# I Let an AI Agent Run My Business — Here's What Happened

Channel: The Next New Thing
Video: https://www.youtube.com/watch?v=n8mXBQpzqQA
Duration: 32 min
Language: English
Words: 6670
Transcript page: https://viewrankai.com/tools/youtube-transcript/n8mXBQpzqQA

---

[0:00] He got his AI to do his work for him. I have another agent that generates the action items, and then it attempts to finish the action items. Everything is coordinated by skills. This skill looks at Granola, Slack, Gmail, you name it, and helps surface a bunch of social ideas. He instructs his AI to keep finding more

[0:20] work to take on. Review all the work we've done in the last week. Based on what you observe, tell me new ideas for automations, workflows, et cetera. he uses are in the description of this video, so you can use them. Let's do a screen share so you can see how it all works. You're going to demo for us how your AI chief of staff helped make you more productive, and all these tools that we're using we're going to be able to share with the audience. Let's take a

[0:40] look. So, this is Cursor. Cursor is a coding agent like Claude code or code X. Uh but I don't do any coding in here. I'm not an engineer. My agent does some coding for me, uh but I don't. I just mostly talk to it like you would talk to Claude or chat GPT or something like that, but the cool thing about a coding agent is it can code for you and do stuff for you. And if you want to build a AI chief of staff, the chief of staff needs to be able to do

[1:05] things. So, that's why I use Cursor. Okay. And everything we're showing here could work in code X, Claude, and so on. Why don't we start with um you said this whole thing happened this post because of this process that you have. I have a person I collaborate with for social media. And every time we would have a weekly check-in, he'd say, "Okay,

[1:26] Wade, what ideas do you have for me?" And I don't think about social media all day every day. I think about running Zapier all day every day. And so, every time he would ask me that prompt I would kind of freeze. I'd be like, "Oh, crap. Like, I'm not prepared. I haven't thought about this. What ideas do I actually have?" And you know, I would inevitably like you know, pop open my Granola notes, I'd pop open like Slack, I'd be like just scrolling through a bunch of stuff trying to just jog my memory on like what what did I like was there any interesting things I

[1:57] talked about? Is there any cool hooks? Um, and one thing I realized was well, hey, I I don't have to go do that myself. Like, I have Cursor that can help me with this. I have the Zapier SDK that it can help me with this. So, why don't I I have it go um use the Zapier SDK to connect to my Granola, to connect to my Slack, to connect to my email, and I want it to go loop loop over all the conversations that I've had in the last week and tell me like, did it was there any interesting ideas that could be good for uh social? And so,

[2:30] uh that led to uh this weekly social idea generator here. And so, this is now a skill, but effectively what this skill does is exactly that. It goes and looks at uh Granola, Slack, uh Gmail, um you name it, and helps surface uh a bunch of social ideas. And actually, let me just uh pause this real quick. One thing I forgot to do here is I want to do this in demo mode. Uh Help me demo social media

[3:04] ideas from my last week. So, let's see what it comes up with here. And for everyone listening, we will give you the demo mode skill Wade uses it when he shows people how he's using different tools on his screen, so it could scrub out private data and we won't have to go back in and and hide it

[3:19] and people have loved it. Totally. And you know, this uh AI Chief of Staff idea was proposed by this. It noticed uh it basically was looking over all the work I'd done every single week and was like, hey, here's some uh you know, things that you should consider. There's one I actually just published this week. We We have an open job wreck right now. That's for uh senior software engineers. And one of the things our recruiting team figured out was that 80% of the applicants were

[3:47] fraudulent. And when I heard that number, I was like, what the heck? Uh and that is just like perfect social media fire cuz I'm like like it's unbelievable that it's that many people. And not only are they fraudulent, people are inventing fake profiles. So like they will find your face on LinkedIn and they will take your face, they'll take your name, they'll remix it with somebody else's stuff, and they'll present it as like Andrew Warner, the senior software engineer with whole different things. I'm like, "What's

