Grok Bot for Sales Engineers

0:00Cool. Awesome. Welcome to day two of Grockbot Galaxy. I'm super excited to be here. My name is Amita. I am a field engineer here at Cursor and we have a really great great suite of workshops for today. Um we're starting off with Grockbot for sales engineers and then we'll have Grockbot for sales later today. Grockbot for SDRs. Um we have uh one specific for customer support. Um, and then throughout all of this, we'll be following along with Matt, Lauren, and Roshan on how their company is going. So, super exciting day. Thank you all for tuning in. Um, and yeah, let's

0:34get started. All right, so today's session is Grockbot for sales engineers. Um, what I'm planning on doing with today's session is kind of similar to the other ones. I'm showing three bots. Um, these are three bots that I actually use on a day-to-day. Um, two of them will be in a demo project. So, you'll see kind of how

0:51it works in sort of a sandbox scenario. And then one of them I'll show I actually use every day for curating particular customer kind of case study slides when I need to do uh external facing calls. Um, I'll do a quick overview of Grockbot for folks who maybe this is the first time turning in um or

1:09tuning into the Grockbot Galaxy event. Um, so we'll do a quick overview about Grockbot. I'll talk a little bit about why we created it, what makes it different from some of the other products out there, and then we'll get into the use cases that I'll talked about, and then a demo. And we'll have some time for questions at the end. So, if you do have questions, please save them for the last 15 minutes or so. And we'll have some build time for folks in person. So, I'm happy to come around and

1:32see how you all are using Grockbot. There are little uh coasters, I think, at your table that show you different use cases of how we use Grockbot. So, if you want to scan the back, there's a QR code that kind of shows you a template of a Grockbot you might be interested in. Okay. Um, we showed this slide yesterday, but I want to kind of hone in on exactly what this means, where we're moving towards a world that's less kind of task based and one-off kind of request to an AI tool and more so a team of bots kind of at your service working together on a task. Um, so we're really

2:05going from a, you know, a chatbot type world to more like a team of colleagues. And what's nice about this is when I think about Grockbot, I think about a tool that actually gives me finished work. Um, and that's the beauty of it is it's not meant to be something that you have to constantly babysit or something that is always, you know, just at your computer that you always have to keep your laptop open for. It's literally meant to be like a teammate where, you know, it can work maybe while you're on vacation, you're away, it can work while you're asleep. it can work because you

2:34can, you know, talk to it on your phone like you would a normal colleague. And so we're going towards a world um where that's going to be happening all the time for all different types of domains. So whether you're in sales, you're in product, you're in engineering, you're going to have a staff or a team of bots at your service that can help you basically accomplish work that maybe you didn't have time for before, maybe you're not looking forward to, but you're also kind of having them ready to go with the context and memory that

3:00you're kind of providing them. Okay, so for folks who haven't seen Grockbot before, um first I recommend downloading it. Uh it's on Android, iOS, and the Mac. So if you haven't downloaded it yet, please do. Um and it's kind of the best way to to see it and try it out for yourself. Um there's

3:17a couple main things to highlight here. The first is kind of the interface itself. So if you've used iMessage or maybe any form of like a a chat or texting interface, you'll see that it looks pretty similar. Um, on the left is where all your bots live. Um, and as you can see, you can message them like teammates. So, you can say good morning to your bots. You can thank them. You can basically go back and forth them because they have this long lived memory that it's able to learn from, you know, what you've said before if maybe you've told it that you didn't like how it did

3:46something a particular way. It keeps that in its memory and it's able to do better the next time around. And we'll see an example of that actually in this demo. Um, you will also see that bots have their own computers. And so this is kind of what I mean where bots are like colleagues is no longer do you have to maybe connect each bot to every single service or provider that you want to have access to. Your bot can do a lot of that for you. Um an example that we did yesterday was a bot making its own Google form. Um an example that I've heard from the crowd is a bot logging

4:19into MongoDB and kind of analyzing your database and making sure that everything is, you know, up to snuff. And then I think another cool thing about computer use is you can monitor what it's doing in the computer itself. So if for example I want to take over its computer, I totally can. I can also control from an enterprise level what it has access to in that computer. So if for some reason I don't want it to access a particular website or I don't want it to download a particular thing, I can always keep that in my admin controls so that you know you don't want the bot making decisions that you

4:49wouldn't take yourself. And then automations and routines. That's a huge one. Um, and I think it's something that people constantly look for with AI tools is, hey, how can I automate or have a repetitive task be done without me having to think about it? Um, a common one that I have actually is I subscribe to a bunch of different, you know, AI newsletters and podcasts and I've started not to keep track of all of them. And so every morning I'll have bot actually look through my email, find the newsletters that are relevant to me and then just create them in a daily digest and summarize them in Slack. and then it

