(upbeat music) And with Agentforce, because now all our clouds are built now on the same platform, powered by the same data, and the AI stacks. You can get that holistic experience whether you're talking to our agents or whether you're actually talking to your real human, right, in a service or sales. If you use our Agentforce platform, you can get that consistent experience, whether through AI or through human. Welcome to "Ask More of AI," the podcast looking at the intersection of AI in business. I'm Clara Shih, CEO of Salesforce AI, and I'm thrilled to be here today with Muralidhar Krishnaprasad, or MK as we call him.
Who's president and CTO of Einstein 1, Data Cloud, MuleSoft and Tableau here at Salesforce. MK, welcome onto the show. Thank you Clara, for this opportunity. Just awesome to always talk with you.
Well, in addition to leading engineering across Einstein 1 including our AI, which we work very closely together on as well as Data Cloud, MuleSoft, and Tableau, you were one of the original co-founders, if I may use that word, of our Data Cloud, which is a journey that started four years ago. And I wanna start there because when it comes to Enterprise AI, as you know, the really the most important thing is having Enterprise data. And so, I mean maybe for those who are, are newer to Enterprise AI, do you wanna start by talking about the ways in which Enterprise data is so important for AI? I think if you look at our AI journey, it started off along the world with Salesforce as well, with our predictive AI, right? When we move from predictive, predictive works on a small set of data, we are trying to make some predictions, maybe it's a lead score in case and so on. And then the wall to generative AI, where primarily it was meant for summarizing sort of things, creating nice email and all that stuff. And now we move to assistive agents, but imagine that assistive agent told you your next quarter result is gonna be, I don't know, a hundred million plus more or less. And you make business decisions based on that, that can have a huge impact on your business.
And now roll forward to autonomous agents where things are talking on behalf of you, particularly with your consumers or customers. Your brand reputation is at stake as well. Look at simple example, if I'm a service agent, I may ask a human look at five different screens and answer a customer's question. An autonomous agent is gonna reply back to that customer, it better have access to all that data. So that's why as the AI evolution is gone from predictive to generative, to assistive, to more autonomous, data becomes even more critical ingredient to that.
And particularly on the Enterprise sector, because your business is at stake, your brand reputation is at stake. And what I love about what you've led our team to do is, you know, we have those data pipelines for training models from scratch. Many of our customers are doing that primarily on the predictive side.
We have data pipelines for fine tuning, which now customers can do with no code inside of Model Builder, which is just amazing. And then of course we need data for RAG and it's just incredible this journey that you've led us on on Data Cloud from you know, what it was originally, which I'd love for you to tell that story to expanding to unstructured data, and building vector search and hybrid re-ranking and all of the things that are necessary ingredients for AI. So take us back to 2020 when Data Cloud was just starting and what was that vision then? And then take us to today. I think when we started off Data Cloud, the early sort of usage of it came from the marketing side.
At that time, CDPs were all the rage. Why? Because marketers felt they needed to give you that personalized email, right? Or that personalized engagement so that your return is high. Marketers had the big problem, data was scattered all over the enterprise and they wanted to sort of pull that together to quickly create that awesome email to send to you. But we took on a different tack, because at that time CDPs were yet another silo platforms that were coming up in the marketing sector. And so at that time, all our leaders said, "Hey, why don't we build this as part of the platform?" And that was a very monumental decision because if you remember, CRM platform was meant more for transactional data and stored in sort of relational databases. And here we are talking lots and lots of data.
We are talking like engagement, web engagement, email engagement, like pretty tremendous scales that we are talking about. As you know, our marketing cloud for example, handles most probably one of the largest email centers on the planet along with SMS and others. And so we sort of then created that journey to say, "We need to do a few things. One is we need a lot of connectors to bring data to make it really easy. We need to integrate into our Salesforce platform. So all of this appears a Salesforce metadata at scale."
And then we said, "We also need to make sure we can create that unified profile of a person." Because if I see Clara as an example, you probably have maybe different names misspelled or not, or you may have different IDs across the systems. And so we had to go build what is called as the harmonization and unification layer to kind of create that 360 profile. But then more importantly, we also had to get then make sure this was actionable in all of the surfaces, whether it's in sales cloud, whether you want to create a flow to action saying, "Okay, you did something on the website, I wanna sure somebody is aware of it."
So that really formed the genesis of what we now call it as Data Cloud being a data and an actionable platform integrated with Salesforce. But something changed along the way, which is people said, "You know what? We already have existing data warehouses, we spent a lot of money on it. I'm not gonna go now copy all my data again into your sort of data cloud." And so that's when we invented the zero copy architecture, which is becoming now the industry standard practically. And we said, "Leave your data where it is if you already have it, but we can still work on the same thing because we can overlay our metadata layer."
