is a bit of an experiment. We are all learning in this ecosystem right now where AI fits and where it doesn't fit to solve certain things. But if we can have that innovative mindset of appropriately trying and, you know, how we apply AI in terms of how we actually do it. We might talk about that a little bit later, but also where we try to apply it to solve certain challenges. And really just look at that as a listening and an analysis exercise. You will be able to find those solutions, you know, within a platform like Salesforce and and and just generally across the board. Well, we'll talk about some of those challenges that are out there and and hopefully how to solve them. You know, I remember probably about a year ago now, one of our larger customers, CIO, called me up and said, we're going all in on Agent Force, and we're gonna do 40 agents, and we're gonna start with pilots. And they they came at us with this kind of this spreadsheet approach to how to deploy AI inside of the organization. Had some success, had some failures. But I think over the course of the next few minutes, we'll we'll touch on some things that, hindsight being twenty twenty, we could have advised that client on to have a lot more success. And it starts with data. So let's talk about, you know, the I'll call it the elephant on the table, but it's not. It's actually around us all day long, and I was talking to Sam earlier today. You know, there appears to be a pretty significant difference between what consumers are experiencing with AI today, with generative AI, and what businesses are experiencing with AI today. And it it comes down to data, because when an organization deploys AI inside of a application like Salesforce or with other applications, they're dealing with proprietary data that is proprietary to how their organization runs and it has to be accurate. So we're gonna talk a little bit about that. Unfortunately, as you go down the path of AI, as an organization, you can't excuse yourself from being radically focused on your data, preparing your data, managing your data, having data stewards, having data leadership in the organization is really very, very important to making these agents work effectively. It's interesting because our report from Oxford showed that organizations face data problems before launch, after launch, during launch, and when AI initiatives stall, 70% of our respondents point to data access or data quality as the main reason. And the number of pilots you do doesn't necessarily solve that problem. So having an organizational focus on your data, having the right tools in place, having the right architecture in place really becomes critical towards driving the right automation through artificial intelligence. And, I think in the early days when organizations were launching agents, it was for lack of a better analogy, it was kind of like, oh, we can put lipstick on the pig here. I'm just telling you folks, you can't do that. Our most successful clients have taken a step back. They put they look at their data architectures. They modernize their architecture. They've deployed things effectively like data cloud and third party products like Snowflake and Databricks to truly have an operational architecture that can feed the right data into their AI. And this is a big, big deal. I'll start with you, Scott, or maybe. I'll start with maybe I'll start with Sam because I started with Scott last time. I mean, this is an age old topic. Like, get your data ready for AI. Get your data right. Like, people are tired of hearing it. Like, what is your perspective on what, like, data ready means, and how do organizations finally harness it so that we can unleash this incredible power of the technology? Yeah. You know, data readiness is a fun topic. I would encourage anybody who's in the chat. You, you can find us on we're gonna have contact information available. You can find us on LinkedIn. Please, if you're out there and you have an organization that has perfect data, reach out and let us know because I've yet to find one. Right? This has been a problem before AI was on the scene. It was just that we adapted to it more, and we weren't putting data under a microscope in the same way. If we weren't surfacing it up in a dashboard or a report, you know, we could kinda tolerate it because you and I and not everybody here was working in the org and we kinda like, oh, you know, that's all lies in that field over there is a common thing. I mean, we hear it at Coastal too. Right? We see it all the time as well. I think what we are starting to see today is is really there's this nuance around again, it goes to where are the use cases where we want to start with AI, and we we're gonna put a focus on that knowing there are going to be imperfections. One of the things that we're doing a lot of at Coastal is really trying to get ahead of that when we engage with clients. We're doing AI readiness programs right now where we're engaging with people before they're getting into or or just starting on their AI path. And a part of that is doing a data readiness assessment that we do within the org in order to get a general understanding of what the lay of the land is. That is not necessarily gonna be a deal breaker if you have a level of imperfect data for use case that you have for AI. But understanding, for example, what is critical from a deterministic perspective about how the business needs to operate and a nonnegotiable about maybe a process that needs to work as opposed to, hey. If that description field is not populated, maybe when the agent renders an answer on something, it's gonna have slightly less information around that context, and we're gonna give it some prompting in order to adapt to that. So a lot of what we're at where we need to do when we start on these efforts is get a good understanding of what that baseline data quality