[4:19] going on here?" This is wild. Okay, so you were internally talking about it in a meeting, you gasped the way that you just did right here, it was recorded in your meeting note taker, and then at the start of the week when you were looking for social media post ideas, this skill that we're looking at here, which is part of your AI um chief of staff, um saw it and said, "Okay, you should write a post on it." And that's how that one was written. And now you're running it, and let's see what are some of the

[4:44] ideas based on your past meetings. All right, so here are some of the ideas. Uh idea one, only works when it becomes an operating rhythm. Um let's see what it's saying. Most teams are stuck in AI demos. The demo looks great, the workflow does not change, the gap. Uh I'm paying attention here right now. Can I Can AI do the task? Uh but did the team change how work gets assigned, reviewed, measured, and approved? Is that like that's a pretty good one. Uh good you know, institutional AI versus

[5:11] individual AI. That's a decent idea. And that's for Ryan to write up. So this is not a post that you're going to go with. It's just here, Ryan, this is an idea and why I think it would work. Yeah, exactly. Like none of this stuff gets presented in like a publishable state. All right, and we'll give this to viewers and they could adjust it to their own way of doing things. Or more and more, what I'm seeing people do is they give this to their agent and they say, "Based on how I work, what can I use from this? How can I improve my

[5:38] workflow with this?" Now, speaking of, the way that this came about is you have this weekly thing that you do where you find automations like this for your life, how did this come from that? And what is this weekly automations finder? So, to me, once you start working this way, like if you start working with a, you know, a coding agent or something like this, um you will start to one, you'll start to work with it all the time, but you might run out of ideas for like, what cool new stuff can I build? Um it's a common thing I find people getting stuck with. But, one of my

[6:14] favorite prompts is this one. Which I'm going to just copy and paste it into here real quick. It says, uh review all the work we've done in the last week. Review my activities in Gmail, Granola, Slack, Coda, etc. And based on what you observe, tell me new ideas for automations, workflows, etc. Help me, you know, uh be, let's see, more efficient, do higher quality work, or faster. Uh And so, you can just go

[6:46] kick this off. And the power of this one is it's a meta agent. So, literally, you've got this thing now that's just going to coach you on how to do your job better. But, more than just coach you, it can just be like, Wade, you're you know, in in the in the case of this social media example, it noticed, hey, every time you're meeting with Ryan, you're scrambling and asking

[7:09] me to help you come up with ideas. Uh You don't have to be scrambling, Wade. I can tell when your next meeting with Ryan is. And so, the morning that you meet with Ryan, I'll just go run that for you automatically and go ahead and share that with Ryan, so you and Ryan already have that content

[7:25] prepped. It doesn't have to be this thing that's happening last minute. I saw that suggestion, and I was like, great idea. Build that for me, please. And it was like, I got you. And so, went and built it. And that's the power of this thing is it notices things that you just don't notice. I And I think this is so important like I I benefit from it a lot. I think anyone can benefit a lot from this because we so often find ourselves just on the hamster wheel where we're like, you know how to do a task, you think you're a good at doing the task, whether you should be doing

[7:55] the task or not, you may not step back to ask yourself that. But when you have an agent that can go look at all the things you've been doing and say, "Hey, wait, there's a better way. Here's the better way." Often times you just look at it and you're like, "Oh, yeah, of course. Like why didn't I Why why wasn't I doing it that way?" Uh and this is what that prompt does for me. It's the It's the accountability buddy It's the like, you know, automation inspiration buddy that's like, "Here's good ideas for for what

[8:22] you should go build." By the way, I typed all that as much as I could of what you said into just Claude that Claude chat that I have and it gave me four ideas for automations. It said, "The interview prep, it's the highest leverage. James Camp prep was 20 plus search churn that produced something genuinely useful. Chronicle uh chronolo- chronological exits, dollar amounts, the don't confuse him with the other James Camps warning and so on. You do this every week. Build a guest prep skill that takes the name plus And you know what? I have it. I just forgot and was out of the habit of using it. And so

