Slack Code Makes AI Coding a Shared Team Workflow (Full Transcript)

[00:00:00] Speaker 1: Slack is the OS of so many companies, most cutting-edge teams are hooked on it.

[00:00:09] Speaker 2: No matter if you're a developer, a designer, or just someone with an idea, Slack is a new evolution of multiplayer vibe coding.

[00:00:24] Speaker 3: We don't have to build a new interface here, there's already one, it's just called Slack.

[00:00:28] Speaker 4: There's no way we could have built everything we've built and grown the way we have without Slack.

[00:00:52] Speaker 5: Welcome and thank you so much for joining us. I'm Jillian Bruce, and today you're going to learn about what happens when AI becomes part of the team. We cannot wait to show you Slack code, a brand new way for humans and agents to work and build together all in Slack. Joining me is Slack EVP and GM, Rob Seaman, and Chief Product Officer, Jamie DeLange. Welcome guys. Thanks.

[00:01:16] Speaker 6: Thanks for having us.

[00:01:16] Speaker 5: We have a lot of great things to talk about today. So Rob, I wanted to go to you first. It's really wonderful to have you here because you have your pulse on what's happening out there in the industry. Everyone's trying to solve the AI puzzle right now. Can you give us a little bit more insight into what's really happening for enterprises across the board?

[00:01:36] Speaker 6: Absolutely. If you look around, we're in just an incredible moment. The pace of AI development over the last 18 months to 24 months is really faster than anything I've seen in my career. But if you go to our customers and you ask a room full of executives what measurable productivity impact they've actually gained for their employees from all the AI spend that's happened, the room generally goes quiet. And it's the case that the models aren't the problem. The problem is that there's a few key missing pieces to every company's AI stack. And without those, we don't think it's possible for true multiplayer work and therefore value to be created to happen. So companies are spending millions of dollars building out every layer of the AI stack. If you go to a customer, it typically looks something like this. They've got a relationship with the foundation models. They've typically got a data lake above that. They've built some MCP tools around all the apps that they have. They've got some approach to guardrails and the apps themselves. But they're missing a few key things, the bookends of the stack, if you will. First of all, those tools are disconnected from where the conversations and decisions get made and problems are getting solved. This is a company's alpha. This is their projects, their topics, their teams, their work items, their ideas, the things they choose to pursue and why, and the things they choose to pursue, not pursue, and why. But they're also missing a runtime where employees can use these AI agents together. And those two things are where we think Slack comes in. Importantly, those two things, when used together, are ultimately going to compound context. So if you take these AI tools out of what we call single-player mode, where it's just an individual in a terminal or a SaaS chat app, and put it into a multiplayer runtime, the exhaust of those interactions becomes additional context that's available to your agents and is available to the rest of your employees as knowledge artifacts. And that creates a compounding flywheel of value that ultimately results in differentiation for the company that's using it. So Slack ultimately becomes that multiplayer runtime. It grounds all the agents in institutional memory, identity, the relationships of your organization. And when you bring those things into the fold, again, that flywheel of value emerges and produces even more and better context. And so instead of individual employees gaining productivity, the entire company is gaining productivity at that point. It's your runtime. It's your context. It's your AI work operating system. It brings all of your people together, all of your agents at every platform you run on all in one multiplayer interface.

[00:04:13] Speaker 5: That's incredible. And multiplayer is what it's all about because we're all working with agents all the time. And it's becoming more of the norm, especially for builders. And coding agents aren't just coming, they are already here. We've already even seen some snippets in the opening video. Jamie, can you talk a little bit about how that's factoring into what Slack is building?

