Slackbot as an AI Co-Host: Contextual Work Help (Full Transcript)

[00:00:00] Speaker 1: Have you ever wanted your own personal A.I. agent for work? Well, you've got one now. Let's go meet Slackbot. Hello and welcome back to Slack School. My name is Mike Reynolds. I'm your host. I'm part of the Slack team here at Salesforce, and I'm very excited today to be talking to you about Slackbot. Slackbot is your A.I. agent for work. What does that mean? Well, we're going to learn right now. I thought about a few different ways that I could do this, but the most interesting is to kind of throw Slackbot into it and see what happens. So here's what I'm going to do. I've made four prompts that we can use with Slackbot, and I'm going to tell Slackbot that it's going to co-host today's episode with me. Let's see how this goes. One last thing before we dive in. It's important that we make all of our purchasing decisions based off of generally available products and features and not the forward-looking statements that I am definitely about to make. So here's what I did. I've got these four prompts. We're going to tell Slackbot that they're co-hosting the show, and then we're just going to see what happens. This should work because of context. Context is king. So what is context and why does it matter, and what is Slackbot? Well, Slackbot will tell you what they are in just a second, but context matters so much because with all of the information that Slack is in possession of, right? Think about it. It has connections to my calendar, to my inbox. It's got all of the actual conversations I have with all of my coworkers about everything I do at work, which means it understands what Slack School is. It knows where it came from. It knows what it's trying to do. So I should be able to simply tell it that I want it to help me co-host the show, and that all should work because this is an AI agent for work and this is what I do for work. So let's see. So I've said, hey Slackbot, I'm filming the Slack School episode all about you. How about you and I co-host this episode together? How about you kick things off by introducing yourself? So here's what Slackbot said. Hey Mike, and I love the use of this emoji. That's an emoji I use all the time. Well, it's got a lot to say. So let's see here. I'd be absolutely thrilled to co-host this episode with you. This is super cool. Let's do this thing. This is very much my voice and tone, so I can already tell that it's definitely talking to me, right? Introducing myself. Hey everyone, I'm Slackbot, your personal AI agent for work. Think of me as your always-on AI teammate who lives right here in Slack, ready to help you navigate everything that happens in your workspace. This is really pretty good, right? I've got a very good structured formatted response. It understood exactly what I was doing. And I also like this. What can I do? Well, it can help you with all sorts of things. Better searching, helping me summarize information that I've missed. I like that it wants me to get some dad jokes. That's great. Finding people and expertise. That's definitely a thing I've used Slackbot before. I had a technical issue. It turned out that it was a big like SEV1 bug. Slackbot found the problem and the solution and the team to fix it immediately. Was really, really impressive. Helps me work with files, connecting to Salesforce, managing my time. Who gets to work with Slackbot? This is another good detail. I'm glad it told me this. Here's the really great news. I'm included with Slack's enterprise grid plans at no additional cost per interaction. So there's no consumption, pricing, or usage. That's fantastic. Love that. I particularly like this last line. So are you ready to dive into some use cases? I'm excited to show everyone what we can do together. I love that because in the draft that I have for this episode in the canvas, I did say the three things that I wanted to talk about in the order in which I wanted to talk about them. So it's good. It knows what's going on. I'm going to grab my second prompt here and paste it in. So now I've said, what do you think are some of the best ways that people can put you to work? Also, based on all of the ways you've helped me, what do you think was the most impactful thing you've done for me? I'm interested to see what it thinks. So looking over this response, the best ways people can put me to work based on what I see working really well. There are three powerful use cases. Using it as a catch-up assistant, the finding things expert. This is a really big one. I will talk about that just a minute. And then the meeting prep and context builder. That's also a huge one. Let's talk about each one of these. The catch me up. Very regularly, particularly now after the holidays, I've come back in and there's been lots of information from a lot of different people planning an upcoming event. There's this really big event we're getting ready to do called TDX. You might have heard of it. And I missed some of that because I was out. Well, I came back. I was able to talk to Slack bot. Hey, can you catch me up on all this TDX stuff? Because we've already done a lot of planning and I want to make sure that I know what's going on. I got really good summaries out of that. So this is a great use case. The find that thing, oh my gosh, I cannot tell you the number of times. I actually did this at Dreamforce. I needed an updated version of the deck. But the person who sent it to me sent it somewhere else and I couldn't find it. Slack bot found it for me. It only took a few minutes. And all I said was, I have this version of the deck, but I need an updated version. And I know that somebody else is the owner did a really good job. The meeting prep and context builder. This is huge. And one thing that Slack bot's not pointing out here is that Slack bot is connected to my Salesforce instance. And so if I'm getting ready to meet with a customer, I can ask questions about that customer by name and they will go find that customer and bring back and gather all the information that I need that I have access to from Salesforce. That's really awesome. And then the most impactful thing that Slack bot thinks it's done for me, looking back at our conversations, the most impactful thing has been helping you quickly synthesize complex information when you're under time pressure. And that is true because I love to procrastinate. It's one of my favorite things. And Slack bot is really good about digging out of a hole when I procrastinate too much. So for this next question, I've asked Slack bot if there are any tips or tricks that we should know. And particularly, I wanted to ask about how natural the conversation could be. Because Slack bot's