Name: Ryan Gavin, Slack at Salesforce
Uploaded: 2026-09-25T17:15:00.000Z
Duration: 17 min 9 s

Clips
More from Dreamforce 2026

Ryan Gavin

EVP & CMO

Slack at Salesforce

play_circle_outline Slack: An Agentic Work Platform and Salesforce's Fastest-Growing Work Operating System

play_circle_outline Breaking App Silos: Slackbot Orchestrates Internal and External Agents via a Unified Conversational UI

play_circle_outline On-demand generated bespoke UIs from Slack/Slackbot for harmonization

play_circle_outline Governance: agents inherit user permissions for secure data access

14:52

15:19

17:10

Info

Ryan Gavin, Slack at Salesforce

Ryan Gavin

EVP & CMO Slack at Salesforce

George Gilbert

>> Palo Alto Studios, connecting Silicon Valley and Wall Street.

Ryan Gavin

>> I'm John Furrier, the host of theCUBE here with Gabe Altaba, my co-host.

Alison Kosik

>> Welcome to theCUBE, NYSE Wired program and community coverage of Dreamforce 2026. We're coming to you live from San Francisco. I'm Alison Kosik alongside George Gilbert. And we're about to talk Slack, one of my favorite topics, actually. And we're seeing with AI agents becoming part of the workforce that Slack kind of wants to become a place where people and AI actually get work done together.

George Gilbert

>> I think that was a big motivation behind what made Slack so strategic to Salesforce was that it is the work surface now where humans, digital workers, and the work all come together. And Ryan, I'm sure, will be able to tell us how that's going to unfold over the next couple of months.

Alison Kosik

>> All right. Let's introduce our next guest. It's Ryan Gavin.

George Gilbert

>> Hey.

Alison Kosik

>> He is the Executive Vice President and Chief Marketing Officer of Slack at Salesforce. Welcome. Thank you, guys.

Ryan Gavin

>> Thanks for having me. It's great to be here.

Alison Kosik

>> Great to have you. So, Slack started as messaging. I'm curious how you think about it today. What does the agentic work platform actually mean in practice and how are you growing it?

Ryan Gavin

>> Yeah, so, I'll start with a customer today like Box. They think about having 75% to 80% of all their time working in Slack. So it's way beyond collaboration now. We've kind of hit this renaissance period where, started as this beloved collaboration tool and now it's really turned into what we call kind of a work operating system. What do you need? You need to be able to obviously have all your people. Now you need to have all your agents, but then that has to be connected to all your enterprise data, all your enterprise systems. And then the real trick is how do you pull that all right into the flow of work so people aren't switching apps, going between tabs or different applications, learning different UIs. They can just talk to Slack, they can talk to our AI agent called Slackbot, and they can orchestrate all their work from one place. It's truly transformative and it's kind of what's taking Slack to be one of Salesforce's fastest growing products right now.

George Gilbert

>> So Ryan, maybe expand on that a bit. The notion that we— part of the value of agents is that they break down the silos, that we've been building for 60 years with the application silos. But it sounds like Slack's role is to give us a user interface that also sort of paves over the silos.

Ryan Gavin

>> Well, yeah, I'll build on that a little bit. The problem with silos is actually going to compound. So today you talk about an enterprise, they have, big enterprises have thousands of applications inside. Well, all those applications are about to have little agent babies. And how are employees supposed to navigate all this? how do you know which agent to go to, which has which skill? Where do I find it? It's actually going to make the din of the enterprise actually worse unless we do something differently. And that's where, Slack and Slackbot, we turn Slackbot into an MCP client. So that means it can talk to all your different agents. So you may have 200 agents deployed inside of an organization. Do your employees need to learn those agents? No, they don't. They just need to know, I go to Slackbot, Slackbot knows all the agents that are deployed and permissioned to me, and then Slackbot can pull in the right agent at the right time, at the right moment, with the right skills. Slackbot is now, it's not just a way to summarize the threads or act as your chief of staff, it's a super agent that can do everything that you can imagine, including things like build UI on demand. It's unbelievable. And it's one of the reasons why you're seeing companies like Meta, that was recently reported, is moving their entire company to Slack because you need a platform for agents now in this modern era. The agentic enterprise needs an agent platform. And the last thing you want to do is have your agents be someplace else that are different than where your humans are because it's all conversation that binds us together.

Alison Kosik

>> Can you give me an example of something I might ask an agent to do inside Slack, not just tell me, but actually do it for me?

Ryan Gavin

>> Sure. I'll tell you something I did this morning. I said, hey, I'm meeting with these 10 customers today. I want you to go through and look at all the account history that we have with them. What are the deals that are currently in place? Go through and build me a rich canvas that tells me what I should be focusing on. What are the things that are top of mind for these customers? Oh, and give me a couple anecdotes about the people I'm meeting with so I can actually kind of connect with them personally and then put that on a canvas that I can have right on my phone. And make it a skill that you can run.