5:20just messages me every morning saying, "Hey, here's what you missed from, you know, these podcasts, these newsletters, and it helps me keep up to date, but it's also something that I no longer have to think about." Um, similarly, there's a very common like morning brief, morning digest for your work if there's Slack messages that you've missed overnight. Um, I do like to highlight that for distributed teams, Grockbot is really a gamecher because you can go from, you know, having to wait on work being done from other teammates or other tools to just having something that's really always on and always at your service. And so that's

5:50kind of the beauty of Grockbot is that it's really great for distributed teams, but also it's for it's a multiplier. It's a force multiplier for you wherever you are. Um, and then the last thing is sharing your bots with others. So you see here on these kind of little coasters that we have QR codes. Uh that's for you know public bots that are available on the marketplace. If you want to share bots with your team at work, you can do that as well. Uh for example, we have bots that are specific to the sales team that we all you know share. There's bots on the engineering

6:18team that they share with us as well. And so bots are meant to be multiplayer. They're meant to be used and shared. Um and also creates a sense of consistency across the company is you know you want to make sure that let's say the the sales bot is not using different resources that maybe someone else's sales bot is. So having that consistency

6:35of bots is also very important. Okay. Um I touched a little bit on some of this like why Grockbot and what makes it different. Um I think the main things to highlight here is it's very easy to use. Um I think a lot of tools when you first spin it up you don't know where to start. You don't know how to kind of connect things. Grockbot kind of learns all of that for you. Um, you'll see it actually when we work with a bot from scratch today. Uh, it will basically ask me like, hey, what am I being used for and how can I best help you? And we'll

7:04also do something cool. We'll actually ask one of the Grock bots today to spin up its own team of bots. I'm basically going to ask it, hey, what bots do you think would be helpful for you? If I had to ask you to hire your own team, what would be useful for you? And a bot can actually make that decision. And so bots spitting up other bots or bots working with their other you know teammates is what really makes Grockbot this like multiplayer force multiplier tool. Um and then as I mentioned computer use and the fact that it can do an entire task end to end. Um one thing that I really

7:36like about Grockbot is if I give it a task and again it's in the cloud so it's always on. I can come back and trust that the task will be done. It's not going to sit waiting for my approval. It's not going to go back and forth on like should I do this decision, should I not? it will finish the task and then I

7:50can tweak it or steer it however I want. But the goal is that you're not just sitting and waiting for it to get do something. You're actually coming back to finish work which really really saves a lot of time. Cool. Um so as I mentioned these are some sales engineer use cases that I have seen um at SpaceX and also three of them actually I'll be showing today. Um, and so the engineered technical resource, David Gan will probably go more into detail about this in his customer support support session. So if you're interested in that, definitely take a look. Um, but what's nice about this is we get a lot of customer

8:26questions about security. We get questions about how Grockbot was built, how is memory being stored, um, how is it, you know, what type of computer is it using, is it a Linux VM? All these technical questions, some of them are in our docs, but some of them also are just in the implementation of Grockbot, right? in the repository. And so what I do as a sales engineer is I actually have a expert I call him Sherlock and what he does is he basically has access to a repository. So he has access to you know the Grockbot repository and so if a customer asks me a question I can go ask

8:56Sherlock hey a customer is curious like do we support X Y and Z? Sherlock comes back. But what's nice about this is I've actually steered Sherlock to make sure that he doesn't ever, you know, release any IP that he phrase this thing in an easy to understand way if I need to communicate this to maybe someone more high level, someone that's not down in the weeds. And so that's kind of the beauty of this particular Grockbot is that it's meant to understand things technically, but then phrase them back to me as to how I would, you know, tell

9:24a customer. And then competitive intel. This is a really fun one. Um, competitive intel is basically a bot that I created. I named it Serena Williams because I'm a big tennis fan and she's also very famous for studying her competitors and watching how they play. Um, but what Serena does is she actually uses competitor products in her own computer and then tells me, hey, by the way, this competitor is using or is doing something different than how your product is. Um, is that something that you want to put on your road map or something that you'd want to change? And we'll see an example of that today

9:58actually of Serena actually going in trying out a competitor product, giving me feedback on, you know, what's different between my product and theirs. Um, and then maybe suggesting like what I can what actions I can take later. Um, Echo is something that I think Christa and Mark will show in a later session. It's basically live curating a deck during a customer call. Really neat tool. Um, I I don't use it often, but what I usually do is before a customer call, I'll give it context. I'll give it notes. And then what I'll do is I have this giant master deck of slides of anything from you know uh case studies

10:31that we done with customers to benchmarks everything in one kind of master deck and based on the notes that I've given it of what the customer is interested in it will curate those slides and basically highlight what's important in you know what the customer wants and then now I have like a specific deck catered to that customer that I can show later. So we'll show an example of this today. Um, but in the build session, if you all are curious about this type of bot, I'm happy to