[8:58] it's absolutely right. This was just a distracting waste of time. Slide deck uh slide deck rendering pre-flight. The gradient text bug uh Oh, it's true. Burned a real chunk of your April 30th session. It's true. For some reason every time it creates PDF, there's an issue where it downloads it with the first the first slide is completely dead and it has to churn through it four or five times and it takes forever to solve

[9:22] it. Wow, all right. Yeah. And you know, like that's a great example cuz it's it's sort of picking up on like you forgot to do it. Like you had this skill there. So, you know, the suggestion should be like, "Hey, how about I just run this for you and so you don't have to even remember to do it? Um And you know, I bet the reason you probably like Well, answer this. Like, did you forget to run it, or did you not actually think it was that valuable? And so that's why

[9:50] you didn't run it. It's a combination of things. One is I happen to just be in Claude chat more lately than VS Code. I use VS Code the way they use Cursor. And it was all in VS Code, but not in the chat. And so I was just like quickly winging it, and that was a problem. So using a new tool more than I should, or more just because it's there, and

[10:10] winging it instead of being intentional. What would you say about that? I mean, that makes sense. Uh you know, I think that I find Yeah, you know, this is one of the things I find for myself as I'm building out these these skills and these tools, is you can get addicted to doing this stuff, and you can build a skill, and then it the skill is just not actually that good, or that that

[10:31] valuable. And so that's where it does take some iterations to get this working right. Like, so if we go back to that social media one, for example, one of the things that I could do to improve upon it, is I could have it actually go fetch um you know, my last month, or my last several months of like social media posts, and then pull the analytics in from that. And I could say, "Here's examples of ideas that have performed really well. Here's an example of ideas

[10:58] that have performed poorly." Use that as a rubric to help come up with your ideas. And so if I did that, I bet my ideas that I would get out of the social media generator one would actually improve. Um and so like this is the type of thing I think that the best people are really doing is they're just like constantly like tweaking these skills, versus yeah, I think some folks will just do the first version of it, and they'll see the output, and they'll go like, "Eh, I'm better than that. Like, I don't

[11:27] really need to use this skill." Uh instead of thinking, "Actually, yeah, I might be better than this version of the skill, but like why am I better than this version of the skill?" It's like, "Well, I have this extra context that this skill doesn't have. So, how do I give the skill extra context? How can I keep coaching it to where it actually has a chance at being better than me at

[11:45] these things?" Uh and so, that's where I see the best folks are just really good at, you know, going through that loop and um you know, pushing the agent to be better than them over time. You know, this reminds me of the book that my entrepreneurship professor gave me in college, which was the E-Myth Revisited. And the story in there was an entrepreneur who builds a company, does everything himself, hires somebody to do some random thing, like put a put the product in a box and ship it out. The person he hires doesn't do just as good a job as he does. He corrects him a

[12:13] little bit, then he says, "You know what? No one can do it as well as me." And he takes it back on, and he does this with all the tasks, and pretty soon he's back to just being a solo person instead of running a bigger grown company. And I feel like you're right. I do that, and I didn't catch myself until you said it. Sometimes I don't instinct I don't naturally say, "This is not as

[12:31] good as me." But I just don't love it. It doesn't do great work, and so I move on instead of coming back and correcting it and improving it. Yeah, it's interesting. This is, you know, one of the reasons I'm bullish on managers being excellent leaders of agents is because managers have gotten good at this task. They're like really good at delegating the work and

[12:53] giving clear instructions and guidance. In fact, the person that I think is the best suited is a frontline manager who is both a good manager, but still close enough to the craft. Like they can remember how to actually do the job. They're like, "Oh, you need to do this. You need to do that. You need to do this to be what like they're a good craftsperson, but they also know how to teach. And that person probably can be the best agent builder inside your entire company. I get that completely, and I also want to become more of that. I feel like I need more guidance as like a manager of AI, as a