Alison Kosik

>> Okay, but do you check it because you have that trust?

George Gilbert

>> Yeah.

Ryan Gavin

>> So here's the magical part because you hit the key, key point there. The trust comes from knowing that the data it's pulling from is a trusted system like Salesforce. And because it's connected into that Salesforce trusted system, it's pulling data from that system. And so I know I can trust it. The other reason I know I can trust it is because it knows me and it knows my context. So Slackbot can see all the conversations that I have permission to see inside of Slack. So it can see that, hey, the account team talked to this customer 2, 3 weeks ago and there was actually a pretty spicy topic. Now, I may not be actively monitoring that Slack channel. I have hundreds and hundreds of Slack channels, but Slackbot can. And so Slackbot's like, hey, be aware, there's a bit of a spicy topic on this renewal. Maybe, be careful of that. And so those are the things that engender trust, knowing that the underlying system of record is connected with whether it be Salesforce or it be Workday or Atlassian, all your enterprise systems there. And then knowing that Slackbot has the context of your business, those are the reasons why Slackbot's been the fastest adopted product in Salesforce history, it launched in January, 1.5 million monthly active users already. It's because of that trust is a big part of it.

George Gilbert

>> So there's something I want to dig into that you said that's really potentially profound because my first, where I first started getting really close to Salesforce and getting excited was when I saw Data Cloud. 3 years ago because I'm like, okay, finally we have someone who can have a semantically harmonized platform for the infrastructure. But what you said Slack might come at this from the other point of view, which is I have agents from all the different applications which are still in their own silos. And Slack might, the Slackbot might know how to talk to all those agents, but then you said generate a UI. Yeah. So I want to kind of build on that, which is, we at one point had to have user interfaces that came with the applications.

Ryan Gavin

>> Yeah.

George Gilbert

>> Now it sounds like we can talk to the silos and then have a bespoke UI generated for us that bridges the application.

Ryan Gavin

>> Yeah, that's right. Yeah. In Marc's keynote, he talked about this thing called AI Force. And in Slack, we have Slackforce, which The simplest way to think about it is if you have, Slackbot is connected to the structured and unstructured data inside of your enterprise, all your enterprise systems, Salesforce included, and that unstructured conversational data. Now, to really have Slack be the front door to your enterprise, it needs to be able to do a lot of different things. 8:00 AM, I need to prep for a meeting. At 9:00 AM, I need to look at what's happening with my pipeline. at 10:00 AM, I need to see what's going on with the latest features and bugs. All those things have historically required different UIs, different products, visualization. But now you can just use, everyone can code because you can speak English, which is kind of crazy. I can just talk to Slackbot and I can say, hey, build me a dashboard based on our Slack analytics. I want to see the latest adoption of Slackbot. In fact, give me a breakdown of how it's progressed over time. And then I actually really want to see a prioritization of the utterances people are saying into Slackbot. What are the things that people are asking most for? And build that into a rich visualization for me. Now Slackbot will go and generate an on-demand UI that is dynamic. It's shareable. I can share it with my team. It's multiplayer. So my team can get in there and ask questions on bits of the data that they're seeing. And suddenly we have this bespoke UI for our enterprise that took no time to build, took no expertise to create.

George Gilbert

>> But it's harmonizing the mess of the islands and silos. That eventually we'll re-engineer and most of them and we'll integrate them, but it'll take many years. But in the meantime, we can harmonize it at the UI level.

Ryan Gavin

>> Yeah, and it's pretty powerful. The way I think about it, you are democratizing access to some of these systems that have really represented trapped gold for many organizations. I'll take Salesforce as an example. That's a very good one. Companies spend lots of money with Salesforce. They built up this customer truth, this context. But how many people inside of these companies access that? Well, if you're in sales, you do. Maybe if you're in service, you do. Yeah, that's probably it. I've been in technology for a long time. I've got all the gray hairs to prove it. I looked at Salesforce very little before I got here. I saw Salesforce data very little. I never logged into Salesforce. Today I access Salesforce data hundreds of times a week. Because it's now been democratized to me. My team accesses the Salesforce data hundreds of times a week when before they didn't, because I can simply ask Slack, I can ask Slackbot, and all that data is pulled in a trusted way. And that is very powerful because these enterprise systems represent a lot of trapped gold that is now being democratized through AI and through things like Slack.