10:55build it out with you. Okay, cool. Awesome. So, let's get started with the demo then. Um, we have three bots today. We have Mimi. Um, she's going to be kind of the slide customer proof point researcher. Sherlock, as I mentioned, is my technical expert. Um, and then Serena Williams is my competitive Intel one. Um, and that will be fun to see. So,

11:17let's see how they do. All right. So, the first thing I'm going to do is I'm going to show you my bots. They are right here. Um, I'm going to start with Mimi. So, what's nice about Mimi, you'll see that I had a conversation with her over the past I think couple days. And the goal of Mimi, if I actually click on her here, is let me actually click into, sorry, into her um into her description. Um, basically what Mimi does is she has access to a slide deck. So you can see it here. She has access to this slide deck once the computer connects. And what she's able

11:58to do is she's able to curate the deck and create customer case studies for me. So I have a bunch of slides about, you know, SpaceX and this is what I would present to a customer. But what's nice about kind of my case studies is they all look the same, right? So I have one, for example, for Starlink and how they use SpaceX. And this is based on a public blog post about how Starlink uses, you know, the API. And they all look the same. They all have the same template, which is problem, solution, impact, and then a quote from the from the blog post. And often what I do,

12:29right, is let's say I'm meeting with a hardware customer. Um, and I need to present to them an example of how one of our existing hardware customers has used Cursor or has used SpaceX. Um, I basically give Mimi a blog post or maybe even notes from a previous conversation with a customer. And all I do is I basically tell her to create a slide like this. So, we're going to do a live example of that. Um, you'll see that she

12:55did one for Jellyfish recently. Jellyfish just did a blog post about how they use cursor for code review. And so, all I did was I say, "Hey, we actually go up here. I said, I just want you to do your job for Jellyfish." And here's the blog post. And boom. You know, she went through. She said, "Okay, here's what I'm going to do for Jellyfish based on reading this blog post. This is the problem. This is the solution impact quote." And then the nice thing about Mimi is again because she has access to her own her own computer, she can create these slides for me and then take

13:22screenshots. So she's like, "Okay, this is, you know, the jellyfish logo." She was able to actually go to the jellyfish brand page, get the right logo, and put that on the slide. And then this is the problem, solution, impact, and quote. And so let's do an example of this live and see how how Mimi does. So what I'm going to do actually is I'm going to go back to the browser here. Um, and I'm going to take this blog post. I thought it was very apt because it's a Salesforce blog post about how they use cursor to cut legacy code coverage time by 85%. I'm just going to take this guy

13:53and I'm going to go to Mimi and I'm going to say, can you create another one of these slides for Salesforce for me? And I'll just link the blog post here. And actually I'll specifically just say blog post here so she knows. And so what she'll do and you know we'll come back to this at the end of the session to see how she does is she'll be able to take this um create a slide from it and then you know actually insert that slide into the slide deck here. So again what I'm opening here is Mimi's computer. As I mentioned I can take control of it. You know I can click

14:30around in it. But what you'll actually see, which is very cool, is Mimi will have her own cursor in here. So Mimi will start editing the slide. Um, she'll give me an update on, you know, when it's ready and how it's doing. Um, and then once she's done, she'll go ahead and present it back to me just like she

14:45did with Jellyfish. So, I'll let Mimi go here. I just want to make sure um that she can access this. Um, but while she's working, I'm actually going to go to the other two bots. So, Sherlock and um and Serena Williams. So, I'm just gonna switch my account here so you all can see the other two bots. Um here we go. So, Grockpot for those engineers. These are the two bots. Um I have something very similar to Serena in my in my other account that specifically looks for Grockpot competitors and you know what they do differently. Um but for today, I'm actually going to work in a booking

15:20app that we at SpaceX just created as a demo project. So the app that we actually have is something called Flyo. Uh think about it as a very simple booking flight booking app. So basically what we did is we've created this entire company called Flyo. Um they're meant to book flights for anybody. You can upgrade your seat, you can change your

15:39flight, you know, whenever you want. Very similar to Expedia or Google flights, any other flight booking app that you see. Um what's nice about this is, you know, we have created a whole like suite of engineering products that power a fly. So, there's a front end, there's a backend, there's a web app, there's an internal app for the crew to use. Um, and what's nice is my Grockbot Sherlock bot has access to all of these things. So, if I actually go back to Sherlock here, you can see I gave him kind of a description. I said, "You're a technical expert on Flyo. Um, and you

16:13support the website floowower.com." Um, I gave him access to the particular repos. So, the booking backend, the booking front end. Um, and then I also say if there's a customer issue, I want you to be able to kind of investigate and then communicate what could be wrong to a customer. Um, so I worked with Sherlock a little bit yesterday. I was just curious. I was wondering, hey, how do we support a race condition for maybe someone trying to book two flights at the same time or they're trying to book the same seat at the same time. And so what you'll see is that Sherlock actually uses cursor cloud agents under

