The New BRAIN Of the Enterprise | Ryan Gavin

0:00I'm old enough to remember typewriters in carbon paper. At what point do we hit sort of as fast as we can work? How is this going to affect not only the enterprise but the economy? For many years, like the narrative is, well, you know, A is not going to replace jobs. I'll be honest, I've said that line many times. This is probably the first time I believe it. They have now more agents deployed in their organization than they have employees. They have more digital workforce than they have human workforce, which is incredible. Where do you see this going as these agents become more capable? Do you imagine Slack as becoming kind of the brains of the enterprise and employees are talking to the brain and the brain is then going out and executing actions on the part of it?

0:46So, yeah, give me your background, how you got to Slack, and what Slack as the agentic OS operating system means. Sure. Well, Juan, thanks for having me. It's great to be here. So, I'm Ryan Gavin. I'm the chief marketing officer for Slack. Really, just middle management overhead is really what that means. I've been here for about two years. I have good fortune to be able to work around kind of machine learning and artificial intelligence for many, many years. I was at Microsoft for 20 years in which time I got to work on search business, which obviously was heavy into machine learning. At times, I helped to build out the business and marketing practice at AWS on the machine learning and AI side.

1:35This is now five, six years ago. I worked at a startup that was looking to disrupt the supply chain and trucking industry using machine learning practices. And so I was fortunate to kind of come over to Slack two years ago to really kind of be part of this kind of incredible transformation that's happening right now, which is this kind of new moment for Slack, which is almost like what it was built for, which is really to become this kind of we call it kind of an operating system for work, which is bringing together a platform that has kind of not only all your people and your teams connected, But now connecting all your data, your structured data, your unstructured data, and connecting all your workflows, all the enterprise applications that you're using, all the systems of work that you bring, and bringing all those together now even with agents.

2:19And how do you bring all that into one kind of workflow for organizations and for people, which are some of our most important assets? And how do you have them interact with it through one of the most powerful tools that we have, which is conversation? I don't want to have to learn another tool or a tab. I just want to be able to talk and have it do what it wants, what I want it to do. And that's this moment that organizations are seeing right now with kind of this transformation, this AI period. So happy to talk more about it, but that's the short overview. Yeah, well, we're all familiar with Slack bot.

2:50I mean, anyone who's used Slack knows Slack bot. You have the latest Slack bot? Can I just interrupt you? And the Slack bot kind of has existed since the beginning as like notifications. But do you have the new, do we have you on that? I don't think I do, but that's what I was going to ask. Is this now a Slack bot that will actually do things rather than remind you of things? Yeah, that's okay. So at the end of this call, we're going to get you hooked up on the latest thing. So we'll take that as an action. So think about Slack bot. Yeah, Slack bot has this rich history of Slack. It's kind of this playful character that gave you notifications and things like that.

3:27But it wasn't anything like a personal agent for work or a new assistant. And what Slackbot is now, honestly, is like the new interface for work. It's incredible. And it is the closest thing I've seen to a chat GPT moment inside of the context of the enterprise. Like it is the chat GPT moment for work. And the reason why is because, you know, while AI tools are unbelievably powerful and we all experience them in our consumer lives and, you know, using these LLMs to do just unbelievable things and they understand the Internet and they understand all the context of that. What they don't understand is your business context and they don't understand you in the context of your work.

4:05And that is the holy grail for getting agentic AIs to be able to truly unlock employee productivity. It needs to know me. It needs to know my work. It needs to know my business. It needs to know my data. And it needs to have access to that in the same way that I would have access to it, which means it's permissioned. It's controlled. It has governance over it, et cetera, et cetera. And that's exactly what Slackbot is now. Like the kids would say Slackbot has a glow up. like Slack bot has had a complete glow up in the world of Slack. And it is now this personal agent for work that is allowing me to do anything that you can really imagine as a personal agent.

4:40So I can, you know, the basic tasks have a scheduled meeting system, easy. The more advanced tasks, something I did at the beginning of this week, Slack bot, what are the areas that my team needs my focus right now? What are the five areas that I'm probably getting behind on my priorities? give me a briefing overview for any customers I'm meeting with this week, and be sure to update these four opportunities from these customers I met with last week along X, Y, and Z. That's a very common prompt that I will put into Slackbot. And what's amazing about it is I'm not telling it which customers I met with.