That is so interesting. Yeah. So, you took this existing category of CDPs, customer data platforms, which is really marketing data platforms, right? JustCorrect. Customer data for marketing. And then you said, "Okay, how do we category create and make sure that this data isn't just about marketing engagements and touchpoints, but that it includes sales engagement, sales activity, customer support cases, everything across that customer 360." And to do that, we had to really up our infrastructure to be able to power much higher transactions, much greater volumes of data. And then to your point, harmonize these data sets that traditionally no one had harmonized before.
It's a great thing that you bring up because one of the biggest failings of what we call now as MDP is that you make it an awesome email personalized, but then you go to the website, it doesn't know who you are. You call the sales person, they don't know who you are or your service person absolutely has no clue who you are. Yeah, or that person calls into a contact center and is super angry, but the marketing campaign still goes out to that person making them even angrier.
Even worse, right? (laughs) Yeah, you're angry about the product and then you get the marketing campaign saying, "Hey, please buy my product again," or vice versa. So I think that's really to your point, we've really expanded the CDP platform to really be that data cloud for the Enterprise where you truly have the customer going through every part of the journey, whether it's advertising to marketing, to sales, to commerce, to service, and then of course analytics with Tableau. Now this journey continued, right? This is all about structured data. Now as LLMs became sort of more smarter and smarter where it can now crack open all your unstructured content, we had to go expand this sort of platform, right? To support that too. Because 80% of the world's data is blocked in PDFs and Word documents and Excel sheets and so on.
And so that's really where we sort of, with the data and AI platform we have, we really again changed the game. We didn't go build another silo vector index, we didn't go build another silo search index. Instead we said, “Why can't we bring structured data, unstructured data together, but then build powerful RAG and indexing techniques.
So again, we can use all this data in the line of work." And that's really the next evolution of this platform. That's so interesting, so if I take a customer contact profile, in the past we would have, you know, structured data, so database rows and columns and fields about that customer, you know, their name, maybe their date of birth, their address, their salary, you know, their income level. And now you're saying we're able to bring in unstructured data about this person, maybe the transcript of their calls into the contact center emails that have been exchanged with this person. And so it's a much broader view of that individual. That's correct, and we can scale it to really high, like for example, internally we do this now where if you have all the meetings that is happening, the video is getting indexed, as the meetings progress.
And now I can go ask even questions, "What did Clara say in this meeting? What was the sentiment about this particular product?" And so on so forth. And so we are expanding it from structured to unstructured coupling them together, for instance, if you have a lot of cases coming in, you wanna know, you know what, kind of issues that are people reporting, and if so, are they clustered? Is it with one product, is it one region? And that's where the structured unstructured sort of comes together. And so, and again, it's not just with any one particular cloud, it works everywhere. So I can run a Tableau report on it, and really start analyzing all this data, or I can make sure our service agent as you called out before, can get the right knowledge article, with the right sort of information along with the context of what the customer has been talking about. Or even enabling things like SDR from a sales perspective as customers sort of bring in questions and answers to be able to pull in the right data behind it to answer it. So it's kind of quite, pervasive across where we are going to really take unstructured documents, which again, they are in a silo before, at best you could do search on them, nothing more.
We are really gonna then bring them into the line of work. You make it sound so simple, but I know it's not, and it's been a tremendous amount of work over the last four years. I know your teams' work nonstop, because I get Slack messages from them late on a Saturday night. So, what was hard about those transitions, right? First extending beyond the idea of a marketing data platform to a true customer 360 data platform. And then what's been hard about extending from structured data only to also unstructured data? I think the first thing is as we expanded from a technical sense, obviously a lot of work had to be done to make sure our core platform, metadata platform that was meant more for relational scale to expand beyond it. That means supporting thousands and thousands of tables, supporting billions and trillions of rows.
And how do we sort of handle some of these ingestion at scale? Like for example, every event from Ford cars, Mach-E cars comes to us at scale. So to how I handle these kind of spikes in scale and so on so forth- As a fun fact, I have a Mach-E, so you're saying that when the battery is low and runs out or there's an issue with my brakes, that data is going in a data cloud. That is correct. Very cool. So when you start, stop car it actually sending a signal up there and you know what, that signal doesn't have the driver information because that's just a vehicle thing, right? Okay, that's a relief, because I'm not a great driver so I wouldn't want you to think any less of me. So they couple it by the way they do like a real time transform to say couple it with their driver information so that that's how they know to send you an email, either congratulating you on using the product, you cross some milestone, or if you want to kind of alert on other kind of things as well. And so that's a great example of why we had to go then change and enhance our platform to be able to handle this kind of scale, right? So it's beyond transaction, it's like we are talking web scale.