already got tons and tons of context. So is that mountain of context that's already sitting on enough? Or do I have to really think about how I'm crafting prompts? Because I've noticed personally, when I use LLMs that are disconnected to Slack, I spend a lot of time giving context that makes the answers better. And without all that extra effort for me, every single time explaining to Gemini what I'm doing, or that I work for Salesforce and that I do have access to Enterprise Grid Slack, even though I'm using a biz plus instance or something like that, Slack bot knows I can do that. Other LLMs don't because it's missing that context. So I asked specifically about that. Let's see what it says. Great question, Mike. And honestly, one of my favorite things about how I'm designed, you don't need to overthink it. Thank you. This is a great example. Instead of search channels for marketing budget Q4 2025, you could say, where's the marketing discussion or the marketing budget discussion from last quarter? And Slack bot will figure it out, which is really great. We do have some tips here. Be specific when it matters, right? If you want info from a particular channel, a person, just mention it directly. Like what did this say about whatever? That's great. That type of context, that specificity does help narrow down searching. You get better results and a little bit faster results. I love this. It's showing out. I always cite my sources. This is one of the things that I love about Slack bot so much. I learn your context over time. The more we chat, the better I understand what matters to you. When you ask, what should I focus on today? I'm looking at your calendar, your mentions, your threads. It's building this big picture of priorities based around your conversations. This is true. And I've seen this play out over time. I've been playing with Slack bot since about Dreamforce. So I've had a few months to really kind of build this context with Slack bot. And it does truly learn and grow. I've seen some simple things like it's using emojis the way that I use them and it's using the same voice and tone that I do. But I've also noticed that Slack bot is giving me more bulleted answers where it knows that my context, I'm looking for a specific thing. I'm rarely looking for these long explanations and big paragraphs. I'm looking for key points of detail. And so it's actually structuring answers in ways that I communicate. And that's what I'm getting back. It's really, really great experience. This fourth item it's pointing out here, my scope is tied to your permissions. And this is key, right? And it's specifically calling this out. It knows who I am, right? And Slack bot actually not only understands who I am, but it understands what I'm allowed to have access to. Our Slack is connected to a Salesforce instance, but I'm not assistant administrator. I don't have access to every piece of data over there. And so Slack bot's access has been limited to mirror mine when Slack bot goes to Salesforce. And that's an excellent way to make sure that I don't get access to something I'm not supposed to have access to just because I was using an AI tool. The bottom line, don't overthink it. Just start chatting with me. I really love that. It's very simple. Again, another emoji I love to use. So let's ask this last question. So I said, thanks for helping me co-host today. I really appreciate it. But before we say goodbye, what do you think we should be the most excited about as we look towards the future of Slack bot? I'm going to start explaining some of this while Slack bot is thinking. What we're going to have is Slack as an operating system. And what that means is that Slack is going to be the primary user experience for most of the work that we do. We've already seen that with tools like Salesforce channels. I don't have to go into lightning for the vast majority of my job. I can stay right in Slack in the flow of work and see that record, edit that record, do those things right there. But I can also do that with other apps. The whole third party ecosystem is really showing up and they're showing up with agents. And so almost all of the systems that I'm working with, they have an agent and their agent is in Slack. Slack bot is going to be able to sit in the middle and know all of these different agents and what they do and how to work with them. When we think about the future, we're going to have lots of different tools that are natively living inside of Slack. And those tools are all going to have their own AI agent. It's going to be complicated trying to remember which agent does which thing. And heaven forbid I accidentally at mention an agent and say something that was really designed for a different agent. That could lead to bad outcomes or leaked data. If I have Slack bot in the picture, I can train Slack bot to understand what each of these different agents is for. And Slack bot can help me remember what each of these agents is for. And so as the job continues to get more and more complicated, Slack bot can be the thing that helps me stay on top of everything that I'm doing. Because Slack bot understands all of these relationships as well. And so driven by me, Slack bot can help all of these systems really, really work well. And then I can be incredibly impactful. But let's see what Slack bot says it's the most excited about. So right now, I'm your personal AI agent for work. I can help you search, summarize and find information and navigate your workspace. That's pretty powerful. But the real magic is coming when you become the single interface for all of the agents at work. That's just what I talked about. So it's right in line there. I particularly love the way that Slack bot wrapped this up, because I said before we say goodbye. So it's got that context and knows I'm trying to end the show. And now here Slack bot is saying, don't forget to remind everyone to like and subscribe. I do have a tendency to forget that. You always forget that part and have to run back to do it. And to everyone watching, thank you so much for spending time with us today. You're doing amazing work. And I'm genuinely excited to be here to help make your days a little easier and a lot more productive. So keep being honest or keep being awesome and see you in Slack. Well there you have it. Slack bot, your new AI agent for work is here. I am so excited for you all to get hands on with Slack bot. Head over to the Slack community workspace, slackcommunity.com and you can try it for yourself. Don't forget to like and subscribe. Thanks for reminding me and have a great day. Hey, you really are awesome. Be cool. The amount of cat hair you managed to get on my desk in such a short period of time.