is and then have a discussion really around that use case to say, hey. You know, for a first phase or for an MVP or for a pilot, you know, what is our level of tolerance for what we want the AI to be able to do? And then we can go back and kinda marry those two concepts together and see what we need to fix in advance or not. And do that with every single pilot that we want to be running, and then that will become just part of a general motion on how everybody approaches this work. Yeah. I mean, I think AI is not gonna clean up your data. Right. It can I mean, there are ways it can help? Right? We are finding we can build new processes to augment that, but it in and of itself isn't going to fix the gaps in the data that you may have today. So, Scott, I'll turn to you for a second here. You know, what's the data conversation that should happen during the design phase but usually doesn't from your perspective? No. Yeah. That's a great question and, plus one to what Sam was saying too. I think one of the things we have to do is kinda have, like, a a therapy moment of our data. Right? It's not good or bad data. It just is kinda is what it is. Right? It's what it's what your company has built up over time. And if you're trying to get to this point of perfection of, okay, my data is gonna be perfect as of this point, you're never gonna get there. Because think about it. Even as we're in this webinar, your data is evolving in your organization. Right? So it kinda is what it is. I also think kinda what Sam was saying, you know, AI can help us clean it up, but I also think sometimes we get into these these meetings with customers, and we're like, oh, we'll just throw AI against the problem, and it'll intuitively know how to fix my data. It'll know what's bad. It's a it it doesn't know what's bad, right or wrong. It just knows that this is your data and you can move forward with it. So I think one of the things we we try to get into at Salesforce is, you know, have that conversation upfront and say, okay. Your data is what it is. Let's, let's really just evaluate it, start testing, start playing with it, knowing that we're gonna have to go through some nuances. Right? A lot of the training we go through internally is all about how to not just clean the data, but then make sure that it's actually giving you the results that you need and those kind of things. And I wanna say something too. I was just listening to the conversation. This is where I think we need to take a step back a little bit in terms of what we used to do, you know, years ago. Think about it. If you had a SQL database, right, basic SQL database and you're gonna be, you know, sending data to an external website, you know, letting your customers query something, what would you do? You'd build out your database and you'd run a lot of tests to make sure that as I'm asking different questions or I'm creating different information, I get the right information back. And I feel like we've kinda lost that over the years because we just assumed that all these tools would be able to get through that data and understand it and figure out, oh, what's right and what's wrong. And they it, you know, it doesn't know. It it just knows that this is the data we're giving it. So I encourage our customers, hey. It's okay to go into this knowing you're gonna have to do a lot of testing. Know that your data is not gonna be complete. It's really just the knowledge that is built up over time. It's not good or bad. It just is what it is. And one of the things we've really been trying to push forward our customers is coming up with this best data architecture. Right? So taking a look at all the tools you have in place. Right? Snowflake, Salesforce, you know, whatever, cloud, anything, and really running those scenarios, you know, end to end and saying, okay, what is the best use of this data tied to the use cases that I want? Right? Not trying to just, you know, squeeze data in where it doesn't belong, you know, reinvent use cases around my data. And and again, it's I think going into it with a realistic mindset of, yeah, I'm gonna have to do some changes. I'm gonna have to go clean this up. But I think let's all agree, your data is in constant flux. So chasing some perfect alignment. of it is not gonna happen. It's like, let's get it to a good place. Let's make sure the a is not not doing hallucinations and biases and those kind of things, and then we continue to monitor as we go forward. And it's the best practice that we can put in place. I think it's that test and Yes. validate process that is we're hyper scaling that now because Right. we need to be able to see the answers that we're getting before we put them into production like never before. And. that gives us clues as to where we need to still tweak our data architectures and and how we. need to work with these data sources. Sam, I know you had a comment that you wanted to bring in here as well, which I think is really important because we're not just talking about c we're not just talking about SQL databases anymore. Not? anymore. Yeah. Our yeah. Our our definition of data is expanding. Right? Yeah. I pulled some studies on this a while ago. 90% of an organization's data by storage is unstructured data, documents, recordings, transcripts, emails, And that generally gets the inverse amount of attention by the business and usually a far lower budget than by the business previously because you could never operationalize and automate on that data previously. We're now AI allows us to do that. And so even sometimes when we think about some of those initial use cases around things like using knowledge to surface up for, like, the the airline FAQ chat agent experience like talk Scott was talking about, data quality can also just be whether or not you've got consistent structure in your fields or something or whether they're