[00:04:32] Speaker 7: Absolutely. So, you know, the most innovative companies are already building in Slack, and they're already working with coding agents in Slack. We know that they're doing this because that's where their teams were already working. So they were already working on a developing code in Slack. So when the agents joined the conversation, it just made sense to add them into the mix. But a lot of companies that aren't those innovative, bleeding edge AI companies are really struggling to figure out how to manage agent workflows, how to help keep their development teams connected. And they're losing a lot of corporate context to isolated work. So that's the challenge that we wanted to tackle. We wanted to make agentic coding just natural for anybody, bringing it out of isolated environments and into the open so that everyone can work together. There's no new developer tools to learn. There's no single player terminal. There's just a place where cross-functional teams, developers and PMs, designers, and of course, coding agents can all work together in Slack. So that's why we're launching Slack Code. We built Code Channels as a dedicated workspace for designers, agents, humans, engineers, PMs, all to work together to build software in the open. The team together can review code, they can run live previews, they can ship software. With the introduction of coding agents, I feel like it's made it so easy for anybody to develop. Coding has literally become a conversation. And if you have an idea, you should just be able to turn that into something. So Code Channels make that possible for everyone. For you guys out there, there's five things that you need to know about Slack Code. First of all, this is kicked off by agents. The agent can spin up a code channel to tackle whatever task you've assigned it. They're specialized for building with agents. The agents can create tabs inside of the space so that everyone in the channel can see the plan, they can see the code diff, they can see HTML artifacts. All of that happening on top of the conversation. They're project-based, so when the task is done, the channel is going to get cleaned up automatically. Everything is organized, focused, it's tidy. Don't have to worry about it cluttering up your sidebar. The fact that this all is happening in Slack also means that when the channel is archived, the content is going to remain searchable. It means that exhaust that we were talking about before, all of that's going to be available to your team and to the agents that are working in Slack. They also take full advantage of Slack's flexible platform, which means that every company can configure the agents to work in a way that makes sense for them. And since it's all a part of the platform, it's also going to inherit Slack's layers of trust, enterprise security, governance. All of that's going to apply to the channels, the work, everything that's happening inside of the code channel.

[00:07:21] Speaker 5: Wow. It really changes the game on how we're building, and it's a whole new way, actually, if you think about it, of building and coding, and it's all in Slack, available to everyone. I mean, this is, it's truly amazing. It's amazing. Now, I know we're not the only ones that are excited about it. Let's hear what a few of our partners have to say.

[00:07:40] Speaker 8: I'm really excited about the Slack launch code channels.

[00:07:43] Speaker 9: Slack is already where we do so much of our work. Code channels are taking this to the next level.

[00:07:48] Speaker 10: We've been using code channels and we love them. You can connect it to whichever tools and data and code bases you're already using.

[00:07:54] Speaker 1: Anyone can join, catch up on the history, understand why decisions were made, and keep contributing. Everyone can collaborate and drive the solutions.

[00:08:02] Speaker 10: Viewing code diffs directly within the channel.

[00:08:04] Speaker 8: Being able to just have screenshots of the experience we're talking about, and I can tag marketing, and they can come in and give feedback, and it just goes direct to the agent.

[00:08:11] Speaker 1: What gets especially interesting is that these channels can get better at their job over time.

[00:08:16] Speaker 9: Discussions and planning, shaping, it all happens in real time in Slack channels. That's a game changer.

[00:08:21] Speaker 10: We're super excited about this because it allows you to bring the power of Cloud directly into a Slack channel as a member of your team.

[00:08:29] Speaker 8: We've been using it to build our integration for GitHub in Slack, which is so cool and so meta.

[00:08:35] Speaker 9: It's going to be fantastic, and I can't wait to see what everyone builds.

[00:08:38] Speaker 8: Really excited for us to continue to use code channels, but also for our customers. We're really excited for you to try them out with Cloud Tag.

[00:08:45] Speaker 11: It's been really fun working with Slack to bring code channels to life. Can't wait to see what's next.

[00:08:56] Speaker 5: All right, we're going to switch it up a little bit. We'll see Rob again later, but joining us now, we have VP of Product Management, Katie Stegman, and VP of Product Marketing, Satya Parlakar. Welcome, ladies. Thank you. Now, Katie, you have been working on this day in and day out. Tell us a little bit about why Slack code is so different from a regular workflow or channel.

[00:09:16] Speaker 12: We are so excited about Slack code. With code channels, we've created a dedicated surface for humans and agents to work together, so code channels democratize building. Anyone can build with code channels, not just engineers. And they're multiplayer, so you can bring the whole team in, designers, marketers, product managers, everybody in one place. And this is all backed by Slack's open ecosystem, so you can bring your own models and you can bring your own agents. So we just showed you guys some of the highlights of our partners who are bringing their coding agents into Slack, but you can also build your own. So we're going to do a demo with our internal coding agent that we call Spec, which is customized for our codebase. Satya, should we demo?

[00:09:54] Speaker 13: Let's do it. Awesome. All right, so this is really timely because we were just talking about how our customers love our seasonal themes last year that we did. So we want to explore doing that for Halloween this year. And this is one of those fun, surprise, and delight features for our users. So in this project channel here, you can see that we're already talking about it, and all I have to do is tag Spec, our coding agent, to kick it off. Now, the coding agent is automatically creating a code channel and bringing all of the context into this code channel here that you see on the screen. And it includes all the files we shared, all the background, all the context that we're already talking about. And it's already started to bring all of the right people in as well, including Katie.