[13:29] manager of these systems to remind me what I'm looking for, how I need to keep correcting, and and how to get better at it. Totally. Do you do that internally at Zapier? Do you train your people like that? Uh honestly, we probably can be doing much better at this. Like, there's a lot that folks are self-discovering, and we do have like all hands where people show off like their favorite skills, and we have a shared skills library where people can see what other people have

[13:56] built. Um but yeah, I think there still is something to be said about having like workshops and things like that where you can just look over somebody's shoulder and and see what they're actually doing. Kind of like what we're doing here. Like, you can like, yeah, we've got this agent running in the background right now that's just like kind of chilling along, and you can see like, you know, here's the little progress it's making, here's what it's doing, etc. And it just the more you see somebody the way somebody work, the more you just like pick up on little tips and tricks. Um That's

[14:24] honestly like one of my favorite things. It's just like watch somebody who's better at using these tools, or who does a different job than I do, use these tools. Cuz then I'll go like, "Oh, that's cool. Like, I I didn't know you could do that." Like, um an example for me this week is I'm I made a slide deck.

[14:39] Mhm. And I usually don't make slide decks. Like, I just it's just not a part of my job. I hate it. I'm bad at it. I don't like doing it, etc. And, you know, I had to make a slide deck, and you know, this time instead of delegating it to somebody else, saying like, "Hey, can you take this thing and make a slide deck for me?" I was like, "Well, I've got my talk

[14:57] outline. I've got my talk transcript. How about I go do it?" And so, I started going messing around with all these like slide design generators, and I had one that was working pretty good, and I went to like talk to the the some folks on the team. I was like, "Hey, like, I was messing around with three of these different slide generator ideas. Like, this one I thought did the

[15:14] best. This one was really the worst." And they were like, "Have you tried Claude design?" I hadn't tried Claude design yet. I'd just totally forgotten that it didn't even exist. Yeah. And I was like, "Oh, I should go try that." And like our team had already put like the Zapier like um style guide all that stuff in it. And so I put it in there and it like spat it out and I felt like Cloud Design was tied with um I used Rep It slides for the other one. Like those two I thought were um the best of the two that I tried

[15:41] for generating slides. Anyhow, it looks like we have ourselves an audit here. Okay. So, let's see what it says. Uh you can see all the stuff it reviewed. So, I had a lot of meetings this week. This was a very heavy meeting uh week. A lot of code and 60 hours of meetings? This in the last 7 days, yeah. This is like a typical 40-hour work

[16:05] week, okay? Uh this was a nutso week. This is not This is not what I like to have. I like to have a lot more um build time Okay. for myself. But you can see it's, you know, scanned a whole bunch of different sources. Um and it's got some recurring patterns. So, same strategic narrative explained

[16:24] across multiple meetings. So, basically I'm saying the same stuff in a bunch of places. So, it's like, "Hey, you should make a one-pager for all this stuff you're saying over and over again." Uh meeting notes generate all follow-up commitments without a universal sync. Route commitments to a single work queue with owner and due date. So, I do this for myself. I think it's suggesting do

[16:43] this for others. Uh piecemeal email batches. Many similar drafts created over several turns instead of one intake. Oh, I know what this was. I was sending out invites to a new thing. Mhm. And I was just kind of doing them as they came in. I don't think this is actually like a very good suggestion. Uh task metadata gaps. Most active tasks are small, but many lack an explicit next step. Auto clean task shape before daily planning. Oh, that's kind of

[17:08] interesting. What does that mean? So, I think what it's proposing here is that it takes some of my ambiguous tasks and tries to break them down into something much more explicit for me. Um which that could be useful because then if it does a good job, I could just have it go do the task.