16:44the hood, which is what makes it um really easy and what makes it able for me to close my laptop is that Sherlock is basically able to say, okay, I'm going to dig into how booking handles two people grabbing the same flight at once. Um you'll notice that it says here, all right, Flyo doesn't let two different customers confirm the same last cabin. Um the protection is in Postgress, which is good to know. Um and then it kind of covers two different use cases, right? a last remaining ticket or if someone is booking the same actual

17:10seat, what do you do in those scenarios? And then what's nice about this and this is what I would recommend for anyone working in like a customerf facing or just support role where you have to respond back about how things work is this is what you should tell a customer is that we hold the last remaining cabin space when you start checkout for 10 minutes and if someone else started first, you know, you'll get a complaint when trying to book. Um, and so this is

17:30kind of what maybe I would even tell it. I can say, "All right, draft this in is it in draft it in an email and send it to this customer." Or I can even say, "Um, okay, this looks great. Can you actually send it for me myself?" And then I can give it a particular, you know, approval to send an email or a Slack for me. Um, so this is the great thing about Sherlock is that it's able to be a technical expert and answer questions for me if I'm ever, you know, confused about how something was implemented inside of my product. Now, let's go to Serena. Um, so Serena, as I

18:00mentioned, um, is a competitive Intel bot. And so what she's going to do is she's going to use her own computer to test out competitor products, see how they compare to Flyo. So I'm assuming she'll probably test out products like Expedia, maybe Skyscanner, Google Flights. Um, and then also check and talk to Sherlock and see, you know, hey, do we implement, for example, this whole race condition thing, like does Expedia implement that or is this something specific to just how we do it? Um, does Expedia do it better? Do they talk about it in their technical blogs? Um, something that I do whenever I'm kind of

18:34researching competitors and want to try out what other products are doing is I always read their change log, their feature releases, their blogs, because it's kind of a way for me to see like, okay, what are they doing differently or how are they implementing things differently? And Serena can kind of do all of that for you. In fact, you can even set up a routine here and say, "Hey, Serena, like every week maybe, can you summarize uh what has happened recently with the Expedia technical blog or with the Skyscanner technical blog and let me know um if they've shipped anything recently that I should be aware

19:05of." Um, a very common I think tool in travel apps and websites right now is like an AI travel agent. Um, so what we'll do actually is we'll talk to Serena and see like, hey, how are you able to see that in some of these other products? Are there products that are doing that right now? And I'm actually going to have her talk to Sherlock and see what would be the technical lift actually of being able to build like an AI agent inside of Flyo. Um, so let's go ahead and get started here. I'm actually just going to say um what are some competitor products that you think

19:41are worth testing against? And you'll see that what's nice about Grockbot is it thinks and it's able to kind of communicate what it's saying if you ask it to. But for the most part it'll actually do a lot of its work behind the scenes. Um so I can always ask it sometimes which is like what are you thinking or what's going on? Um but sometimes what it does and because it kind of operates pretty fast is it'll tell you exactly how it's going to go about for example testing something like this. Um once it's done kind of percolating on this um so as you can see

20:16it's going to say okay I'm you know connecting. I'm going to be able to find you some competitor products. Um as you can see it messaged Sherlock already. Um, and so it said, "Hey Sherlock, I'm doing some competitor product testing against Flyo. Um, quick baseline ask, what can the current Flyo booking codebase support today? Do they search, you know, do they support searching, fair selection, seats, bags, um, any known gaps versus a typical booking flow that I should weigh when I compare?" So again, I didn't really give Serena any context. All I did was um, if we go back to Serena, all I did was give her a

20:48description and then she immediately was able to kind of understand what she had to do. Um, but now she's going to give me some options here. So I said, let's go with let's go with Southwest and Spirit. Let's see how those booking experiences compare to Flyo. Um, and then as you can see again, she will work with Sherlock to kind of go back and forth. So Sherlock is actually now going to say, "All right, Serena asked me for a Flyo booking flow baseline. I'm going to pull what the code actually supports, and then I'll reply back to her and let her know like, hey, we support these

21:18things that maybe Southwest and Spirit Airlines don't. Um what I'll also do while Serena is working and this is a very cool part of Grockbot is I can steer her in a particular direction. So for example if I want to say um you know I want her maybe not to test just the booking flow. I want her to test like what are the agent capabilities that are there in these flight booking apps right now. Uh, I'm going to go ahead and just say, can you see if there are any AI travel agent capabilities in any of the competitor tools that we should be aware of? And what you can also see here

21:56at the top right is that Serena is actually doing the Southwest booking flow on her own. And so, as I mentioned, a lot of times I don't get the time um or, you know, I I don't I don't have the time to be able to test out booking or any sort of competitor products what like step by step. I don't get that developer experience. And so, what is nice is that I I'm kind of outsourcing that to Serena. So, she can tell me, hey, like the the Southwest booking experience is actually great. Um, and