populated. Yeah. But it can also be are the answers that we are providing the AI to then answer its questions. Are they accurate and relevant to the audience at hand? And that's a muscle and a movement that most organizations are not used to kind of flexing, and frankly, we're adapting with them, you know, in this new kind of world as well. Can I can I add one more point to that, Sam, yeah. about the data? it, Scott, please. The bucket that we didn't add to this is think about this. So you have your data and you're doing your, you know, information, your AI, but then the AI is also learning. So as customers are asking questions or your internal stakeholders, whoever it is, the AI tool is also learning off of that. So it's another feedback loop of data being added to your data that could then create some other nuances. So again, going back to the Eric's point, it's like we have to flex that muscle just constantly monitoring and iterating on this because, you know, you don't want it to get too far down a path of as it's learning, you know, yeah. Okay. Maybe your data internally is getting cleaner, but now it's getting feedback that may not put it in the right direction. So, again, it's it's one of these new muscles we have to really put in place to keep it clean. I wanna move on in a second, but I have one more question for both of you. So, you know, and I'll hark back to the days when it was like, oh, we're gonna migrate data into Salesforce. You know, the IT organization is gonna get us the data. Eventually, we'll finally get it, and then we pump it into Salesforce. And maybe there's a cleansing exercise that would happen. Maybe there was some minor integration that would happen. I'm kinda dating myself. What is different today that organizations have to think about in terms of how they're investing in their own organizational structure? Like, is there a different ownership imperative here when it comes to data? Is there what are we recommending to clients that they think about in terms of how they set up their teams to deal with this ongoing data challenge. Yeah. That's a that's a a very good question. Asim, do you wanna go first? Well, I'll take a little bit of a controversial take maybe, which is I don't think anything changes. I think the fundamentals well, now nothing changes, but I think the fundamentals still apply. We still have to build a business that operates in a deterministic realm with structure that regardless of who's accessing that data or working with it, you're getting that same experience and the same numbers that that are coming out. either way. The fundamentals are still there. Now what I think we are seeing is changing is the volume and the types of data that we are bringing into the Salesforce ecosystem, particularly through data three sixty, which we just wouldn't have done in the past because we didn't have that kind of scale capacity that tools like Data three sixty plus zero copy capabilities give us. But I would say a lot of the fundamentals still apply. If anything, they're just more prominent than they used to be, frankly. Yeah. I would I would agree with what Sam was saying. I was actually thinking the same thing is that if you think about Salesforce, you know, ten years ago, there was a limitation on how much you could bring in. Yep. You know, it was mostly, you know, accounts, contacts, you know, opportunities, those kind of things, where now you have just this sprawling mass of data you could go act out you go get p three sixty, you know, zero copy LLMs. So it's again, the same problems are all there. It's just the the volume and the mass has just grown. Okay. Well, we'll move on. But, obviously, investing in your data infrastructure and being really thoughtful about this piece of it is critical to having success. Let's talk about employees. So the other trend we found in our report is that employees are ready for AI or peep organizations think their employees are ready for AI, but AI isn't ready for them because 73% of all organizations still fail at adoption. And all cards on the table, I see that inside of Coastal today. We have various AI tools that we use to drive faster programs with our clients to assimilate information quicker, to put presentations together quicker, to do testing faster. I mean, there are all sorts of tools out there that we're leveraging, but driving the adoption with the teams is more of a challenge than I thought it would be. I really felt like AI would just go viral inside of our organization, and we have had to invest in the right level of change management and enablement to drive adoption. And, obviously, our report is showing that organizations are seeing that as well. So AI that sits unused is a is a sunken sunken cost, sunken asset, maybe. Although, Scott, you're gonna tell us about consumption pricing a little later on here. But on the adoption front, on employees being ready, and I'll start with you, Sam, you know, every leader watching has had this moment where they deploy an agent to the team. They start to watch the adoption. The curve goes like this, and then it flattens, and then maybe even it dips. You know, what's your take on why this is happening and what we can do to make these really powerful tools go viral inside of organizations? I'm thinking of that meme that goes and then I took that personally because Eric Eric knows that I'm also responsible at Coastal for driving this. And and we have seen that. We absolutely have. I think there's, there's a there's a phrase I've talked to a number of clients about this, which is really to a large extent, you have to respect your users and you have to respect the gravity of your data. And a lot of organizations that and look, no organization just has one AI tool today. I would be you'd be hard pressed to find one that does. This is an evolving ecosystem. Obviously, we have a core belief that