5:13I'm not telling it who my team members are. I'm not telling it what my priorities are. It knows that because all that context lives in the unstructured conversations of Slack. It lives in the structured data that's supported by Salesforce. And it can pull and access all of that and bring together an incredibly powerful response that sets me on my way that otherwise would have taken maybe a half dozen meetings. Maybe I would have to call a staff meeting. Maybe I would have had to send six or seven slacks or read maybe two dozen channels. And all that just happens like that. It's transformative.

5:43We have a customer, Mr. Beast, who Mr. Beast Industries, one of the world's largest content creators, is incredible. Their employees are all up on Slack bot now. one of them saying they're saying like 20 hours a week of their time is being saved 20 hours a week it's like okay that's interesting you do the math that's six months that's six months of time given back to go focus on you know creating their next major project or whatever it is um it's like you know there's a lot of buzzwords that are thrown around this stage but it's transformational it is like chat chapiti moment for sure yeah well and i'm i'm excited to hear about it because i i I've been waiting for these agentic systems to be integrated into products.

6:24I use a product called Descript for video editing and things like that. And they now have an agentic assistant that's incredible. The one thing in your example there, you were telling me what you would tell Slack, but you still have to type it, or is there a voice interface where you can just talk to it? Yeah, we're innovating really, really fast. In fact, we just recently launched. Now I can just talk to Slack, so I can just grab them. I'm going to my car. After the end of the day, I can say, hey, Slackbot, do all those things I just said. I can speak to it. We have new capabilities that allow Slackbot to be your meeting assistant, so I can ask Slackbot to be a participant in a meeting, capturing notes, following up in actions.

7:14And what makes it really differentiated again is unlike a lot of other meeting assistants out there on the market that can transcribe notes and do all those wonderful things that we've seen, what Slackbot can do is because it knows your business and it has access to the systems of record of your business, when you are asking for an action in a meeting, Slackbot can actually go take that action. So go update that opportunity record or go schedule that meeting or go get this set up with, you know, inside of this system. Slackbot can actually take the actions from the meeting. So when you're finished with a meeting, actions are already being done and taken for because you have a system that knows and has access to your organization's system.

7:50And that's the powerful unlock that differentiates a lot of these features, which we're all kind of blending together. You see a lot of AI capabilities. They all kind of like, yeah, they all do roughly the same thing. This is what makes it different. Yeah. And what's amazing for you guys is you're already installed in a huge section of the enterprise. I mean, I don't know what the metrics are. What are the metrics? How many enterprises in... Well, I'll tell you, there's certain data we don't share, but the things that we share, there's like 6 billion messages being sent in Slack a week. And what's interesting of that and why I'm excited to get you stood up with Slackbot is when you think about those messages, you know, I'll tell you just a small personal anecdote.

8:43You know, those messages represent the long-term memory of your company. But historically, that has been largely inaccessible, right? You have the conversation, it goes away. Like if I ask you, hey, what was the conversation you had in Slack four weeks ago with so-and-so? You're like, yeah, I don't know. Like, what was the context you gave? Well, you don't know. It just kind of goes off into the ether. Well, that's unbelievably valuable context. When I joined Slack two years ago, you know, I was fortunate enough to have people who were assigned to help me onboard or help me get up to speed in the company.

9:15And here's a document you can read and here's people you should meet, all that kind of lovely stuff. Number one onboarding tool was our new AI search capability inside of Slack, which allowed me to access that long-term memory. And I could go and just sign Slack and I could ask all the quiet questions like, you know, what's a V2 mom, which is our tool that we use here at Salesforce for establishing growth. Or like, you know, someone would put an acronym in like, I have no idea what that is. Or, hey, this conversation has been happening in this channel. I have no idea what they're talking about.

9:44Can you explain this to me? Like I'm a layman and I could go in there. It was my number one onboarding tool. I would go up to it like 30, 40 times a day. And it was giving me this rich context of the company, the history, the accounts. I'd go to a new customer meeting. I could go through and see the journey that customers been on, not because it was all in the CRM system. The CRM system had how big the opportunity was and what their phone number was. But what was really valuable is looking at the 400 thread conversation when they were really trying to fight to get this deal done and what was happening and what were the dynamics and all that.