[00:10:39] Speaker 12: Yeah, so Satya, you spin up the channel, but I'm in here too, so I can see everything going on. So let's take a closer look at what Spec has made. I see we have the color themes in here that I can check out. I kind of like the pumpkin patch one. But I'm going to tag you and Satya to see what you think about the color scheme so we can all talk about it together.

[00:10:58] Speaker 13: Awesome. So now, it's hard to tell from the colors on this palette how it's going to look in the product. So I can just ask the coding agent to mock it up with the Slack UI. Now, what you see here on the right is all of the different color palettes. And now, the coding agent is taking those and actually taking our design system and bringing in the right color palettes it just created. So I can preview them right here. And if I want to make a change to this, I can do it on the fly by just tagging Spec. So here, I'm actually not a big fan of this green and orange situation over here. I want a little bit more orange for fall, so I just tagged Spec. And it's actually updating this mock-up in real time. There, I have a nice new orange for fall.

[00:11:43] Speaker 12: I love that. I love that. So what do we think? Should we ship it? Let's do it. Okay. So I'm going to ask Spec to spin up a PR. And Spec will create a pull request for this update. But what I'm also going to do is tag in Chris. Chris is an engineer on our team. He's the engineering lead. And he's going to make sure this is a reasonable technical approach. And he'll act as a little bit of a technical checkpoint for us. So Chris is checking it out. You can see that Chris actually has the ability to talk to the agent, ask the agent to make updates to the code. He's doing that right now. It looks like we're going to get some updates. But once all of that looks good and Chris is happy with it, looks like he is, we actually have a link to our testing environment. And we can open that up and take a look at what we've built right now.

[00:12:26] Speaker 13: All right. So here is the working dev environment in my browser. Let's test it out. Trick or treat. All right. I like it. Here we see that pumpkin patch theme all mocked up.

[00:12:37] Speaker 12: Katie, what do you think? Should we ship it? It's awesome. I think we should do it. If we wanted to ship this right now, I would just go into GitHub and I would merge this. We will not do that right at this moment. But that's what we would do if we wanted to proceed.

[00:12:49] Speaker 13: Awesome. So this is how we're building this fun Halloween treat for our users. Now this is a fun, silly example. But it's actually how our teams have been able to work together to build cross-functionally all these new features even faster. Yeah, exactly.

[00:13:02] Speaker 12: We actually built Slack code with code channels. We are really excited about how much this is increasing our engineering velocity. But what is especially cool is how non-technical people like me and Satisha can have an idea, build with an agent, bring our teammates in, and ship code.

[00:13:17] Speaker 5: Well, I'll debate that you aren't technical people because you're here talking about Slack code. But that was an amazing demo and a wonderful way to really see this in action. Now what's your biggest takeaway, Jamie, seeing that in action right here?

[00:13:30] Speaker 7: I think that there's just so much that Slack code can do now. And it's going to be so amazing to see how our ecosystem and our customers use this in the future. I mean, look, Satisha's in marketing, Katie's in product, and the output of what they built together is stronger because they could collaborate on it. For me, I think software engineering is really just the beginning of where Slack code starts. It's not where it ends. There are so many departments that are having very similar problems. And you can move that agent interaction, whether it's working on a prototype, working on a document, move it out of those private chats into this unified multiplayer experience. That exact same multiplayer pattern can extend across the entire organization. Just imagine a marketing team using a dedicated channel with a custom agent to draft, review, and ship maybe even something like this, right? With coding agents, there should be no reason that a total normie can't tag up with a coding agent to ship.

[00:14:32] Speaker 5: I agree. Again, for normies in the room, I'm very excited about this. But I want to talk a little bit more about these coding agents and our partners that have built them with us. Jamie, can you tell us a little bit more?

[00:14:43] Speaker 7: Yeah. So we chose to build Slack code not just for us, but for our partners. A lot of our partners were already building agents doing these kinds of coding behaviors inside of Slack. And we wanted to incorporate their feedback and make sure that we were accounting for their needs as we were building. Because we wanted to make this pattern work for everybody and truly work for everyone, not just for ourselves. So Katie actually had the chance to catch up with three of our partners to not only talk about their experience, but for them to show us what they're building in action. So first up is GitHub. You know GitHub. They're the folks that built the home that basically everybody lives in if you're a developer. And Katie had a chance to talk with Mario, their CPO, about that experience with developers and how they thought about working in and with co-channels. So let's roll the video.

[00:15:33] Speaker 12: Mario, thank you so much for being here and talking with me today. I am Katie Steigman. I'm a VP of product here at Slack. And why don't we start by having you tell us a little bit about what you do at GitHub?