Salesforce Salesforce needs to be at the center of that and a first class citizen in that environment. But there's multiple tools. If you're not experimenting and trying different things and challenging your different software partners as well and your own teams, like, every every good organization that that's succeeding in this is doing that. But one of the things that that we have seen is that you want your your AI to be as close to your data as possible. One is just the function of reducing that swivel chair. Right? Another is the technical lift to get that context, that data that now you've been curating to your AI agents and to where your users are working with it, so they're not having to literally copy and paste. Right? And and we still we're still challenged on it as an organization. We're still growing into it and deciding where to invest in in terms of that integration in that time. So that's that's one key part of it. I think the other that kinda what comes along with that is this concept that we're talking about is is embedded AI, which is not necessarily surfacing up AI in a standalone chat interface. Right? Most organizations are gonna have multiple interfaces like that that your users can work in. And so now it's like, okay. Well, I'm going over to Claude for this thing and Slackbot for this thing and AgentForce for this thing, and we're also at Google Shop, so we've got Gemini over here that I can use as well. And one of the things that I would encourage any organization, you know, is talking about that success of running multiple pilots, Is pilot you should be piloting the user facing, the agentic user interface kind of layer. Slackbot with the recent innovations absolutely needs to be at the center of that. We are finding immediate close success of that with Coastal right now. So a huge shout out to that team and the amazing work they're doing on that product. But the other angle is look at where you can kind of look at the existing business process that you have and just say, hey, what if we were to, like, sprinkle AI into a couple parts within this process? You can do headless agents, but you can also just do what I think is the unsung hero of the ecosystem right now which is Salesforce's prompt builder capability. And if I can run a flow that otherwise would have required a user to kind of interrupt that process and review something or give an input and now I can have a prompt template that maybe is running in a part in a flow that acts as a validation rule or acts as that additional check. We can reduce where users have to stop what they're doing and go back into another system in order to keep that process moving. And the other thing is that everybody you kind of get more consistent adoption because you're not requiring your users necessarily to change their behavior in order to get the benefit of AI. It's pausing what I'm doing and going and opening that chat interface as opposed to an automatic process where AI is running under the hood or maybe you're just surfacing it up in a button on a record on a page, and that's all that they need to do to initiate it. So take some of that cognitive load off of them and find those alternative ways additionally to be running and benefiting from AI within your org. It's a little less sexy, right, than some of the front end agentic use cases, but we are finding a lot of organizations are getting immediate and much faster confidence and value from their AI solutions by also taking that approach. Yeah. No. I think one of the things our report showed is that 50% of users don't trust the AI outputs enough to act on them. 46% reported that the AI solutions don't fit how they actually work. So, like, these are some nuggets that were and, look, anytime you try to get an organization to adopt a new technology, there's always pushback. There are always change management opportunities, I'll call them. But you've gotta build the trust first. And I think some of those nuanced examples, Sam, which you then can take to the broader organization and show them that, hey. This is working for this set of users, starts to build that trust. And, hopefully, that number goes down significantly so that the organization can leverage the technology. Scott, Yeah. from your vantage point at Salesforce, like, what does good adoption look like where AI is working? Mhmm. Yeah. Great question. And I was thinking about what Sam was saying, and I and I'll give two kind of internal things that have been playing out at Salesforce. And I think it's good advice for anybody on the call, in the webinar, is, one of the things we see at Salesforce is that, you know, obviously, we we we eat our own Sunday, and we use Slackbot and everything else in Agent Force. And what I've noticed is that there's a lot of noise, right, where it's like everybody's doing something innovative when it comes to AI. And one of the things that I we're trying to work on is let's pick a repeatable, painful process that we can use AI to solve. Because then we feel like then that will drive the adoption. You know, and we want to kinda double down on education. So I'll talk about the painful, repeatable process is every quarter, we have to do quarterly business reviews. And in the past, it would take, you know, weeks to or at least a week or not longer to pull data together, pull the slides together, everything else. And what we started doing now is using in Slackbot Slackbot skills to help us quickly pull this information together. So one of the things we're working on is, okay. Let's make sure everyone is educated to do this one process because everyone's a part of the QBR process in the company in some way, shape, or form. And a lot of times, we get the meeting on the calendar and we have to get this day together in, like, two, three days and it's it's a it's a quick build. And so that's where we're focusing on keep