10:15All that happens in conversation. And now that's unlocked. And that is becoming unbelievably powerful. And those are the tools that really start to go through. And that's why Slackbot's kind of having this moment, because when you unlock that and you couple it with the structured data, you can take workflows that previously might have been days to weeks, and you can take them down to literally seconds. Yeah. You're talking about CRMs. Does this integrate with any CRM system or is it with specific systems? Yeah, no. So that's, again, beautiful part of Slack. And, you know, why we talk about as an operating system for work is it's an open platform.

10:57So, you know, whether you're a company like Vercel or Perplexity, you know, which recently released their computer agent on Slack. You know, Slack is one of the fastest going platforms right now for agents and people building agents and they're deploying them in Slack. And the reason why they're doing so is because that's where the workers are. That's where people are working. So instead of asking employees to go to yet another tool or to learn yet another thing, bring it into the flow of work. And that same thing is for CRM. We have native CRM built right inside of Slack, which allows small businesses to go through and get started with customer management, deal opportunities, etc.

11:31and allows them to seamlessly upgrade to Salesforce as they start to grow and scale. But it's all done through conversation. And that same capability is now extended out to third parties. So anyone who wants to have Slackbot, for instance, talk to their agent that may be wired back to a system of record like a CRM. Slackbot is now an MCP client. So any MCP server can call Slackbot. Slackbot can talk to it. And so you can use Slack bot as that interface for your system of record, whether it be a CRM or an HR system, DocuSign, a notion, whatever it is. And that's the that's the real transformation.

12:08You know, people talk about like the Sasspocalypse and Sass is dead and whatnot. It's not it's not right. The right way to think about it is like bad UI is dead, like bad tabs and browsers and me having to go figure out how to work with your thing. That's dead. And what it's being replaced for is a conversation the way I want to have it with an agent like Slackbot that can now be orchestrated across all these systems of record and increasingly all the other agents in your organization. You know, I spoke to a customer just last week. It's incredible. They have now more agents deployed in their organization than they have employees.

12:46Wow. Just let that sit in. Like they have more digital workforce than they have human workforce, which is incredible. But there's a secondary problem that comes from that, which is, you know, very familiar if you've been around enterprise software for the last several decades, which is now like, well, how do all these employees know which agent to go to for what task and what skills they have? And how do I find them and et cetera? Well, that's where orchestration comes in, this idea of like a super agent. And Slackbot can being this MCP client can now know all those other skills of all those other agents and can invoke the right agent and prompt them on your behalf.

13:19So we have an incredible sales agent here at Salesforce that we use that really knows and understands our customers, et cetera, and helps our sellers sell. I don't use it, but Slackbot does. When I ask it questions, it goes and queers the sales agent, gets that special expertise that was built by AgentForce, and then brings it back to me. I never have to learn and find out where it is or how to manage it, et cetera. that orchestration becomes a key component to kind of unlocking the gains of this kind of agent proliferation era which we're about to be in or we probably are in right now yeah i've got to ask the company with more agents and employees what industry is that and are they a a young company because there there are these agent native companies that are springing up to challenge legacy enterprises and variants.

14:08Yeah, they're a large and growing AI company. You've heard of them for sure. I'm not going to say their name because I didn't ask about this specific stat that they gave me to share publicly, but you've heard of them. They're on the bigger side versus the smaller side, just to kind of put it that way. But you're right in that there are these small startups that can go through and can act like very, very large companies because they are understanding the leverage of having something like Slack as this operating system, the system of engagement, and then you're bringing in all your agents and having a single human who can orchestrate work.

14:45That pattern is very powerful. It's never been a more exciting time to be a startup, no doubt about it. And I think it's going to be a challenge for many enterprises to say, can you transform at a rate, can you transform how your teams work at a rate that's going to keep clip with those startups that really have very little barriers right now because, you know, they don't have all the historical governance and all the historical software stacks to deal with. They all don't have the data policy privacies that they're, you know, struggling through. And that'll be the interesting question for these large organizations.