[00:15:45] Speaker 14: Yes. Thank you for having me. My name is Mario Rodriguez, and I'm GitHub's chief product officer.

[00:15:51] Speaker 12: So just to start us off, why don't you tell us a little bit about, you know, why GitHub chose Slack as this dedicated environment for your agents?

[00:16:01] Speaker 1: Yeah.

[00:16:01] Speaker 14: I mean, if you think about it, a lot of software work already begins in Slack. You know, someone reports a problem, my team is discussing it, and then eventually we kind of put that and translate it into a conversation that ends up in a pull request, right? And that translation at times has a lot of friction, and you use a lot of context because of that. And we could remove a lot of that friction and actually bring the conversation where people are having it, and they're already collaborating, and Slack is the perfect place for that. So bringing the full power of Copilot into that conversation is nirvana, it's actually what we always wanted to do. The interesting part is that everyone benefits. Like I get to see what you are asking Copilot, and I'm like, oh, my God, that worked really well. And then I could then try it, and then we could just build on that. So even the same tasks that we're doing over and over, we're getting better and better at it. So that's a very interesting byproduct of all of this. So the multiplayer without context losing, and then continuing to raise the ceiling of what we could actually achieve is one of the primary reasons why we chose it.

[00:17:11] Speaker 12: Maybe you can tell us a little bit more about your vision for multiplayer collaboration between humans and agents.

[00:17:18] Speaker 14: We always say, you know, GitHub's mission is about advancing human progress through developer collaboration. And that advancement of human progress is through software, and that collaboration is across the surfaces, and Slack is at the center of that, as you know. One of the things that we're very excited about is within the new capabilities that we're adding is that there's a diff view now in Slack. That doesn't mean that Slack is becoming, as an example, an IDE or even an ADE, but it is bringing context to where it's needed, and it's keeping you in the flow. Where you're most creative, that is when you have the most flow as well. So for me, it's about keeping developers in the flow, keeping the team in the flow, and then continuously then bringing not only new entrants into that creation, but continue to up-level the craft of what is being created, too.

[00:18:11] Speaker 12: I am right there with you. I'm really excited to kind of dive into some of those parts. I know we're thinking a lot about the same things. How are you bringing non-technical users into this? How are you making this feel really accessible, as you say?

[00:18:22] Speaker 14: We usually talk about lowering the floor and then raising the ceiling because of that. Like, lowering the floor is, how can we invite more and more people into that creation? And then we made some of these people into masters and experts, and then they are the ones that push in the innovation, both invention and innovation forward. So we need to continue to do that. And now AI just gives us such a platform to take that to 100x leverage, more or less.

[00:18:50] Speaker 12: I'm curious what trends you guys are seeing in non-technical users working in GitHub, sort of on that note.

[00:18:56] Speaker 14: Professional developers now will continue to have more and more leverage because they understand the fullness of the stack. At the same time, what we're seeing is now everyone can code. So the reality is we can easily now, not through skilling, but just through the accessibility of these coding agents and through Copilot, can create 1 billion developers on Earth. That's around 10% of the world population, which would be amazing to actually have. In Slack now, with the creation of these panels that you could bring in, I think we're starting to see the same trends, which is you want to interact in natural language. You want to make creation possible really, really fast. You get, at the end, the outcome that you want.

[00:19:40] Speaker 12: How do code channels and your expanded API capabilities, how do those lay the groundwork for the larger vision that you guys are driving towards? Tell me what you'd love to see or what you're thinking about.

[00:19:52] Speaker 14: All right, let me. So yes, I think the best way for me, like if I played the future out, right? Like code is just one of the artifacts overall. The app itself is another one. So if you think about it, well, the code will translate itself into the app. So for me, it's like code might be the low level. We will start going more upwards and we'll start actually even maybe thinking more about the customer that requested that code and maybe inviting them into a huddle and then kind of going in and trying to get real time feedback on it. So I think we just are the very, very beginning of what collaboration could look like within a channel, a code channel, a customer channel, all together to create, at the end, the application that we want for that consumer.

[00:20:37] Speaker 12: Mario, thank you so much for taking the time to speak with me today. It was an awesome conversation, and I am so excited for what is to come with our collaboration.

[00:20:45] Speaker 14: Thank you so much for having me. Like you said, we're just at the very beginning of this, so we got a lot more to do together.

[00:20:52] Speaker 12: All right. We've been talking a lot about what we've been building, but Mario, I'd love to see in action what you guys are working on.