22:23it's actually a lot better than Fly Lo. And these are the ways that Fly Low can be better. Um, and I don't have to do this just for Southwest. I could have Serena do this for a million different, you know, booking flow websites. And so if you're ever wondering like, hey, I don't have time to see what the competitors are doing in my space or how they how they fare compared to my own product, you should really spin up a competitive Intel product. It's very powerful and also just really interesting to see like how these products do in comparison to your own product here. Um, so we can see here

22:52she's actually like picking on dates and actually trying to see how booking works in Southwest versus Flyo. Um, what's nice about this as well is, you know, you can also ask it to open up another tab and do the same thing in Spirit. You can ask it to do the same thing in Expedia. So if there's any particular things that you wanted to test, you can actually explicitly say that inside of Serena Williams um, uh, the bot. Okay, cool. So, as I said, I kind of steered her in a particular way. Let's say I want to, you know, test something more than just booking. Let's say I also want

23:26to test like the AI travel travel agent capabilities. This basically allowed her to not abandon this per se, but also just incorporate this as a part of what she's doing. So, as I mentioned, bots have memories and bots can learn. But what's nice about that is that they're also able to remember if you've asked it multiple things at once. I think sometimes what can be frustrating with other tools is you ask it a question and then you're like actually I want to do something else but the minute you ask it the second question it forgets exactly what the first question was and you know you can't really give it proper

24:01instructions on how you want to incorporate both of those things. Rockbot is actually able to really understand that. Um so as you can see here it did some few it did some like initial discussions into like what the travel agent place is for Flyo. It tells me exactly what we offer. Um, and then it's also seeing kind of like where uh or what I change for Fly Low here. So, she's able to understand my question immediately and kind of come back to me, but at the same time, she's also still doing the booking flow testing that I asked her to do earlier. So, it's kind

24:32of a mix of both, which is really great. And what makes it a force multiplier here is that this bot is able to multitask and do many things at once. Okay, I'm going to check in on Sherlock here and see how he's doing. Looks like he's looking again at like our booking flow. baseline and what we're offering as our product. You can see here that the cloud agents are being launched in cursor. Um so one is being launched in uh I believe the back end, one in the front end and I can always open them and see exactly how they're being launched, what they're doing. This is kind of the

25:02integration between Grockbot and Cursor is once you know we're able to get any technical information from the repo itself, we immediately can talk back to Serena and give her information as to how we've implemented something. Um, and this is great again when you want to talk to customers because a very common question I get is, you know, competitive differentiation. What makes one product better than the other? How do you guys implement something differently? And this is where I actually work with these bots together so that they can talk to each other, understand what the competitive threats are, and then see what we do in our repo to kind of make

25:35that or assuage those worries. All right, so I'll let these two guys run. Um, as you can see, still doing the Southwest booking here. Um, which is good. I want it to be able to get that full experience. Um, let's check back in with Mimi and see how she's doing. Um, so if we go back to this account here and I open up Mimi. Oh, let's go ahead and just resend that. Um, and that way she'll be able to get the information from that blog post again. Hopefully,

26:04she'll be able to see that. What's nice about having bots that talk to each other, um, I'm actually going to switch back to Sherlock and Serena. There we go. Um, is once they have a conversation, you can use that conversation as a template for how you want them to interact in the future. Um, so I showed this yesterday, but if folks are curious, instead of kind of going in between both of these bots and kind of seeing what they're messaging each other, sometimes what I like to do is I put two bots in a group chat and then get that visibility of how they message each other in one place. So I'll

26:39actually put Sherlock and Serena Williams in a chat. Oh, sorry. Looks like it put two different people. One second. We'll do another group chat here. Let's do Sherlock and ah Sherlock and Serena Williams. There we go. Oh, looks like it's automatically putting me and another person in the chat. So, if I add Sherlock and Serena Williams here, basically what you can do now, if we go back to that, where do we

27:10go? Sherlock and Serena. Go back here. There we go. Now both of them are in this chat here and now I can say something like um what have you found is the key differentiator in our competitors that would be significantly you know loweffort to build in the product and what they can do is they can basically have a conversation with each other both based on what they've learned already. So Sherlock now knows kind of that Serena cares about the booking flow. Serena now knows that I care about, you know, Agenta capabilities and they're able to use the context of what I've given it already and be able to

27:59talk to each other about all of these things in one place. Um, and so you'll see here that Sherlock and Serena Williams are going to work together now to come back to me with an updated kind of response or solution on what they think would make sense. Um, and what's nice about this is you can basically take the context of both these agents together and then create a PR or a product doc that you can then send off to your technical team. Um, and so you'll see here in the marketplace that we have a lot of different bots uh that are published by, you know, our team