15:16And when I spend time with customers, whether it be like, you know, sitting down with Engine or Remarkable or any number of slalom, et cetera, I don't, as a chief marketing officer, I'm not really selling Slack really at all. I'm almost exclusively just talking about transformation of your employees and how do you transform what your employees are able to do and to get the productivity levels that they're going to be required. Because most people aren't adding people, but they need to add impact per person. And how do you multiply the impact per person by a multiple, two, three, four, five X? That's the question.

15:52And that's the transformation that's kind of happening right now. Yeah. And so Slack is acting as an orchestrator. The actual agents, are they coming from AgentForce? The answer is both. So AgentForce, as you know, is Salesforce's platform for building out these highly specialized agents that can be customer-facing, that can also be employee-facing. And these agents can be tuned, trained, managed. You can have deterministic controls over them. You can go through and monitor their performance and tune it over time. These are kind of robust agents that are doing a job. And you think about these specialized skill sets, like that sales agent example that I was told you.

16:33They are trained against that, and they can work with your team. Now, there are also other agents from other companies that are being built out and deployed, etc. Now, what you want is a place for all those things to come together. So whether it's agent force agents that are showing up inside of Slack, which we have, as well as third-party agents, whether it be coming from OpenAI or Anthropic, or I mentioned Perplexity, and the computer agent that they recently just deployed in Slack for Cell. There's another great company that deploys incredible agents inside of Slack. There's literally Slack is one of the fastest growing platforms right now on the planet for agent deployment.

17:05All those things can come inside of Live in Slack. And because we're an open platform, things like Slackbot can work equally well with all of them. And so that is really incredibly important for these enterprises as they go through and they think about, you know, obviously they're going to have a rich stack of enterprise software that they're looking to transform. And how do you bring all the elements into it? Every enterprise worth its salt is building, you know, an agentic layer at some point and they're adding agents inside of all these things. You were even talking about your video editing tool.

17:35So the question is, can you bring all that together in a thoughtful way that doesn't just multiply the problems that we had in the past, which is, hey, Mr. or Mrs. Employee, here are your 1 ,000 enterprise applications. Good luck figuring out how to use them. And that's required to get your job done, even though it's really not helping you get your job done. Yeah. Yeah. And you mentioned Perplexity's computer agentic system. Um, if, I mean, for example, I'm, I'm using OpenClaw. Sure. Yeah. Although I'm about to try Perplexity's computer, which I, from what I hear is, is similar, if not the same.

18:21Um, can you connect to things like OpenClaw or, or are you restricted to sort of approved, or is that up to the enterprise to manage? Yeah, so there's two things. We have a marketplace that has all the governance that you'd expect of a marketplace. So we take very seriously applications that are going to be deployed in Slack, that they have security, privacy policies, respect PII, et cetera. So we have a governed marketplace. So if you want to deploy an application inside of Slack that's available to third parties, it goes to the marketplace just the same way you would, you know, an application published in the Android store, the Apple store.

19:06The other side of that is in custom applications, which can be built inside of Slack. So an enterprise can go through and build and many, many, many do. We have over 100 ,000 AI applications that were built in Slack over this last year and a half. These are custom applications that organizations can go build and deploy and they stay inside of their organization. And those can be built and those don't have to go through our marketplace. So any organization can build on our API sets to go and build with our customer applications or then third parties, if you're like a perplexity, would go through our marketplace and then they can offer that application out to the totality of Slack users out there.

19:42Yeah. And one of the reasons why, just to kind of give you the and to that, one of the reasons why the explosion of these applications in Slack has been so pronounced recently is we've made it very simple, safe, and secure for organizations to tap into the kind of this goal that I was talking about earlier, these unstructured conversations. So we have a real-time search API. We have an MCP server that allows you in a safe, secure way, while respecting all the permissions I have, the organization has, to tap into those unstructured conversational data and bring that in. And that enriches these applications greatly.