[00:20:58] Speaker 14: Indeed. I would say demos, not memos. So let's actually, let's dig in. Here's an early look at our new GitHub integration for Slack. And the part I'm most excited about is how much it keeps you in the flow and how it turns GitHub into a multiplayer experience. I've heard from customers that our fitness app shows their waitlist position as zero. That's confusing because it makes them think that they're already bugged. And I report this issue in our Slack channel, and Megan asked GitHub to investigate. GitHub creates a dedicated code channel, connects it to the repository, and starts tracing the waitlist calculation from the API response through the customer facing experience. Now you can see here how it's doing that. Now the cost is not confirmed, both the server and offline fallback intentionally use zero base array. Then it's just thinking through it, which is great. I could go into a code view and look at the depth of the fix that it's doing. What we also want to do, you can see Megan here saying, hey, can you render the app for us to actually check it out? So it's not only working on the fix, they could go in, take a screenshot, and then render the final outcome of that fix as well. So I could see here, perfect, now I'm showing the class and only three spots left. That's exactly what we want. Now I love this, but what I really want to do is also in the list, I want to say how many people actually have joined it, just in case there's like a hundred of them. We don't want the user to think they could get in the class overall. So I'm telling Cobilo to do that. Cobilo picks up exactly from where it left off in that branch, and it's going to start implementing that. You can see here that it created a PR for it. Now Megan picks that up and merges it.

[00:22:54] Speaker 5: GitHub, I know, is loved by developers all over, and what an incredible demonstration of co-channels. Jamie, after watching that, what are your thoughts?

[00:23:01] Speaker 7: Well, first of all, I just love seeing Mario's passion. He's clearly thinking about co-channels in the same way that we are. I also love that they're thinking about co-channels as a natural extension of the conversation. So whether you're working with the development team or even customers, thinking about how you can sort of bring everyone into that development story. Finally, I just really love how he talked about lowering the floor and raising the ceiling. I think there's so much that we can do with democratization with co-channels, but they can also be an incredibly powerful tool for seasoned developers doing critical work like incident management.

[00:23:39] Speaker 5: I mean, everyone has to deal with that, right? And thinking about new ways of solving that that reduce the friction is incredible. So we have a lot more other partners that we also want to feature here. And the next partner up is Cognition. And I think Katie talked to Jeff. Yeah. Yeah.

[00:23:53] Speaker 7: So if you're not familiar, Cognition is the team behind Devon, which is the AI agent that's shaking up how software gets written. So let's just see their take on co-channels.

[00:24:02] Speaker 1: Let's do it.

[00:24:03] Speaker 12: Hey, everybody. Katie from the Slack team here. Jeff, I'm so excited to talk with you. Tell me a little bit more about yourself, and we'll jump into talking about Devon and Cognition.

[00:24:14] Speaker 3: Thanks for having me. I'm the president of new enterprise at Cognition. What that just means is there's a lot of new things going on across the world and in the products. And I'm happy to just jump in and try to help out in all those spaces.

[00:24:26] Speaker 12: Just to start us off, what made you choose Slack as the main environment for Devon?

[00:24:31] Speaker 3: We've always been in Slack. I think we might've been one of the first agents that were available. And it's a place where all the ideas are forming. It's where a lot of conversations are happening. There's a lot of times where someone complains about something or identifies a bug. And that is also the ideal way to set off a Devon, to go and fix a bug, to create or add a new feature. And then people can jump in to keep making sure the idea is the right one, or maybe help debug.

[00:24:58] Speaker 12: I know we have that experience as builders too. How does pairing Slack and Devon together advance your vision for multiplayer collaboration between humans and agents?

[00:25:06] Speaker 3: Devon was always built as kind of a coworker. You have a lot of work you want to give to this agent. The agent comes back to you and the work is done. Actually, Devon has a lot of tools to prove that the work is correct. It might send you a screen recording or give you a report, but at the end of the day, it comes back with a PR. So it kind of finishes the work and gives it back to the humans to approve. So now this new experience where we are able to create a code channel and have an area to track all the sessions and see the different changes in the code and even maybe even see a demo of the changes right in that session is actually amazing. It's actually bringing a lot of what makes Devon great into Slack. And we hope that everyone else is as excited as we are.

[00:25:52] Speaker 12: I know both of us are thinking a lot about single player versus multiplayer collaboration with agents. What changes when Devon's work happens in a shared channel versus having it in sort of a private session, either for the engineer or for the rest of the team?

[00:26:06] Speaker 3: Yeah, when it's in Slack, it's way more visible. If you have a Devon session, it can be obscured in the web app or API or anywhere else Devon is running. When it is in Slack, people can jump in to collaborate. You can pull people when necessary, when you want to hear their feedback on how to collaborate on a specific session. And actually Devon is proactive in creating these sessions as well.