Even last quarter for example, people brought in even single customers, bringing trillions and trillions of growth from everywhere. So to be able to handle that kind of scale. Then the second part is customers understand, because like you mentioned we were changing the category. When you change a category, it's really hard sometimes for customers to say, "What is this?" Because they understand what a CDP is in a marketing context like CDP, they know exactly what it is. Now we go and say, "Wou know what, this can work across your sales service." They were like, "What is this thing? What is the name for it," right? Like, there is no like analyst quadrant or analyst sort of name for it.
And so we really had to explain to them how we are changing the category. And I think that that did take a while, but then once it took on, people really realized the impact, right? Like as then you could see the change in a year where customers are like, "oh, this is awesome, what can I do more with it?" But the more interesting thing is the next level, like you said when the generative AI wave hit, and we are looking to see how could we do better? And as you know, you played a huge instrumental part in saying, "How could we really make knowledge and other things come up to the level where it can be used again in the line of work," not just for one particular cloud but across the thing. And so this required that we start understanding what it mean to do RAG, and a lot of work going on with our research team, that you sort of manage, which is to really understand what embedding models work, because we have so many in the market, right? And then our own research team created some of the best embedding models out there but then what are the techniques we should use? How do we sort of bring that together? So it was a great collaboration across Salesforce, across Data AI, our research teams to then really create that platform that can make this unstructured just work in the line of work. That really has been, I think, the most rewarding, just to see everyone come together across product engineering, post-sale, pre-sale, go to market, across products, working with Sylvia on the research team. I'm so glad you brought that up, because I really think it's our company's finest moment.
Absolutely, because I think when this wave hit to have a team that we could go actually ask questions saying, "Hey, what does this work? Should we do this or that?" And them being able to even run experiments has been hugely beneficial for all of us. And so based on all that work, it's really, and I think that's why customers are coming to us, right? To your point we're we're offering these out of the box AI use cases that they can quickly deploy where their salespeople, marketing teams, service teams are already working. And then our RAG also to your point, thanks to our research group's efforts around Atlas and other experiments, our RAG is really good and that accuracy is really important. And so, so far, you know, we've had many service workers and salespeople and commerce managers benefit from this, but our big push now is all about Agentforce.
And so that for people who don't know what an autonomous agent is, could you explain what it is and how it's used? I think the way you should look at it is if you think Copilots, we call them assistive agents, what are those? You ask a question, it can give you some valuable information, you can have some more conversations with it, but it's primarily helping you as a human. Maybe you're a service agent, you can ask some questions, maybe you're a Tableau analyst, you can ask questions. Where we wanna take it to the next step is for it to really then automate some of these tasks. So assume there is an end customer that is coming in, and they wanna say maybe the shipping order was delayed, and they're like asking a question saying, "Hey, where is my order?" And then the system needs to be able to quickly figure out who the customer is, say it's Clara, what kind of order you had placed, why is the order delayed? And then be able to even resolve it for the most part, to say, "Okay, maybe it is stuck somewhere, maybe it needs some approval or something else, their credit card it was expired or something." To be able to stitch all that together and make that seamless is where autonomous agent comes in.
But even more important, it's important to recognize that this is not full autonomous autopilot, meaning there are gonna be clear cases, where this cannot resolve the problem, it needs to be able to hand off to a real agent. And so that's really where our agents are differentiated across industry. We differentiate on many fronts. I think this is one key differentiation, because we are built on the same sort of bots framework which used to have all of this multichannel, omnichannel as well as sort of this human escalation. And so we are carrying that forward to this. So autonomous agents work off of all your data and all your unstructured and structured that we talked about.
Also all your actions that you may have, APIs and actions, and be able to then use that to quickly do the right actions, call the right actions, whether it's a shipping order, whether it's an update to do or whether you need to go update your credit card and so on so forth. But the key thing again is this is all not hard coded, while we ship a lot of out of the box agents, you will see here that our sales service, everything, every one of them is customizable. So you can customize it to your business needs. I love that.
You mentioned that it's all built on the same platform, which is very powerful. I think one question that comes up a lot from customers is, you know the agents sound a lot like bots. So what is the difference between Agentforce and our traditional Einstein chatbots? With bots, it was a fixed call tree, it took a lot of effort to get a good bot because you have to define every single step to say, "If then else, if they ask this question, what to do and so on." Whereas with agents we've taken to the next level.