28:30members, but also publicly, but then we also have plugins or tools that you can connect to that make it really easy to kind of plug in whatever you found out from your bots into some sort of like paper trail. Um, and so maybe what I would do here is once Sherlock and Serena have come up with a plan of, let's say they want to build, you know, a travel agent, an AI travel agent, because Southwest doesn't have one and Expedia doesn't have one, and that would be our biggest kind of moat. What I can do is I can connect it immediately to Google Docs. Um, I can connect to

28:59Atassian or Confluence. Um, if you guys use a tool, maybe you guys use Microsoft OneNote or a tool that doesn't actually have a plugin or an MCP, you can just ask uh Sherlock or Serena to log into that tool on their computer. Um, and so we covered this a little bit yesterday, but I want to hone in on how important it is that a tool not having an MCP or an API is not a blocker. Um, because the the bots have access to their own computers, they can log into whatever tools that they want or that you want on the computer and that makes it a lot

29:29easier to integrate with tools that are historically very hard to integrate with. For example, we get a lot of users that use PowerBI um, and have had trouble using AI agentically with PowerBI. bot can log into PowerBI on it on its computer. Um, similarly, I think the MongoDB example is also great where I don't believe we have a MongoDB plug-in, but that doesn't mean that you can't use MongoDB and Rockbot anymore. So, again, remember that if your bot or plugin doesn't exist inside of this marketplace, your bot can still use it because it has access to it on its

30:01computer. I want to cover one last thing while I'm here, which is you can see that I have some private skills set up. Um, this was one that I created yesterday, which is actually showing a bot how to do a slide deck animation. Um, what's cool about this is you can teach it multiple different skills and it can apply those skills at the right time depending on when it should be used. Um, so we can actually teach Sherlock a task or teach Serena a task. For example, you can see here that, you know, she's doing something with her computer with Booking Southwest. So, I'll let her keep doing

30:35that. But let's say I want to teach Sherlock something. Let's say I want to teach it um how to let me think maybe how to go and see for itself if there is a technical post released in like the GitHub change log for a competitor. So what I can do is I can just take over Sherlock screen. Um and I would recommend everyone actually doing something like this during the build session with me. It's one of the biggest lever ways to use Grockbot is actually by teaching it something via video recording. And I'm just going to say teach a task. And I'm going to go open up Google. You can see it's pretty

31:08fast. And I'm going to say something like Expedia technical blog post. And then I'm just going to go ahead and click this guy. I know, ironic. Um, and then we're going to scroll in here. And you can see, okay, they have like a medium. Cool. And they have a bunch of stuff. This is about GraphQL. Maybe I want to search and see is there anything AI

31:31related here? And then look into that. Looks like nothing really. Oh, maybe this one. Reimagining platform engineering for aentic future. Scroll here. See exactly what they're doing and are building. And then that's it. I pause it and I'm done. And now what's nice about this is Sherlock will now have all the details about that skill or that video. And what I can do is I can say, okay, you know, now use that skill and apply it to other other competitors. So do the same thing for uh for Southwest or for Spirit or for a Skyscanner for Google Flights. Um and now it's going to be able to know

32:10how to scan technical blogs, look for AI related things. You can see here it says, "Okay, the recording is finished. Learn the task from it." And then I can even highlight particular things from that skill or that video and tell Grockbot to particularly pay attention to those. So I can say something like, and I'll use voice mode actually for this. I'll say notice that I specifically looked for AI related blog posts in my competitor products. Um, and I wanted to make sure that we are staying on top of our AI tooling inside of our own product. So, I'm going to try and apply this to as many competitors as

32:44possible. Um, as you can see, I did this skill inside of Sherlock. I could have done it inside of um, Serena as well. It's kind of up to you on where you think the skill will be leveraged most, but as you can see here in the marketplace, the skills are basically being used by all the bots at all times. So, um it doesn't matter where you create the skill, the skill will still be learned and used by

33:07by the different bots here. Cool. Okay. So, we're going to go back and see how Serena is doing. Um actually let's check on our group chat where I basically asked them to come together as a team and figure out okay where can we actually make a change or a difference with this product. Um so you can see Sherlock was saying okay from the codebase side of things looks like uh every competitor treats this as basic table stakes which is for managing

33:34booking on the back end uh for trips. And so that's one easy way that we can fix things. Also the second is a flexible date calendar. Um, looks like this is live. The front end client already has it wired, but it never calls it. This would be a classic differentiator for us and very easy to wire up. And then Serena says, "Okay, roundtrip and bags look big, but they're not low effort here. The bags aren't a skew. I'm curious which of manage booking versus calendar felt most

34:01painful when you were on those tools." So, they're working together here. They're kind of figuring out um what exactly is, you know, the biggest lift here or what would be the easiest thing to implement. I want to notice or I want to point out something here which is that Sherlock tags Serena Williams in this and this is a cool way for bots again to communicate with each other is if uh Sherlock exact wanted exactly you know something specific from a bot you can either do that yourself so for example I can tag Serena myself and say hey Serena I'm specifically talking to you but the bot can also do that on its