[00:26:29] Speaker 12: So you can just be sitting there talking about a bug or a problem and Devon could pop in and say, actually, we can do something about that. Hey, let's fix this, that type of thing.

[00:26:37] Speaker 3: That's right. And there's more advanced things you can do in the future, I hope, where it's pulling product owners or people that have run into the issue before, or people that have seen the bug multiple times from other parts of the org, and then collaborating together to fix it. And then obviously at the end of the conversation, the fix is done. You give this work to an agent, comes back with the fix, and the idea to production should go down over time.

[00:27:00] Speaker 12: What are you seeing about how people's roles are changing when it comes to these sort of shared channels and engineers and designers and non-technical people also being able to build?

[00:27:10] Speaker 3: Yeah, I mean, we're going to see a lot of non-technical people push a lot of code into production. And this is a good thing, by the way. We have, for example, at Cognition, some of our sales reps identified bugs reported by customers and then in the reply, this is something we'll change with sessions, there will be a fix immediately, right? So getting the entire company mobilized to work with agents and getting things to production quicker is what we're trying to do here. I mean, the future is already here with a lot of startups we work with, and I work with a lot of CIOs to really think about how to transform their organization into an AI native organization. And really, it is getting the ideas into production as fast as possible. This really shortcuts the traditional ways of working in an IDE or CLI or any other surface. You can have things that just start in your conversation and then just be out there in the public, right? And that is a very different mindset and almost feels like magic in some cases.

[00:28:11] Speaker 12: Okay, we have been talking a lot about Devon, but why don't you show us how it actually works?

[00:28:17] Speaker 3: All right, let me get this ready and let's pull this up. So here we have one of our engineering channels where Devon is watching. This means that when someone drops a message that Devon could help with, we don't even need to tag it. It will just jump right on the task. One thing we focus a lot on is Devon's Slack etiquette. Starts by replying in the thread, and unlike some other agents, it doesn't pollute your notifications by responding to every message. Other team members can chime in and Devon will take into account their input as it's working. I can see that Devon created a ticket to track this issue. And when it's ready, it drops a PR and a video showing the fix. Devon even tagged the correct code owner for review. Devon has its own machine in the cloud, so it can run the app and test it in Chrome. Here we can see Devon using Chrome's dev tools to test the UI. In fact, Devon can spin up any infrastructure your app needs for end-to-end testing. Today, we're giving Devon a new place to work in Slack with code channels. In our engineering threads, when a task gets big enough, Devon will now move the conversation into a dedicated code channel. Code channels are new spaces to work with agents, and they allow us to bring more of Devon into Slack. In a code channel, I can clearly see Devon's status, PRs, and actions with the cloud session always pinned right here in the context bar. And anyone can join a code channel to help Devon finish the work. Let's add Joseph, our designer, to help with some new UI layouts. If he wanted to, he could kick off some other Devon sessions to show us more prototypes in parallel. Thankfully, he has a design ready. As Devon is working, I can click here on the session artifacts to see any code changes, plans, screenshots, and videos that Devon is working on. Because Devon has shown us a demo of the feature, we can be more confident Devon has done a lot of the work correctly. All right here from Slack. Devon in Slack has access to all the same skills and tools you use on Devon Cloud. Let's try one of my favorites. Great job team at Devon. Get the team their usual Boba order from DoorDash. Deliver ASAP to the office. Code channels are live today in Devon and Slack. We can't wait to see what you build.

[00:30:39] Speaker 5: If you're a marketing agent, they can also get you Boba. That's pretty amazing.

[00:30:44] Speaker 7: I didn't expect that in the demo. Jamie, share your thoughts. Well, you know, what I actually think is even more amazing than the Boba, which is deeply impressive, is how proactive Devon is. I mean, I think, you know, they talked about how critical it is to go from idea to production as quickly as possible. And that's what makes the difference between an AI native company and a company that's just kind of like hobbling along with AI. And Devon is doing that every pass of the way, from being proactive to pulling in the code owners to make sure that it's safe, the screenshots and channel. I just think what the team has built is really incredible. And it makes so much more of what makes Devon really awesome, really shine inside of Slack.

[00:31:25] Speaker 5: Yeah, I know it's a developer favorite these days, and I can absolutely see why.

[00:31:28] Speaker 7: Yeah.

[00:31:29] Speaker 5: So we also have one more partner that Katie got to sit down with, and that was Vercel. She sat down with Malta. Jamie?

[00:31:36] Speaker 7: Well, Vercel, if you don't know, is the platform that makes shipping web apps ridiculously fast and easy. Let's roll the tape.