innovating, but let's pick one or two of those things that we can just continue to drive home and train on, that we know will be a value add to people like the QBR process. So that's one thing. The other thing that we're trying to do or we're about to start doing is, again, because of all the noise, everything that's going on, is do an AI spotlight. Right? Where either once a month or an all hands call, get the team together and focus on an AI, you know, spotlight tool that we're using. Right? Again, something in Slackbot skills or something along those lines or, hey. This is how we're using Cloud now. And, again, do a dedicated training. I think the other thing that we've kinda missed and I think other organizations miss this as well is, oh, well everyone knows how to use technology, everyone knows how to use AI, you use AI probably every day in your life whether it be chat, gbt, or on Google. And I don't think we realized like, yeah, that's in your personal life where maybe it doesn't as it's not as important or critical to get it right, where at work, you wanna have the right metrics, you wanna have the right information. So I think doing the dedicated training sessions for people, not just assuming they just know what to do, will become more important for adoption. And again, if we're focusing on maybe those one or two repeatable use cases that we know are a pain point for a larger audience, and continue to train and invest on, we feel like we'll get better adoption. So that's when and that's how we coach our customers as well is, you know, again, pick those couple processes, find the pain points, and solve those because that will make employees gravitate, to adopting more I love it. I'll I'll throw one other. nugget in there, which always has worked in the Salesforce world is get your leaders bought in, get the senior most people in your company using their own versions of AI in. a way that is driving productivity for them, and use those examples with the broader organization. The top down adoption approach, which is more around the carrot, not the stick saying, hey. You know, our CEO or our COO or our CMO, like, they're using these tools effectively, and it's working. So think about a top down approach from an adoption perspective so that we can conquer trend number two, which is employees are ready for AI, but AI isn't ready for them. I think it's somewhere in the middle there. Let's move on to our third trend, which also deals with the C suite, which is what we found in our report is that AI is only as strategic as the problem it's solving. Only one in four of the organizations that we surveyed start their AI initiative with a clearly defined business problem. I told you the story earlier about the CIO that delivered a spreadsheet to us of the 40 use cases that they wanted to deploy agents on. None of those cases came with a clearly defined business problem or a clearly defined business outcome. Like, put it down on paper. What are you actually trying to achieve? How is this gonna drive more productivity? How is it going to save time? How is it gonna deliver a better experience to a customer or an employee? How is it gonna help us to sell more products? How is it gonna help us to make our products better? Like, clearly defining the business problem, the solution, and the outcome is critical before you go down the path of 40 pilots, because three out of four of the organizations that we surveyed started somewhere else. They fell in love with a platform or a vendor recommendation, and there was no defined problem at all. So this is a this is a really important trend where there's a little bit of a less is more approach. Like, really hone in on the business case that you think is achievable and is measurable. And, you know, I think the thing that has changed so much is speed. Agents aren't hard to deploy. There's a lot of competitive pressure out there, and the AI hype cycle, news cycle are are pushing leaders into reactive decision making, before anyone's really articulated the business strategy. And what is the what is the tool actually need to do? Your strategy should come before the technology. Your business outcome should define which platform is gonna fit the problem, how you're then gonna modernize your data infrastructure to serve that problem, not the other way around. And I'll actually start with you, Scott, because you work for a vendor. Sure. And, you know, I think your your senior executives would would love to see more and more clients using Agent Force, which becomes a tool conversation. You know, some of the folks that are on this webinar have already bought a platform. Some are weeks into a pilot. They're wondering if they should have started differently. Like, what do you say to them, and and what's the move that actually recovers what. they've already invested in as we think about business outcomes and business value? Yeah. I I think it's, you know, I I would say this, you know, I it's it's keep doing what you're doing. Right? One of the things I've always loved about Salesforce and and technologies and in our space in general is that it allows for, I would say prudent risk taking and failing fast. Right? Where you can work on some things, get, you know, do some testing, take some prudent risks especially with AI, and and fail fast. Right? Assume that you may fail. Right? It's okay. You know, you learn more when you fail versus when you succeed, the old mantra. So I think it's just that idea of, like, keep doing what you're doing, keep iterating on it, don't don't get discouraged. Right? Just keep working on what you're doing. And, again, take the failures and learn from it. Do the postmortem. Don't be afraid to fail. I was actually thinking when you were talking, we have a mantra, or at least a mantra in the enterprise architecture team at Salesforce. It's think big, start small, and move