George Gilbert

>> Okay, so I have to ask a pricing question. Sure. Which is that now, people in the enterprise who didn't or wouldn't have access to Salesforce before, are they going to be full-fledged Salesforce license customers? Or because I know there's different pricing models, but my understanding was it's still sort of a user entitlement, and then there's some credits, there's a consumption component.

Ryan Gavin

>> Yeah, so the good news is we've got all sorts of options for customers, so there's not a one-size-fits-all. And when you can sit down and you say, hey, I have a whole host of customers that, or employees that I want to have be able to read into Salesforce data and be able to use that on things. Like we have ways of licensing that, that's not at full freight as you described. Okay. The other thing you're mentioning is there's consumption-based licensing with things like Agentforce. One of the reasons why Slackbot, I love that you brought up pricing, has been one of the most quickly adopted products in Salesforce history is because most AI today, a lot of the concern is the cost. You know, you're bringing in, you got these rich tools. We all heard about token maxing, etc. Slackbot is one fixed cost for your employee, unlimited usage. So what enterprises get to say, wait, I can give this to my employees. It's consumer-grade kind of AI, meaning it's ready, it works out of the box, and I don't have to worry about costs. I'm paying one fixed cost and they can use it unlimitedly. And that's what we've found to be a very powerful model that's eased the anxiety of cost when it comes to utilization of things like Slack and Slackbot.

Alison Kosik

>> Switching it up a bit, as agents take on more work inside Slack, how is Slack thinking about governance, permissions, and giving companies control over how agents operate?

Ryan Gavin

>> Yeah, it's such a good question and it's certainly front of mind. One of the things that we've figured out inside of Slack that nicely translates into agents is I have a set of permissions that have been set up for me inside of Slack. Okay? So when I use Slackbot, Slackbot just inherits those permissions. So now I have an AI tool that only can see the things that I can see, only can access the content I can access. And so my permissions just extend to my agent in this case. And that's true for how we've set it up. And again, it helps really allow administrators to feel confident that, okay, if I've set up the permissions correctly for you, I know your agent is going to follow those permissions correctly. And now Slack's also an open platform. So we have all these third parties like Claude and Claude Tag that deploy. And, the way they think about it is they deploy Claude to act inside of Slack. They just treat it like another employee. So I'm going to set up what are the permissions that this agent can have, just like the same permissions that might be set for a human. And then the agent can go do its work. So the individual and human-based permission model extends pretty nicely into agents.

George Gilbert

>> All right. Yeah, I know we don't want to drop too much in the weeds, but I'm just dying to understand how Slackbot can talk to these other systems. it's permissioned, I assume, maybe through something like MCP.

Ryan Gavin

>> Yeah.

George Gilbert

>> And then it— okay, so then, but it's got to understand that, this data from this system may not mean the same thing as data from this other system. So when it puts the UI together, how does it know to, do that?

Ryan Gavin

>> Well, there's two things there. So first, you're exactly right. So these MCP servers and MCP client is the interconnectivity between these agent-to-agent conversations. What Slackbot can go do is it can go and learn the agent capabilities. So when you're deploying an agent inside of Slack, you can connect it with Slackbot. And one of the first things it's going to do is like, all right, what is the dataset that you have? What are the capabilities that you have? What are the permissions? What are you specialized in? And so just like you onboard an individual, you're going to learn all the capabilities in the underlying dataset. So Slackbot will then learn and retain that. And then when Slackbot's invoked, one of the things that's really powerful, it's not working in a black box. So when I ask Slackbot a question, I like to say it shows its work. So Slackbot then says, hey, I'm querying Salesforce right now. I'm querying Atlassian, and it shows where, and then you can stop and drill into those. So you can always audit what's coming out on the backend. And so kind of showing your work and not being a black box is a critical element. Again, back to your question about trust.

Alison Kosik

>> All right. So a year from now, what's the biggest change you expect to see in how people and agents work together?

Ryan Gavin

>> A year from now. A year from now. My goodness.

Alison Kosik

>> Sorry, how about 6 months?

George Gilbert

>> Yeah, it's 7 dog

Ryan Gavin

>> years.I think we're going back to this idea of composable UI. It's a really— it's a mind bender if you start to experience it because suddenly there's nothing out of bounds. And as you know, I mentioned earlier this idea that everyone can code and everyone's a builder. Because we can all speak language and we can all talk and converse. And that is the new code. And we can code in the flow. And to your point, one of the things that's been remarkable that's happened literally in the last 6 months is with coding agents, coding is no longer a bottleneck. Product backlogs are done in like a quarter. It's incredible. And so the new backlog is things like taste, judgment, customer insights. Are we building the right things? And so we launched actually a product called Claude Code that brings together all your individuals, your design team, your research team, your marketers, and your product and engineering team so you can build collaboratively. So we can pull up all those things that have historically been pushed down in the development timeline and you can do it real time in multiplayer. So it's like multiplayer building. And I think what you're going to see in 6 months is everyone in your organization is going to think of themselves as a builder. They're all going to be writing code. And they're all going to be in there creating new content products, capabilities as a team. And it's pretty powerful.