[00:31:45] Speaker 12: Hey, Malta. I'm Katie Steigman. I'm a VP of product here at Slack. I am so excited to talk to you today. Before we jump in and start talking about agents and code channels in Vercel, why don't you tell us a little bit more about yourself and what you do?

[00:31:59] Speaker 15: Absolutely. Great to meet you, Katie. Thanks for having me here. Again, my name is Malta. I'm CTO of Vercel, and we're building agentic infrastructure in part on Slack.

[00:32:09] Speaker 12: Okay. So just to start off, I would love to hear a little bit about why Vercel chose Slack as the environment to work in with your agents.

[00:32:17] Speaker 15: Yeah. One thing that's, I think, very, very fair to say is that Slack is the absolute life blot of Vercel. We're absolutely living in it. It's both for... It's where all work happens. It's internal. We try to get our customers in here, and that was true before agents, but it definitely has been accelerated with them. When I report a bug, why wouldn't I just send an agent on the way to fix it? This has been something that's just very ingrained in our culture now. I think one of the big, maybe surprising things about the advent of agents is that the conversational interface is really here to stay. I mean, quite obviously, Slack has been a place where people talk to each other, and so it's very natural to bring the agents into that conversation.

[00:33:02] Speaker 12: So on that sort of coding agent note, what are you seeing so far with the code channels? That's an experiential difference of working with agents in Slack. What are your sort of first impressions, and what are you seeing out of the gate about that being sort of another way that you can manage the agentic experience in Slack? What are your first impressions?

[00:33:24] Speaker 15: Yeah, I think there's a few things, but it primarily comes down to essentially uplifting the complexity of the tasks that I can complete in this environment by one order of magnitude. And I think it's actually quite easy to kind of visualize this, where previously your interaction was probably confined to a single thread. And we all know that that can get somewhat unwieldy if you want to talk about design, and you want to talk about the system architecture, and you want to talk about optimization. And those are really three topics that might relate to the same overarching topic. And so with a coding channel, you just get a space where you can have these three conversations, but confine themselves into their own threads.

[00:34:07] Speaker 12: There are plenty of times where you're working with a coding agent, and you can handle it in the thread, but I love the way you say that. It sort of expands the complexity of the task, especially a particularly collaborative task that you can actually get done with the coding agent. What's the coolest thing that you've seen being built in Slack lately with agents in your world?

[00:34:25] Speaker 15: To be honest, we built some substantial portion of the software that we built in Slack. So actually, I will say for myself as the CTO, all production software I write is in Slack.

[00:34:41] Speaker 12: Thinking ahead, what do you think gets missed? It's going to be a big part of working with agents in the future, in the near-ish future.

[00:34:47] Speaker 15: Well, how long of a conversation do you want to have? First of all, it definitely is a transformation that's happening where folks are coming from a world where they're using coding agents in a relatively lean-forward type of way versus saying, I'm just going to fire up the task, and I'll have the agent ping me back when it's done. And there's zero expectation that I babysit it along the way. So that workflow, obviously, that's why it's so compatible with Slack because you're in this asynchronous world where you get a notification back when something's actually happening or when the agent does need feedback.

[00:35:25] Speaker 12: As more coding workflows move into shared spaces like Slack, how do you see the role of a traditional IDE changing over the next few years?

[00:35:34] Speaker 15: I feel like I haven't really used an IDE in years. The transformation that we are seeing is one that some folks call going from the traditional way of doing software engineering to something that's more like a software factory. And in that world, I'm definitely triggering tasks, but I also get feedback, and then there's background work. And so we need this interface to that software factory that's very focused on the asynchronous work. And Slack, quite obviously, is kind of the place to do this. The other thing that we are definitely seeing is this. So we recently released a new framework for building agent called Eve. With our customers, we've seen just this absolute explosion of them building both internal agents and software factories. And so because Eve is so well integrated with Slack, you have kind of this natural collaboration happening there. And what I see with coding channels is that those agents now, again, they get up-leveled to have a space that is more holistic, that can have more facets in the discussion about the same underlying change. And so I think these things are starting to fit together really well.

[00:36:50] Speaker 12: Okay, Malte, we've been talking a lot about the Vercel agent and how you guys are building. I would love to see what it looks like. Would you mind showing us?

[00:36:58] Speaker 15: Yeah, absolutely.