fast, and I cannot think of a better way to approach AI projects, especially here if you're already invested. It's keep investing in what you're doing. Think big. Right? Think about what you ultimately wanna get to maybe in one to three years, but really start small. It's like we were talking about. Do those FAQ agents. I know it's, here at AI is only strategic as a problem it's solving. But you know what? Hey. It could be solving a very, a very painful small process that'll make your employees that much happier or your customers that much happier. So it could become a very strategic thing even though it feels very small in context. So I would encourage everyone just to keep doing, you know, failure is an option, and you will learn way more from your failures and your successes, and just keep doing what you're Do you agree with this concept, though, that you you truly need to document and and articulate what the business outcome is that you're trying to achieve? Oh, yeah. A a 100%. And, again, it that should never change for any projects that we go down. It's what do you ultimately want to be, whether it's just increased CSAT score, ROI, redeploy people. Really come up with what those metrics are gonna be so you can measure against it. I will say too it's you were saying to back when I was saying it again, like, we've gone through the hype cycle of AI. Right? Now we're being, more we're scrutinizing more of, like, what it can do, but it always goes back to what is the pain point we're trying to solve. You know, what is the thing that will make life better with this AI and really focusing in on that. Sam, what's the one question every team should answer before evaluating a platform? How well does it integrate and tie into where our gravity of data is? You know, it's a phrase I used before. Right? If you are looking at an AI platform that kind of sits over on the periphery here is kind of outside of your stack and whether or not it's got native integration or not, the more that you're gonna have to move data back and forth, the more that you're gonna have to keep those in sync, the more that you're gonna have to define on both sides what that data is and how we're working with it, the more friction you will have as you scale out, or just the more investment you will need to make in order to maintain those two systems together. And from a platform question, the question I would be asking a platform is saying, what on the other side of that, what kind of metrics and insights do you provide out of the box around how my solutions are performing? Because a lot of tools that we do see in the ecosystem, they they give a really good really good marketing, really good use case. It might even work really well. But it's they're so new and it's such a nascent industry that many of them haven't worked on that other side of actually surfacing up the data and the insights around the how the AI is performing so you know where to invest further to to make it work even better for you. But if I heard the first part of your answer, which is the one question, you think finding a platform that's going to integrate with the right data sources is the most important thing or or needs to be considered I think it's I think it's? if it can be native and it can be as close to your data as possible, I think that's one of the most critical things because the amount of time that then you the less time that you're going to spend on getting the data aligned to whatever solution it is that you are trying to drive towards for a business strategy, that's less now you can just focus on building the AI to do what it needs to do, and you can have more flexibility around that as opposed to a greater lift to get the data over to where the AI lives. So then you can achieve that goal. And like we said, if you are going to fail, is that now a sunk cost that you've had to do for that integration and that additional work? So that level of marrying the AI to the determinism that you have and kind of your platform of operation or something that Salesforce usually sits in, I think is a really, really key thing that you need to focus on. K. Great. Well, before we get to our fourth trend, folks, we're gonna actually preempt it with a poll. I think I have that right. And I'm supposed to ask the question for the poll. So I think we're gonna see it. There we go. So the question is, who owns AI decisions at your organization? IT leadership, business unit leadership. Is there a dedicated AI team, an external partner? There is no clear owner. These are all answers that we found in our report, and I'm curious to see what we have here in our audience. So I believe Miriam is tabulating these results as this is being filled out, and I should see some results. The well, the number one result, there's no clear owner. Wow. Is that actually true? We didn't gain that one, did we, Mary? I love that answer. I mean, that's one of the issues organizations have right now is that no one owns AI. There are a million AI champions out there in your organization, in the world, on your teams. People are using it daily. They're fascinated by it. But ownership is a critical question, and it and it it really kinda answers the next question, which is what sets apart the organizations that are seeing success from those that are struggling. This is trend four that we found in our report. AI champions are everywhere. AI owners are not. One of six organizations that we surveyed have a dedicated AI or AI transformation team. Five out of six put AI under IT or business leadership or an external partner. And the organizations that have dedicated AI teams are almost four times more likely to have success and have AI in production delivering true business value. This is a a a really powerful statistic. I'll date myself again. I was actually in the professional world when the Internet first came to be