But the power of LLM, we can take the intent of what the user is trying to say. If I say, "Where is my shipping order?" You don't have to predefine if they say where is shipping order, you should go call this. Instead the LLM decides, "Oh, this is talking about a shipping order."
And all you have to define is what to do, where to go look up a shipping order, whether it's an API or maybe it's a knowledge article. You may have to look up, you just have to define what kind of data actions you are allowing that agent to do and we automatically figure it out and we plan all the steps. Like I said, "Hey, what happened to my shipping order?" It can automatically look up, okay this is the object on which the shipping order is kept, this is Clara, and this is the state, oh the credit card had expired, maybe we should ask her to go update the credit card, as an example. And so that's really where agents and bots differentiate themselves. So you do not have to define all the culture because I can ask all sorts of questions as a human, and bots are much, much more natural in answering because we can detect the intent and plan dynamically. It's such an exciting opportunity.
Where do you think the first set of Agentforce agents will be deployed by our customers? We are already seeing that, right? You saw the example of Wiley and OpenTable and others. I feel service would be the first place, because that's kind of where we can really have a dramatic impact right away. And you saw some of those early results, we've seen 50, 60% even higher sort of deflection rates just with, just enabling these bots and not even like fine tuning too much, just with like basic tuning.
And so we believe we can foundationally change how customer service agents work, improve productivity, and really dramatically improve customer success more than anything else, because your consumers are gonna be much more happier if they can quickly resolve things. And then the next thing would definitely be sales as well. Like you can do a lot of automated SDRs, you can start creating the right campaigns, automatically collect them, create the leads, opportunities and so on. And that's another huge area we think will benefit. But agents as a whole, you're gonna see it everywhere. My favorite is also MuleSoft, like you can really make your mainframe stock.
Can you believe it? I can now talk to an AS400, because MuleSoft can make that work. Or even a website that may not have an API, you can actually make it talk English, so that's- So, you can just speak natural language to any machine? Yes, think about it, you can speak natural language to your app, your website, or maybe your dotted visual code that you wrote, I don't know, 15 years ago that may not even have an API, we can actually make that talk with our agents right now. And so that is super powerful. It's just incredible how far this technology has come. And I mean, it goes back to your point about how important it is to have that trusted data, because I don't want my agent to be autonomous if I'm not pretty darn close to a hundred percent sure. That is absolutely true, because you're, like I said, your brand reputation is at stake, your company's business is at stake.
So you better get, make sure you get the right data, and our agent data and our overall platform makes it really easy for you. So it's not like an onerous task. You don't have to go keep training models, and everything, we just make it all easy. So at every moment when the agent is looking up data, it's real time updated through data cloud, across structured unstructured data, and it's both CRM data that customers already have in Salesforce.
But it's also, as you were talking about the zero copy network, it's also all of this, this data that's in, you know, a Snowflake or a DataBricks or a BigQuery. All these data lakes and data silos that companies have. That's correct, and you should also add chat transcripts, video, audio, all of that as well that's kind of pouring in. You can truly get the picture of how your business is doing, how your consumers are sort of responding to your business and be able to act on it.
I think that's an another big thing. Our Agentforce is not a passive platform, it's an action platform. So you can actually act on all of those changes that are happening in the system.
So let's run through an example. I mean how, how would an Agentforce SDR, so sales development rep agent, how would it use data cloud to then take action? So let's take this example. you have to go talk to all the leads, right? So which means you're probably running some campaigns. I can just talk to the agent, you're talking to our agent that says, "Please create me a campaign.
I wanna go target people who are maybe are high net worth individuals living in Southern Bay Area," and it can actually use now Data Cloud to go curate all that data. But this is all structured data that's probably coming in from different sources, salary information, all of those things. And it could go automatically create the segment for you, and then you can even activate it. Maybe it's email campaign you wanna run, maybe it's an SMS thing. And so you can now activate that campaign and as people are responding to that, it now becomes leads in your sales cloud, right? And then the SDR can now take those leads or people who are responding and then automate to qualify which leads would make sense or not.
And this is where the predictive part comes in as well, because you can actually have created predictive scores to say which lead might convert based on the source and based on maybe the questions they're answering in that form or whatever else you have sent. And so you can even automatically say, "Okay, these top a hundred, 500 people are the right one to convert," and Agentforce can now qualify these leads automatically and even create now the follow on emails, or the follow on calls that they have to send or maybe some people are important enough, it can even escalate to the salesperson, and say, "You should go talk to them right away because this is a great opportunity." So imagine that loop from campaign, to lead generation, to really understanding and predicting who probably are the most important people to talk to either escalation to the right sales person, or to even generate that follow on sort of email, or even calendar scheduling sort of with that person.