34:34own as well um and that's a really cool way for bots to communicate and specifically get information that they need from you a bot with memory and context on that information. Okay, so we have these two bots. I think I mentioned earlier that I wanted to do something different which is I wanted to ask the bots what bots would be helpful to them and I do this sometimes when you know I'm working and I have a team but I want to see like okay what can actually what can I do to help this team and to make it you know operate faster or more efficiently or with better memory and

35:07context. So, what I'll do here um is I'll actually go let's go to Sherlock. Sure. And I can say I can say based on the work that you and Serena I'll tag Serena here are doing, what bots would be helpful for you to continue doing great work um specifically around competitive differentiation and sales engineering. and I'll say go ahead and spin up those bots for me. So, this is a little known fact

35:45which is bots can spin up other bots. You can ask a bot to say, for example, if you have a chief of staff bot, which is I think one of our most common bots, um you can say, "Hey, I want you to spin up three bots that you think would be the most helpful for my day-to-day." And usually what the chief of staff bot will do is it'll spin up like an inbox manager, maybe a scheduler and maybe like a a a calendar manager or something like that. Um, but I'm actually very curious here to see what the bot thinks it would be helpful in this scenario and

36:14how exactly it chooses those bots to spin up. Um, and this is also great if you're starting out with Grockbot for the first time. Let's say you're a sales engineer, you work at a particular company, you guys have very particular workflows. Um, and you know, you do demos, you do particular things that are very specific and are not, you know, onetoone with what I showed. Go ahead and ask Rockbot to spin up some bots for you and say, "Hey, I want to be able to create a demo for uh, you know, a PC that I have with my AE in a couple weeks. It's for this particular

36:42customer. Spin up three bots for me that you think would be useful in order for me to accomplish this task." Um, Grockpot is here to help you. It's not here to add extra work or to, you know, give you extra managerial responsibilities. It's here to tackle problems together on the same page. And so that's kind of why I like having it spin up its own bots is because it's able to do that without me kind of having to think exactly what I want those bots to be doing. So good kind of trick if you want to to work uh on multiple things or you know whether that

37:15be competitive differentiation, creating demos, uh working inside of Salesforce, you can do all those things with these bots kind of here. Okay, we have about eight or so minutes left. I want to check back in with Mimi and see how she's doing. Um I know there was a bit of an outage earlier today.

37:33Okay, cool. Looks like she's working. She says, "Already on Salesforce from your last message. I'm checking whether the pair is in yet. I'm still building the Salesforce pair. I'll send both screenshots when it's in the deck." Um, looks like she's having trouble logging in again. Let's see if we can go back in and log in for her. So, again, this is an example of how you can take control of logging into uh Oh, look. That looks great. Um, you can take control of logging into a uh slide or doc or anything in Google Drive for you. Um, looks like it actually created those two slides perfectly for me. So, it the logo

38:08was a little small, but at least it got the right logo. And then what it also did is it took that exact uh same blog post that I gave it and created a slide for me. So, it said here's the problem. Legacy repos were far below the coverage mandate. This is the solution. This is the impact. And this is a quote from the blog. Easy. Done within, you know, the last 10, 15 minutes or so. And now I have a nice kind of template of a slide that I can give. Let's say if I'm speaking to Service Now or any other Salesforce competitor to be like, "Hey,

38:38by the way, we've done cursor with Salesforce and they had a great, you know, uh, use case with us." So, we can do this one more time if you guys want to see it again. We can just keep Mimi in the loop. Um, we have a good use case with Grab. Grab uses cursor. Grab cursor

38:54case study. Let's see. Here we go. Another blog post published by Grab. I'm going to go ahead and do this one last one. Go back to Mimi here and say, "Great work. Can you do one for Grab?" And again, as I mentioned with the steering, I could have sent all of these to Mimi at once. I could say, "Hey, do Salesforce, do Grab, do this other customer story that we have." And then she can create all of these. And then what what I could do, which is kind of what I showed earlier with the Echo uh Grockbot, is when I have my customer, you know, call or I have to present

39:28these, I can tell Mimi, hey, hide the slides or hide the case studies that wouldn't be relevant for this customer. Let's say the customer is Uber. They don't maybe they care more about the grab case study than they do about the Salesforce case study, something similar to that. Um, and so I can kind of send all of these off. Again, I can do these from my phone. I don't even have to have my laptop open. Mimi can work on all of these case studies and all of these slides for me. She knows exactly how to do them now given that I gave her the exact template which is logo and then

39:56problem solution impact and quote. Um and then I can just you know feel free to use these slides whenever I want. Um and that's actually how I make a lot of my slides with uh with Rockbot is I ask it to follow a template because I don't want it to, you know, come up with anything too sloppy or too wordy. I know exactly how I want my slide to be organized. But then once I had the content for it, for example, this blog post, Mimi the Grockbot was able to be smart enough to take the most important parts of this blog post, organize them into the right places, put them into a