and when organizations started to use it for e commerce, the organizations that succeeded immediately built e commerce teams. They built teams that were solely focused on e commerce, and those teams had accountability. Those teams had a certain amount of power in the organization. They had budget. They could make decisions. And those organizations drove the future of successful ecommerce across the Internet back then. I'm seeing the same exact thing happen with AI. It's not enough for a CIO to just give you a spreadsheet of use cases that they wanna see deployed through an agent. You have to build organizational structure around this. You need owners and you need accountability. So, you know, I think we kinda saved the best for last here. This is organizational thinking. I'll ask my panelist a couple questions. We'll start with you, Sam. Some of the folks out there have been, like, named the AI owner, but they don't actually have resources to do the job. What do they do? How do we help them advocate for a new organizational structure or thinking? How do we help them advocate for more investment? What do what do people need to hear? I mean, I don't think we're really bearing the lead at this point, but obviously use our report, right, to help make this argument. We'll have a link, at the, QR code at the end of the, presentation here for that. You know, there's a lot of data that we've been able to mine that helps drill down further on this and really kind of show that delta within organizations. What can you do in the meantime? You depending upon where you're coming from within an organization you absolutely need to build partnerships with IT. A big challenge to many organizations in getting AI rolled out is a level of compliance and security and you wanna be able to get past those discussions early on. We have some clients where we work with where every agent that gets built needs to be reviewed by their security team in IT even if it's on the same platform where they've otherwise built things before. And so that may be the rule that that organization needs to work with. I would hope that many organizations can take more of a holistic approach to that and more of a guardrails on on how AI can be at least piloted before maybe a full launch, but really build those partnerships. And and on the other side, you know, you need to partner with the business stakeholders that are actually gonna be the drivers of the real use cases that you want to bring forth in a pilot. And that's where you're gonna be working with them and understanding those baseline metrics that you hope to influence by getting your AI pilot live. So working with them to understand not just, hey, we need AI to do x y z, but learning the challenge and work with them a little bit further on understanding the why behind that and understanding, well, what how, you know, how can we articulate the potential impact that will make at the organization there? Then it's just a question of scale and and how you how the business wants to invest in that further. But I would just really say, you know, no matter what, even if you have that team, that team still needs to be enabled to be able to bridge that gap between some of the legacy or traditional kind of IT functions within an organization and the business stakeholders that are gonna be the direct participants and recipients of benefits of whatever you're building. K. You know, another component here in terms of who owns AI is one of the things that is holding it back a little bit is the concept of consumption. Platforms vendors are charging for things differently, which ultimately is the right thing for the ecosystem because you're gonna pay for what you use. And we probably won't have shelfware out there a decade from now. But it's also a scary topic because consumption can be a runaway train. If you have consumption agreements across multiple platform vendors, how do you reconcile those? How do you make sure you're using your consumption in the most economical way? And I think AI ownership inside of an organization needs to own that consumption topic and be fluent on it, and it is very confusing. It's new to everyone. So I actually wanna turn to Scott because I know it's it's right in your sweet spot, just given your. title, Scott. Yes. You know, it it's consumption pricing is changing the way that customers are buying technology. Alright. Like, what are the smart ones doing differently? Yeah. That's a great question. Yeah. I think what the I would say the smart ones are are understanding that you don't have to buy into the big the big package anymore. Right? Really rightsizing your, your consumption based on your use cases, understanding what credits you're gonna use, whether it's, you know, credits. At least in Salesforce, we have flex credits. You know, we have d three sixty credits. Some are metered, some are unmetered. And we understand it's complex. But it's really, you know, holding true to say, here's the use case I wanna put forward. Here's I wanna estimate what I think I'm gonna use, put in some sort of buffer, and and buy that and use that first. Instead of buying kind of the big bang, you know, buying tons of credits that may set up on the shelf, I feel like that creates this kind of pressure to consume where the nice part about, credits these days across any platform is that you can just continue to add credits as you need them. You don't have to buy the big packages out of the gate anymore. And what I would say too is ask questions. Don't be afraid to ask questions. I mean, my my team, for consumption sellers, we want to inform. We want to be the trusted adviser to our customers. Right? Please come to us and ask questions. If you're not clear on it, don't be afraid to ask. We know it's complex. You know, we and I'll even say Salesforce, we'd like to change on how we what becomes metered and