All that can be accomplished with our Agentforce platform. That's just amazing, it's like the holy grail of bringing marketing and sales together, and addressing, you know, this perennial issue we've had where so much of a salesperson's time, I think 70% is the latest research, is spent on non-selling activities, just pure admin. And so now for us to be able to use Agentforce, use agent technology to free up salespeople's time and literally serve them that qualified lead and meeting on a silver platter. And even amazing, I just saw recently a demo where based on the transcripts you now have with that customer and that's also coming into Data Cloud, unstructured indexes and everything else, your next email, it can autogenerate with such sort of clarity because again, it's bringing in not just your unstructured data, okay? You may have had conversations, they may say, "Oh I don't like this, I like this," et cetera.
But also everything else that they have with the system because that's structured data, right? Like what have they purchased, what are they using, not using? And so you can now start having that conversation if you may, will be to the next level. Imagine how much time you're gonna save, because you need to probably look up as a salesperson, you probably have to look up five different screens, five different applications, pulling all that data together to create that email. All that is like in your hands right now.
And so you can really do what you do best, which is selling rather than collecting data and trying to figure out what to do. I love that too, because you know that that meme that shows the iceberg where most of the iceberg you can't see because it's underwater. That's how I feel about Agentforce, because on the surface, the agent itself, I mean it kind of looks like a chat bot, it looks like a lot of agents out there, but that's really not where the magic is.
The magic is everything beneath the surface, especially the data, and the flow, and the trust layer, the predictive models that it can invoke as you were saying, that really makes the agent usable. No, you're you're spot on on this one, and like I said, the bar is much, much higher in the Enterprise context because every answer we return, particularly in the consumer side, has to have that high bar of trust, has to have that high bar of accuracy. And so that's really what we are pushing, pushing sort of harder in this Agentforce platform. The reason we had to go build this was people have invested a lot in warehouses and lakehouses and they didn't want to go invest in yet another, "Oh, let me go now copy all this data now to Salesforce."
And so we pioneered the zero copy in bi-directional way. When I say bi-directional, first of all our Lakehouse platform, we already have a, we also have a lakehouse, we are obviously beyond a lakehouse, but we also have a lakehouse underlying storing all our data and we pioneered the use of an open source technology called Iceberg. And before, if you remember in all warehouses and lakehouses, people wanted to, those are all data gravity systems, meaning they wanna pull all the data and they wanna create their own format so nobody could copy the data out, right? Yeah. Because their monetary incentive is to gather more and more data. We actually took a different tactic. We said, "We wanna be open from the start."
And so we said, "We are gonna standardize on Iceberg," an open source sort of format at that time. And that was a risky move several years ago because Iceberg was still relatively unknown at that stage. And then we worked with our partners such as Snowflake, and Redshift, and BigQuery, and others to say, "You know what, we have stored all this on the file system as Iceberg, could you read directly from us?" And that was quite a bit of a change for all of them and they kind of opened the industry, but they saw the value because Salesforce had such wonderful content in terms of all our customer engagement, all the customer data. And so they saw immediate value in working with us.
And so that's kind of where it started with the data out where we can essentially share all the data we have in an open way with the other sort of platform vendors. And that led to a tsunami of all of them supporting Iceberg natively. And really then shifting to that becoming the open source format for all data. On the river side, we also had the other problem, which is data was stuck sort of in all of these warehouses, lakehouses, the businesses weren't seeing much value. The business person wanted to go faster, they wanna do other things, and that required them having to go work with IT to go create changes, and maybe build all these ETL pipelines and all that stuff.
We said we could make that simpler too. So then we built the zero copy data in, where we said, "If you have an existing lakehouse, warehouse, doesn't matter, we can talk SQL to it, or we can talk, we can read from the file as a case maybe, and we can just integrate." And that was a huge unlock, because now every single thing, if you saw the FedEx example as an example, they run on Azure, they have a lot of data sitting there, billions, trillions of thing. We don't have to copy any of the data over to Salesforce.
We literally can work on their existing system but provide the business value in terms of sales, service, and many, many other things. And so that's why the zero copy network is such a huge win. We truly bring it and business together, and we are able to leverage all the existing investments people have done in their data stack. I love that, because I mean I think about all the data that is in these data lakes and data warehouses today. Like a customer, you know, going to the FedEx website, business customer going to FedEx website, and looking up different subscription options.