[17:23] Ah, okay. Yeah. And so then down here it's basically uh automation backlog like um suggestions. So it's like, you know, what what is it help with? So auto extract commitments from granola map owner date source push to Coda drafts. Yeah, this is like a no-brainer. Everyone should build one of these. This is a fantastic use case. And Coda is your project management

[17:44] software. That's what I use to you track all my to-do's, yeah. External meeting follow-up queue. After each external meeting draft email inter doc serum. Another fantastic one. So it's basically the same version of this but for external. Uh I actually already have one of these so I'm not exactly sure why suggesting this one. Um because this

[18:02] already exists. It's funny because after our last meeting that came into my inbox and then you hit reply on and you said, "This came from the automation." Um Yeah. So anyhoo, you can see like the various different things that uh it it'll generate. And then often what I will do is say, you know, you know, great. I like ideas, you know, one, two, and five. I like one, two, and five. Let's build those. Ask me questions, you know, one by one to go

[18:32] build them. And then it'll come and ask me a bunch of questions about this stuff and then I'll go kick off the agent and it'll build out all those workflows. How do you know that you're actually going to get to use it? With all these different automations, how do you remember to actually use them? So the the best ones are automations that run in the background that I'm like, "Oh, this is definitely going to help me." Like background automation versus like things I have to remember to go do. Or if it's a thing that I know takes a lot of time that I do need to

[19:06] manually do a lot. And it's like, "I really would benefit from having one one these." Um that I just it's just like pervasive. And you don't have to do it. The write as Wade skill is one that I just like use for almost everything because I just hate you know, when AI write like tries to write stuff, it just never sounds like me. And so I'm like, man, I just need it to sound more like me. And that skill it it still isn't perfect, but it like gets 90% of the way there. And so it just minimizes some of the last mile work I

[19:34] got to go do on stuff. Okay. By the way, I now have a skill that will improve my slides. It's just ready. It's been sitting here. This is great. I can't believe I never thought to do that before. It was just bothering me and I said, come on, Claude is good at this. Maybe I'm doing something wrong. All right. Um, you and I did a session like this where I said to you, I think we should give people the skill that you

[19:56] use. And you said, yeah, I think I will. I got an email later on with that follow-up with the skill I think was even included in it. And then you hit reply and you said, Andrew, this was an automation. This is the example of the thing that I'm doing. How do you do stuff like that? So let me actually share my screen on this, too. So folks can see exactly what happened on this. So today's recording

[20:16] action item, uh, great discussion today. Dan Slaegan will connect you with Austin Johnson for Claude co-work. Also connecting you with Grace with the Zapier agents. Let's get that April visit scheduled. Message from Steam. Bottom out. And then I I actually added this part. This was drafted based on my end-of-day recap plus the Zapier SDK. I just reviewed and pressed

[20:34] send. Um, and so effectively what that's doing is looking at um, my granola notes. It also has the context of who Dan is, who Austin is, who's who Grace is, like all these things. And so it knows how to go draft an email. Um, honestly, I don't know how else to explain it. Wait, this is the meeting that you and I

[20:58] just had a few minutes ago. No, that was the one from last week where I sent you the follow-up. Yeah, where I sent you that follow-up meeting after where you were like, hey, I need an introduction to these people. God, that's what I was talking about. And I thought the PS was you typing it manually. No, it's part of your automation that the PS is in there. All you did was you looked at it, you hit send. I did talk with him, the person you tell me about Claude CoWork. He was going on vacation and then he said to me, you know, Andrew, I I think the

[21:20] whole thing's going to change when I get back from vacation, but it happened like that um because of an automation. Now I see how it worked. So, just at the end of the day, it just scrubs over all those things and it's like, "Here's the emails you need to send out." Is that accurate? Getting getting AI to write in my voice even with a style guide is really hard. Is it This is even more important because you're sending messages to people you work with. Does that come out to be accurate or is it

[21:42] just a bunch of BS? 80% good, I'd say. Um so, this is why I have it generate drafts is I don't feel comfortable to have it send emails completely like me. Um but it's like 80% good. And for certain ones, like that one was like, "Yep, send." I didn't have to do anything on that one. It pretty much nailed it. Um but then there's some where I'm just like, "Nope, don't" Like it's either like it'll be overeager where it's like sending a follow-up email where it didn't need to send one and I'm like, "Nope, just delete that." Or then sometimes it'll be just like overly expressive where I'm like, "That's not