unmetered, what's a flex credit. Now we talk about agentic work units and those kind of things. Again, ask the question. Don't be afraid to ask. You know, we're here to educate. We're here to build trust and transparency. Use things like digital wallet tools, right, to get that transparency. Don't don't use don't use the fear of what of of credit consumption to block you from using things in AI. You know, we're here to help and support those kind of things. And one of the things I'll put out there too is is challenge your not just Salesforce, but anybody that goes your partners, whoever you're working with is challenge them to come up with what is the best consumption architecture with everything I have in place. I love to tell the story. We had a customer in, Iberia, where they were trying to understand what they they could do all the different credits, they had marketing cloud, they and what we did was we came up with three different scenarios of all their credits end to end and came up with the best way to manage all their credits, and they loved it. And and I said, this is what we should do for every customer is really demystify this stuff. Let them ask those questions. Really prove it. You know, build it as a proof point and say, look. We went through all the different scenarios. Here's all your different credits across all your different apps in Salesforce. This is the best one to maximize your credit usage. And, you know, I feel like that just builds up great trust trust and transparency. The customer actually bought more credits because they're like, yep. We believe you now. So, again, I encourage everyone on this call, challenge your partners, challenge your your where you're buying your credits from, to make sure that you're getting all the information you need. it's a key thing ask the questions. someone needs to own inside the organization. Right? Absolutely. Well, those are the trends, folks. We went through four of them. We have a couple of minutes for a question or two. I'll pass it back to Miriam to run that portion, before we wrap up at the top of the hour. Oh, Miriam, you're on mute. Thank you. You would think this is my first day on Earth. Sorry, guys. Thank you for our webinar. So we're gonna start, with a few questions. So I have one that says, how important is it to decide which data not to include for data three sixty agent force agent force versus what you do include? And that's from Kurt Stoll. I'll pass that one over to Scott. So that's it. I I saw that question. I was like, well, that's an interesting question. What data not to include? You know, that that is a I would say I would say it depends. It really depends on what you're doing, where your data's coming from. I I think it's one of these where you evaluate, I think Sam had said this, like, where's the gravity of your data? Where's your data sitting? Can you use zero copy? And and I would say with with tools like zero copy and d three sixty and some other tool with LLMs where you can put large amounts of data in there, I don't think it's necessarily a question of what to exclude. Obviously, if it's like PHI or PII or those kind of things, you don't wanna bring it in. But I think it's one of these things where if you believe it to be good data that can ultimately help the conversation, bring it in. We don't have the data limitations like we used to have, you know, ten years ago, fifteen years ago where it was like we were paying so much for data space. You know, and if you had too much data in there, it could, you know, slow down your queries and your reports and those kind of things. Everything has become so efficient these days that I think it's okay to bring in more than we used to. I think it's more of just evaluating, what makes sense for each use case, right, and using the right tools at the right time, like zero copy and those kind of things. So I wouldn't necessarily worry too much on what not to bring in. I would more just focus on just making it efficient and and using the right tools at the right time. And I think, you know, this is an evolution too. Right? Like, I think about it as your your architecture, where your data sits, you're you're tending a garden. You're not hard coding something that has to stay there for the next five years. So, and and you should always balance your consumption based on that that notion. Like, it's gonna change, but resources have to be dedicated to watching that change and tending that garden. We got time for one more, Miriam, before we wrap. Perfect. I'll we have one from Mackenzie. Do you think a company should have an AI strategic plan or just have their company strategic plan with end goals and AI enables approach to get results? I love the idea of having an AI strategic plan, which will obviously be in parallel to whatever the strategic plan is for the overall overall organization. But, like, pulling that out and actually committing to getting it down in writing and what that looks like, I think is a isn't is a really good idea. And and, you know, you marry the two at the top. But if I understand the question correctly, I think that's brilliant. Well, Perfect. thank thank you, Miriam. We're gonna sign off here from Coastal and Salesforce. Thank you, Sam. Thank you, Scott. Hopefully, our audience enjoyed that, and hopefully, you can take this report and continue to push AI into valuable business outcomes for your organization. Yes. And a huge thank you to you guys, Eric, Sam, and Scott for putting on a great webinar, and a big shout out to the audience for tuning in. So please look out for an email from us. We will be sending across the recording of a session in Coastal's twenty twenty six AI operation reports as well, and we also will be having a short survey. So please, if you could fill that out, we would appreciate it, and hope to see you guys soon. Thank you. Thank you. Thank you.