That information, the fact that they visited the website, and looked at these certain offerings, that is so crucial for the salesperson to have. But to your point, you know, traditionally, I mean you'd be lucky you'd be at a very well run company if maybe once a quarter you got a report of those website visitors, versus I think what you've done for them is in the industry is now you've unlocked that data so that that real time web engagement activity, mobile app activity, can make its way directly into Salesforce for that salesperson and for Agentforce to be able to use. Absolutely, in fact, we recently had huge success, but our own Salesforce team uses all our platform, and we've shrunk now lead capture from hours or even days to like seconds, seconds and minutes. So, it's quite a lot. So imagine you're on the website looking at stuff and if you're a highly sort of important thing that you get a call immediately to say, "Can I help you?" Right, I think that is a tremendous change in how we do customer success as a whole.
Yeah, I mean just as a customer, that's the kind of service that I want. (laughs) Right, exactly. that's the other beauty which is we are blending the difference between all of these, because as a customer you don't think in silos, you don't think this is marketing, this is sales as a service. You think like, "Okay, I'm engaging with the brand, I'm engaging with that company. Maybe my first question could be inquiring about sort of marketing like content."
Quickly it might become sales, quickly, it might become a service. And with Agentforce, because now all our clouds are built now on the same platform powered by the same data, and the AI stacks, you can get that holistic experience whether you're talking to our agents or whether you're actually talking to your real human, right, in a service or sales. If you use our Agentforce platform, you can get that consistent experience whether through AI or through humans.
It's a true customer-centric approach. I mean it's a customer 360 degree view. That's correct. It's just amazing how all of these different pieces have come together. One of the other very crucial pieces of making this work is MuleSoft.
You mentioned APIs earlier, and how we can use natural language now to talk to old on-premise machines. Could you also talk about how MuleSoft APIs makes Agentforce more powerful? So MuleSoft gives a whole bunch of things, right? One is, it has a whole API management service. So you can actually expose all your enterprise APIs, through a protected service, and through a governed sort of API management layer. What we are doing is to make all of those APIs automatically come up very easily, with few clicks appear within sort of Salesforce. So they become invocable actions that could now be used in agents or even in automation like tool.
That's one part of it. Second is MuleSoft allows you to easily create what are called RPAs. So if you have old applications, let's say like I said a visual, I don't know, VisualPro or whatever old applications you build, doesn't even have API, it's just a website.
You can quickly run these things called robotic process automation that can actually then do the actual clicks instead of API calls, but it appears as an API from sort of through the power of new. And you also have things like intelligent document process where if you even have images or documents, you can then convert them into APIs, that you can start querying about, right? All of these are now available to use within the Agentforce platform. And so with that power now your agents, I can start saying, "What's my inventory of my car?" Maybe, and that inventory might be sitting in your NetSuite and now your agents have access to that. So I can, as a consumer or as a human, I can ask that question, you can quickly pull that up or I can even start doing actions on this, maybe old website you've written that is supposed to go update your loan documents.
You can say, "Can you please update your loan documents," and in English it'll automatically then convert to the APIs and maybe execute the RPA. So with MuleSoft we can make your entire enterprise application come together and sing within Agentforce. So it's breaking down silos, not just across the front office but across really every part of the enterprise. Because you can put an API on every application. That's correct, and so I look at, this is why I look at two sort of parts of the coin.
One is data, we can do all the data integration needs using the power of our data cloud. With MuleSoft you can bring all your applications and APIs together, and with data and APIs together, now you have the full sort of platform benefits to be able to either query the data, read the data, et cetera, or to be able to action on all of the APIs or call the APIs in action, all of that. What I love, too, is you've got two layers of protection.
You've got the Einstein trust layer that governs all of the AI requests, but then you also have MuleSoft, it has governors to make sure that the APIs are accessed securely and according to the permissions of who's asking. Correct. And even bigger, right? You wanna make sure old application may not be able to handle sudden like bursts of volume and so on. So MuleSoft helps you not just with security but also policy governance limits. Like maybe it can only do 10 requests per second, MuleSoft can do all of that stuff for you automatic in all your existing enterprise stack. That's pretty incredible to think about.
You know, a system that was built 50 years ago in Cobalt, that MuleSoft can help buffer it, and make it usable by a modern agentic system. Yeah, it's pretty fascinating, right? Like I think the tagline I use is, "Talk to your mainframe," or "Talk to your Cobalt." It's real and it can even go even further because think about the amount of complicated code you've gotta write to just get two systems together. Let's say if you have a travel kinda system, you may have to say, "Oh, please call this API to say what the traveler is, the dates, the calendar and so on so forth." But now you can write in English, say, "Please book your ticket for me next week. Preferably in like maybe some airline that you like and going between say Seattle and San Francisco and so on," the power and the simplicity we can actually bring to this integration and automation space is also huge with the power of Agentforce.