40:27Google slide for me, and then now I'm able to go to my customer ready and prepped. So it's a really beautiful end-to-end flow. kind of bringing it back to what I said earlier, which is finished work is the best part of Grockbot is you are no longer spending time reading through this blog, figuring out what the main points are, putting that into a slide, designing the slide, making sure everything looks nice, getting the logo. Rockbot did all of

40:50that within the last 10 to 15 minutes. And it can do that for as many different customers as you want. So, as you can see, it says, "All right, grab next. I'm going to check that the Salforce status works." Then pull that engineering post. I'm actually going to say Salesforce looks great. You can go ahead and start

41:06working on Grab. And hopefully by the time Q&A is over today, we can see the Grab slide as well in addition to the Salesforce slide. But this is what I want to highlight when it comes to you know what makes my job easier is a lot of times I have to create slides for you know sometimes 15 20 different customers in a week and having Mimi as like my slide guru but also my customer proof point reader. Um, actually sometimes what I'll do is I'll ask her to to crawl the web and just see all of these blog posts like the Salesforce one, the Grab one. These are

41:38not published by cursor. These are published by these companies. And sometimes we don't really get updates as to when these companies publish these posts. So sometimes I'll ask me be like, hey, have there been any customer posts recently about Cursor or SpaceX recently that would be good to highlight in my deck? And then if so, go ahead and create a slide for it. And that can that's something that can easily become a routine. Um, and so really, really great and easy. Um, and I just love slides that look nice. You know, everyone, I think, has seen a lot of slides that they all look the same. They

42:07all use that same purple color because they're all AI generated, but not anymore. Not when you give bot the template, the memory, the context of how you want things to get done, and then it's able to work off of that. Okay. Um, I'll let Mimi work on Grab. I'll come back to her during the Q&A. I do want to quickly go back to Sherlock and Serena. um and see how both of them are doing. As I mentioned, I taught Sherlock how to use that AI kind of blog skill so that it's able to find out what our competitors are doing. Um but you can also see here that these are the

42:40three bots that um that my uh Sherlock bot spun up. So, if you remember earlier, I said, "Hey, based on the work that you and Serena are doing, what bots would be helpful for you to continue doing great work?" And you can see it spun up those three bots for me. So, let's see what these bots do. battle card blare competitive differentiation and sales engineering battle cards. So this is useful. So in addition to maybe just having customer case study slides, maybe what I can do with battle card player is she can work with Mimi and create slides against my competitors. So not only do I have case studies on what

43:13customers like Cursor, I can do have case studies on, you know, what customers are different from Cursor and what makes them different. So this looks great. It says it combines Serena's hands-on competitor product findings with Sherlock's Fly low codebase. Produce short SE ready battle cards, competitor claims, fly low reality, etc., etc. Looks good. I'll probably keep her around. Demo Drake, let's see what you're up to. Sales engineering demo and talk track specialist. Build confident no hallucination demo scripts and call talk tracks that map customer pain to the fly low live flow. Ground every claim in Sherlock. I love this. I think this is important is that it it

43:49knows that Sherlock is the technical expert and so it refers to that going forward and then when competitive contrast is needed you can pull from Serena or from Battlecard Blair. So this is what I'm saying is that you have this team now that it's all working together on the same outcome and the same projects and that makes things so much easier when you have for example so many different things to manage or to keep track of. You get all of that here and then the last one is AI radar. Let's see what this one is doing. Okay. Track competitor AI tooling from public tech blogs. So, this honestly is maybe a

44:19little too similar to Serena. Maybe I need him, maybe I don't. Um, maybe I I don't want him because he doesn't have a cute aliterative name like the other ones. Um, but you can see here that this one prefers to, you know, specifically go through engineering posts. It opens each blog, finds the AI related post. So, it's actually using that skill that

44:36I taught um I taught Sherlock earlier. Um and then again uses Sherlock as a source of truth always and then hands any you know testing or user user flows to Serena to do on her own computer. So as I mentioned you can always spin up your own bots just by asking them to do it. Um a very I think common uh joke at cursor is or at SpaceX is if someone asks us like can we do this? I'm like ask rockbot. Um and a lot of times that's how we work is there are times where Grockbot will say no. In fact, I think that's actually a big advantage of

45:07the rock model is it's one of the few models I feel that actually pushes back um and will say I cannot do something. It won't always kind of just be sick of fantic and tell me yes or or yeah, of course, you're absolutely right. Um and so sometimes what I'll do is I'll use you know these these bots as my as my like as my helpers basically to figure out what else I can do next. Um and

45:30that's that's the beauty of Grockbot. Um, I'll bring Mimi back up, but I think um, we are about to wrap up pretty soon. Um, yeah, we'll do some Q&A uh, for folks in person, but thank you so much for joining and I I hope this was useful for everybody.