20:18And so by doing so, all these organizations can not only just bring their application functionality in the stack they can then couple it with the conversational data inside of slack and you can really start to unlock some some pretty powerful scenarios yeah and that conversational data um if you're if you're adopting uh the the the new slack bot uh does it have access to uh conversational data that existed before uh the new slack bot was integrated i mean yes Yes. So, you know, the beauty of the cloud, all that data is there stored. It's your data as a customer. It's your permission data. You govern how long you keep your data, et cetera.

21:01So it's obviously predicated on the fact the customer would have that data. But if they have, then Slack bot just honors the permissions that I have. So in the same way that I have permissions to see public channels, I have permissions to see the private channels I've been part of. I have permissions, obviously, to see the direct messages or, you know, MPDMs as we call them, multi-person direct messages. Slackbot inherits those permissions, respects those permissions, and then can go through into the Wayback Machine and can go and see all that context. So the reason I was talking earlier about looking at presentations, maybe even from the founder of Slack, Stuart Butterfield, 10 years ago, and querying and looking at things that he might have done originally for customers.

21:45That was all shared in a public channel and I get to see that now as an employee. That's incredible. So yes is the answer. So you turn on Slack bot, it has access to that rich history from your organization. Yeah. Where do you think this is going? I mean, you've got an open ecosystem and you've got partners. You mentioned Perplexity, OpenAI, Anthropic, Google. Microsoft. Microsoft. Yeah. So is this, yeah, where do you see this going as these agents become more capable? uh it is is i mean do you imagine uh slack as becoming i mean you you're calling it the uh operating system for for the workforce you see it as uh becoming kind of the brains of the enterprise and employees are talking to the to the brain and and the brain is then going out and executing actions on on the part i mean the the reason i'm i'm asking is i've talked to so many companies that are building that orchestration layer or or building agent building layers uh and the mpc or mcp uh you know connective tissue uh and you because you have this installed base you have the advantage of doing that for everybody.

23:30And then all the user has to do is talk to Slack. Yeah. Let me give you maybe a little bit longer answer to your question, but hopefully I'll come all the way back around to it. Start more complex and hopefully end a little bit more simply. If you look at kind of the modern enterprise stack And certainly what Salesforce and kind of our Agent Force 360 platform is envisioning is there are growing amounts of systems of intelligence, right? These LLM providers are providing just intelligence and access to intelligence at unprecedented levels. It's incredible. But the first thing you need to do is you need to access that intelligence and harmonize it through your data and your metadata and bringing together all your various data systems.

24:16And we have a product called Data360 that really helps you harmonize and understand and bring all that data together to make it actionable. Then obviously every enterprise has kind of their systems of work. What are the applications where people are working? In Salesforce's case, we have obviously sales, service, ITSM, commerce, et cetera. Those systems of work applications, those are the units of function of effort that's happening across each of those functions in the line of business, et cetera. and are incredibly powerful. And then, as you talked about, everyone now has their system of agency.

24:52We have AgentForce as our area, which is the platform for building kind of agentic capabilities on top of those three foundations. All those three things are incredibly important and need to come together. And now when you talk to the employee level, which I would argue is one of the most important growth drivers in this AI era, and I can talk about that more and why that is. But when you talk about employee productivity and team productivity, what's required to get the benefits from the intelligence, harmonized data, all these systems of work, all these applications that your enterprise relies on, and now you're a gentile platform?

25:27Well, you need, ideally, one conversational interface, one place to go that brings together both the humans and the agents, as well as the data and the workflows that connects to all the various applications that you go through and doesn't make me think, doesn't make me learn a new way of working doesn't make me have to go through and get trained on anything. It just works. Like I should be able to talk to work day is easy. I'm talking to you. I should be able to talk to open AI is easy as I'm talking to you. I should be able to talk to my CRM as easy as I'm talking to you. And I shouldn't have to go through a training or I shouldn't have to ask my employees to do something different.

26:04And in the future where, you know, employees are interacting with agents as frequently, if not more frequently than they're interacting with humans, the paradigm shift needs to be, well, what was the platform that allowed for human collaboration and team collaboration that really transformed the world? Slack was fortunately a huge part of that. The same rules apply with agents. It's just a broader workforce. I should just be able to at mention an agent like I mentioned a human and bring them in and talk to them about the problem I'm working on and have them brainstorm and have them work on something and then come back and see the progress they've made.