So you also lead Tableau, and the data analytics world is also going through this tremendous revolution. Could you talk about that? Yep, oh, wonderful. I think there's a lot of changes happening in Tableau. I think two big ones that work today, that we are super proud about is Pulse. So if you think about Tableau, it has been primarily an amazing data exploration interface creating dashboards.
But you still have to go, to go get, look at a dashboard, or go create a dashboard. And so with Pulse that we introduced, you can actually get the pulse metrics to your Slack, or to your email every day or even sooner, right? And so that can give you a snapshot of what's important for you. Maybe you're looking at your sales, maybe you're looking at your, I'm an engineer, I look at the bugs that I have. And so we use that across Salesforce. It's really democratized analytics. I mean I'll just say everybody on my team, we're constantly looking at our AI, of course, our sales numbers, but also our product usage numbers and how it's trending.
I'm not a trained data analyst. I wouldn't know how to to build a visualization, but now that it comes to me, I feel so informed in the moment. That's right, and you can now take actions, right? So we made it into a action platform practically. And we also have the Tableau assistive agents where you can start instead of tracking and dropping, you can write natural language to be able to create your dashboards.
But more interestingly, we are also taking it to the next level with what we call as Tableau Einstein. Everything by the way, we talked about in agent platform today, like Data Cloud, AI, all that is usable in your current Tableau, right? You can just connect with one click all this data, unstructured data as well is now usable in your Tableau. But the challenge is still Tableau, and sort of Salesforce two different platforms, right? Your work is done here with your sales service cloud marketing, your Tableau is kind of here, even though we have connections, data connections is still separate. So as a simple example, if you're a marketer, you love Tableau, you create all of these awesome sort of campaigns and who you need to go talk to, but then you're to now export it, copy to your campaign manager here on the other side to run those campaigns. And so with Tableau Einstein, we are really changing the game here, Tableau is now gonna run on the same Agentforce stack, and we are also foundationally changing even there, like if you think about what's the power of all of these analytical systems, we call that the semantic model.
Semantic model is where you define logically your business model. Okay, this is what sales mean, this is what service means and what are the KPIs, what are your ROIs and so on. Till now, in most of these systems that tends to be with the analytical visualization layer. In Tableau we call that VizQL, and other systems, Power BI, LookML, and other things, they have their own thing. But they're still only in the visualization layer.
So which means you define a KPI like sales ROI, that's only usable there, not anywhere else. What we are now doing is we are create, we abstracted that semantic model as a layer, and we are now also bringing it to data cloud layer. So which means you define your KPI once the entire company is gonna consistently use that same KPI, and that same KPI is usable both by your analytics, and by your AI systems as well. So your agents will have the same KPI as your service or sales ROI. I love that so much, because it's that context that is so important to drive the accuracy and relevance of AI, especially when you're looking at agentic use cases.
I mean, just a simple example, I was with a bank this morning, and they were talking about their CIO, and for the longest time in the meeting we assumed that they meant their chief information technology officer because that's what most companies, many companies refer to as a CIO. But in financial services, a CIO often means their chief investment officer. And so that's a semantic, is a unique semantic attribute to this company and to many financial firms.
And there's many, many examples of that, whether it's from a naming standpoint, or what someone means when they say sales numbers, that's different in company A, versus company B. So I love how we've invested in bringing that semantic layer and that semantic model across the entire Agentforce stack. Yeah, and this is why the answers you are gonna get on top of the semantic layer are more accurate, because it is embedding all your business semantics into it, right? Like what do you mean by sales, KPIs? And so it's not left to interpretation. And so the second thing we are also doing is all the goodness that Tableau had, things like I can easily spin up a site and do my own exploration, right? I don't have to wait for some admin somewhere to go tell me, create me an org and all that. So we're bringing all that goodness, we call them personal workspaces into Salesforce platform. And you're also gonna have marketplaces now built around it which, and to be able to sort of monetize and other things around it.
So Tableau Einstein will be a complete reimagination where we are bringing the goodness of Salesforce platform, our data AI stats, the goodness of our Tableau platform in terms of citizen development, exploratory interfaces, the ease of use, and the power of agents, where you can now start doing things like data Q&A easily and Pulse to your email or Slack or other places. That is the, sort of, I would say regeneration, reimagination of our entire analytics strategy. It's so exciting. MK, you have had a fascinating career, and you exemplify what it means to be, have a growth mindset and to bring that into your leadership. Tell us about your personal journey and how did you learn those crucial leadership lessons that have made you so successful? So it's a long career.