[00:36:59] Speaker 16: Let's take a look. So we're seeing a front-end issue on our website. Let's ask Vercel agent to investigate this using code channels. The agent creates a new channel, brings the original message in as context, and gets to work. On the right-hand side, Vercel agent has a dashboard to give everyone a quick overview of what's happening, the latest running summary, available slash commands, action items, and a timeline of the investigation. Since this is a public channel, anyone can join the investigation. My teammate Brian just joined, and we can see the Notion app joined as well. The investigation just finished with the diagnosis, and it even took a screenshot of the bug. Here, I'm going to invoke the implement fix slash command.

[00:37:43] Speaker 17: The agent then proposes a plan, and I just click approve. When it's done, the PR defloads directly inside the code channel, so we can review the changes without leaving Slack. Look, Brian just asked the Vercel agent to take a screenshot of the preview deployment.

[00:38:20] Speaker 16: Meanwhile, the CIN deployments tab shows us that the deployment completed and all PR checks passed.

[00:38:34] Speaker 17: We now have verification of the fix, and the site is rendering correctly again. To wrap it up, I'll ask Vercel agent to write an incident report. The report gets written right inside the code channel, and the team can continue editing it together.

[00:39:05] Speaker 16: So that's Slack code channels with Vercel agent, a shared workspace where teams and agents can collaborate.

[00:39:13] Speaker 5: That was incredible to see that from Vercel, and I want to welcome Rob back to the couch.

[00:39:17] Speaker 6: Thank you for having me back.

[00:39:19] Speaker 5: Good to have you back. So, Jamie, I've been doing a lot of talking, but I wanted Rob to hear a little bit more from you specifically about Vercel. What's unique about their story?

[00:39:26] Speaker 6: Yeah. I guess a little bit of a segue. I just want to say a personal congratulations to G and Gene and Malta and the team at Vercel for their continued success, their most recent fundraising round. They're such good partners to us. They help us make the Slack product better. They've also disproportionately, through a number of their projects, contributed to the increase in the adoption of agents and development of agents specifically for Slack. Something really stood out to me about what Malta said in there, which is all of his production software is now written in Slack, and I think that's immensely powerful, and we're seeing more and more companies look at backlogs and bugs differently. You really don't have to prioritize engineering time and plan in the same way that you used to. So much is doable with these coding agents and some human input, like you just saw. This is a huge deal because it unlocks higher order work for your product teams, your engineers, your designers, and I truly think it's the end of product backlogs. The minute you have an idea or the minute you see a problem, you just do what Malta did and you just address it in Slack with your coding agent.

[00:40:32] Speaker 5: I like that. Do what Malta did and just end all product backlogs. Yes. That's a great note. Anyway. Thank you so much, Jamie. Thank you so much, Rob, and everyone else who joined us in the studio today, Satyaja, Katie. I want to give you both an opportunity to give us any last words before we wrap up. Rob, I'm going to go right back to you.

[00:40:49] Speaker 6: I just couldn't be more excited for this time. Like I said, people are spending a lot of money on AI, and we at Slack, I think, can help them get a better return on that investment. Use the context that's in Slack. Put your agents into Slack in a multiplayer way so people can use them together and learn from each other, and try code channels. I think it's going to make your entire company more productive.

[00:41:07] Speaker 1: Yeah.

[00:41:08] Speaker 5: Jamie?

[00:41:08] Speaker 7: Well, I just wanted to say this is why really multiplayer AI is critical. That compounding effect we've been talking a lot about, that's going to allow you to deliver the next order of productivity and really shift how you're working from doing the same old stuff faster to operating in a completely new way. I really think that code channels are going to be the thing that allows every company, not just the frontier AI companies, to start operating that way.

[00:41:36] Speaker 5: I love that. It's a great message to end on. Thank you so much for both of you being here and all the amazing work the team has done to make Slack code a reality. Now, for our audience, do not forget, if you are using agents from any one of our launch partners, GitHub, Anthropic, Vercel, Cognition, you can get started with Slack code today. We're introducing even more capabilities that'll enable our partners to build the most powerful multiplayer agents possible for coding and beyond. So keep an eye out for more platform, product, and partner announcements. We hope that you actually come and learn more about that at Dreamforce. It's just a few weeks away. It's your opportunity to dig deeper and learn more about Slack code and all the amazing innovation that's been coming with Slack. Now, in the meantime, of course, we want to make sure you're following all of our social channels because that's where we'll be sharing all of these great announcements and more content from our partners, tips, tricks, and everything you can think of. Now, before we go, don't go yet. We want to hear from you. We want you to share with us what you think about this event. So please take a moment, take this short survey. There's a QR code somewhere on your screen, somewhere around me. So take a few minutes and do that. Thank you so much for joining us. Really appreciate you watching, and have a great rest of your day.