Alison Kosik

>> So I know you have one more. Yeah, I can feel it.

Ryan Gavin

>> You can feel it. Yeah, I can feel the heat coming. Yeah.

George Gilbert

>> Because when I first started playing with Claude Code, I'm like, this is the biggest change in personal productivity tools, in 35 years since we had a GUI.

Ryan Gavin

>> Yeah.

George Gilbert

>> And what I'm hearing from you is like we're going to be building instead of building like spreadsheets, word processors, and slides, we're going to be building interactive tools that we share with coworkers and potentially customers. And that this is the new— it's a— it's not just a way of— it's not just a dashboard. It's a personal productivity toolset for, Retrieving, analyzing, synthesizing, presenting, sharing.

Ryan Gavin

>> Yeah, I literally just got off stage with Boris Cherny, the creator of Claude Code, and he was showing how he works. And there's going to be products, there's going to be things that we use, there's going to be UIs. It's not that all that's going away, but the work of work is going away. The, hey, go find this new document, go figure out how to use this new product, go learn this new tool. That is going away. And what you're going to be able to do is just converse with things like Slack and Slackbot. Slackbot's going to know how to work with those tools and products. So then you can actually focus on the things that add value to customers, the things that add value to your business. And honestly, what I've seen is this renaissance with people loving what they do because they're spending less time doing all this nonsense work and they're spending more time actually adding value and feeling fulfilled in what they go do. So it's a pretty magical period right now.

Alison Kosik

>> Ryan, thanks so much for stopping by theCUBE. Great conversation. Thanks.

Ryan Gavin

>> Thank you guys so much. Really appreciate it.

Alison Kosik

>> And you're watching theCUBE, NYSE Wired program and community coverage of Dreamforce 2026. We'll be right back.

Ryan Gavin, Slack at Salesforce

George Gilbert

>> Palo Alto Studios, connecting Silicon Valley and Wall Street.

Ryan Gavin

>> I'm John Furrier, the host of theCUBE here with Gabe Altaba, my co-host.

Alison Kosik

George Gilbert

Alison Kosik

>> All right. Let's introduce our next guest. It's Ryan Gavin.

George Gilbert

>> Hey.

Alison Kosik

>> He is the Executive Vice President and Chief Marketing Officer of Slack at Salesforce. Welcome. Thank you, guys.

Ryan Gavin

>> Thanks for having me. It's great to be here.

Alison Kosik

>> Great to have you. So, Slack started as messaging. I'm curious how you think about it today. What does the agentic work platform actually mean in practice and how are you growing it?

Ryan Gavin

George Gilbert

Ryan Gavin

Alison Kosik

>> Can you give me an example of something I might ask an agent to do inside Slack, not just tell me, but actually do it for me?

Ryan Gavin

Alison Kosik

>> Okay, but do you check it because you have that trust?

George Gilbert

>> Yeah.

Ryan Gavin

George Gilbert

Ryan Gavin

>> Yeah.

George Gilbert

>> Now it sounds like we can talk to the silos and then have a bespoke UI generated for us that bridges the application.

Ryan Gavin

George Gilbert

Ryan Gavin

George Gilbert

Ryan Gavin

Alison Kosik

>> Switching it up a bit, as agents take on more work inside Slack, how is Slack thinking about governance, permissions, and giving companies control over how agents operate?

Ryan Gavin

George Gilbert

Ryan Gavin

>> Yeah.

George Gilbert

Ryan Gavin

Alison Kosik

>> All right. So a year from now, what's the biggest change you expect to see in how people and agents work together?

Ryan Gavin

>> A year from now. A year from now. My goodness.

Alison Kosik

>> Sorry, how about 6 months?

George Gilbert

>> Yeah, it's 7 dog

Ryan Gavin

Alison Kosik

>> So I know you have one more. Yeah, I can feel it.

Ryan Gavin

>> You can feel it. Yeah, I can feel the heat coming. Yeah.

George Gilbert

>> Because when I first started playing with Claude Code, I'm like, this is the biggest change in personal productivity tools, in 35 years since we had a GUI.

Ryan Gavin

>> Yeah.

George Gilbert

Ryan Gavin

Alison Kosik

>> Ryan, thanks so much for stopping by theCUBE. Great conversation. Thanks.

Ryan Gavin

>> Thank you guys so much. Really appreciate it.

Alison Kosik

>> And you're watching theCUBE, NYSE Wired program and community coverage of Dreamforce 2026. We'll be right back.