I started off as a developer back at Oracle, and it was fascinating. I was in fact talking to Marc about this is, if you think about the power of marketing, I learned where we had a very mundane release, it was 814 something and Larry renamed it as Oracle 8i, and then immediately it became the database for the internet, right? Like, so that was my first learning as the power of marketing, right? Which just in simple terms, but it can have a profound impact of how your product sort of works. And that's kind of where, Oracle is where I learned all my technical skills as well as learned what it means to really be able to talk about your product and take it to the next level. And then I continued my journey in Microsoft and there was a lot of learnings, both people and technology.
I think our first learning was in the cloud thing. Microsoft's at that time very early in the cloud, and we learned a lot of what not to do. We had taken bets on various systems without understanding what it means to have (indistinct) and other things. And so it was a hard journey, but we also learned a lot along the way of how you build for the cloud, right? Building for the cloud is not like building for your server sort of software, because monitoring, logging, those are all not afterthoughts, but has to be part of your earlier thoughts, right? And also learning why it is important for us to really have that growth mindset in terms of really learning on a day-to-day basis. Like actually having DEBS being involved with the life site with the customer so that you actually will write better software.
It's not someone else's problem, it is sort of our problem. And then from a career perspective, also learned what it means to run, whether it's large or small organizations, how to really bring the entire team together as we go through a journey, because these are all new journeys, like cloud was a new journey back then, and now we are going through a new journey, whether it's AI or data and others and really be able to inspire them, tell them what's possible, and then also make sure the customer success and other things that they get back. Because that in the end, it's not just about money and others, people are here, they're working long hours because of that mission, right? They need to learn what is that mission, and really feel satisfied and that satisfaction comes when you see that customer using your product and seeing some sort of value in what you've built. And those were kind of the big things that I learned along the way and I try to use it in sort of my day-today thing.
And Salesforce has been another huge learning experience for me personally. I think I've never seen any other company exemplify sort of customer success as well as having that relationship with the customer. And that was a huge learning for me personally as to how more than product and technology, how you have to create that relationship with the customer. So they are involved in the success of the product as much as we are. I really appreciate how much time you spend with customers and you've really made sure that you spend the time with our early customers on both the data cloud, and the AI side to ensure that we're building the right product and platform.
Thank you, Clara. I think I also learned from some of the great minds like you on how sort of you work with customers. I think it's also the bringing them together, because particularly when you have a big change like this, customers are also equally confused, worried, right? As everyone else is.
And so really working with them, talking to them through the whole tech stack, explaining what it is, so we demystify what it is so they understand foundationally what is it that we are saying. I think that is really helping a lot. Well, you've done a fantastic job of that and I look forward to continuing to partnering together on that. My final question for you that I always ask guests is that, you know, whether you're hiring or thinking of raising kids, grandkids, what advice would you have on how we should be educating the next generation to, to succeed in this very new age of data and AI? Oh, that's a really good question.
I have my own kids, and I know I constantly talk to them about it. I think the biggest thing we can ever do is to make our kids curious and then make them have some passion wherever that passion takes them a lot. To me, curiosity and passion is what can really make you succeed. Whichever area that you sort of wanna take a bet on, whether it's computer science, and building the next AI models, or whether it's mechanical or electrical, whatever else it is, that curiosity and passion will take them long. So as the next generation, that's really what we need to instill in them, not the fear of what may happen.
It's really the fact that be curious, be open to new ideas, and really be passionate in whatever you do. Go do that thing that makes you wake up and want to do that work every day. That's really what we should be instilling in our next generation. That is very sage advice. MK, thank you for your tremendous leadership and for bringing curiosity and passion to Salesforce every day.
Thank you, Clara, this was so great. What a fascinating conversation with MK. Three takeaways from me.
Number one is that the most important underpinning of successful, trusted Enterprise AI is Enterprise data. Number two is that Data Cloud and MuleSoft extends the power of the Agentforce platform beyond CRM to every Enterprise API and Enterprise data source. Last but not least, MK believes that the two most important qualities for success are passion and curiosity.
That's all for this week on the "Ask More of AI" podcast. Follow us on YouTube or wherever you get your podcasts and follow me on LinkedIn and X. To learn more about Salesforce AI, join our "Ask More of AI" newsletter on LinkedIn. Thank you and see you next time.
(upbeat music)
2024-09-13