[22:14] how I talk. That's nobody talks like that." Yeah. And, you know, it has my way of writing skill, it has a skill that's supposed to scrub signs of AI writing. So, it has these things. I don't know if it sometimes like forgets to use them or if it does use them, but it still just can't quite get it quite right. I don't know. This is where like if I was I if I spent more time on it, I know I

[22:36] could get it to be a lot better. Well, it's effective enough. I mean, really. That was That was magical for me. Um all right, let's talk about Run the Morning. One of the things that people were using Open Claw for the most when it first came out was was they were just getting these morning setups. What's

[22:52] yours look like? Mine is pretty basic. I I So, if we want to do the morning brief, um, we can do that real quick. So, uh this one runs every morning now and it DMs me the um uh you know, just the agenda for the day. But, we can also do uh we can just run it from here, which is the uh daily brief. Um let's see. Well, let's see. Actually, demo let's just say, run my uh morning brief for this demo. And, you know, in fact, what it does is it just goes and hits um my calendar predominantly and says, here's all the things that you got on

[23:38] tap. It also hits Coda, which has all my to-do's and says, here is like some important to-do's for the day. It hits Gmail, so if there's any like critical emails that are sitting there from overnight where it's like, you need to you know, you here's this customer that you need to go talk to, etc.

[23:52] Um you know, you've got it. Um okay, so you can see here it you know, I have this like QA check as the first thing, which just makes sure that it actually hit the um hit calendar, hit Gmail, hit Slack, etc. Because sometimes I've in the past I found like, oh like it missed Gmail or something like that and it wouldn't tell me. And so, I'd be like, oh that stinks. And so, I have

[24:14] these like checks. And then uh it you know, sends me just like a quick little heads-up on here's what's going on for the day. So, uh you know, three meaningful blocks, exact same, customer prospects, strategy review. Uh you got this context switching, here's your must-do list. So, these are things I got to get done. You know, watch out for these things. So, it's always like kind of trying to coach me to be like, you might get distracted on this, that, or the other. So, try try

[24:37] not to. Um and you know, here's a suggestion for how you should should should open your day. So, do you actually use this? I mean, do you need a morning brief like this? Wouldn't you rather just quickly scan through Gmail and calendar and so on? You know, it's interesting you say that. Like, I think the morning brief is overrated. But, it is a good first um agent to go build because it's fairly low stakes. If it's not that good, it doesn't really matter. I much prefer the daily recap to the daily brief. The daily recap goes and sifts through here's all the meetings you were in today, here's all

[25:09] the action items that were really important. Um do you want me to go do these things for you, etc. And so I find that way more valuable than the prep one for the day. do that? And then it will say, "Do you want me to go do these things for you?" It it has like it basically does do them now. So draft that email for you. It'll go like uh it creates this little outbox of things where I can copy and paste stuff and go like, "Hey, this should go here,

[25:32] that should go here, etc." Um let's let's go do demo mode. Uh run my daily recap. And that that's the magical part that it's not just here's what you need to get done today, here's what's been going on, it's do you want me to do it for you? And that's the part I haven't gotten connected yet. By the way, all this is using uh the Zapier SDK. Why

[25:58] Zapier SDK instead of direct connection? I like the SDK for a couple of reasons. One, I think it's really good for a user because you're you have a safe mechanism for uh connecting to all these tools. If you if the service doesn't have a native connector, it'll often say, "Hey, I'll just use the API to go do that. Copy and paste the API key right here." And now you got plain text API keys hanging out,

[26:23] which is a security no-no. Mhm. Uh so you definitely don't want to do that. I also like using the SDK because I can switch back and forth between Claude code or Cursor or Codex like really really easily versus connecting them all in one service and then I have to reconnect them somewhere else and I have to reconnect other place and I just like to switch back and forth through a

[26:41] bunch of different tools to try them. Then I think the other reason it's really handy is if you work at a company, the company can then um you know, set up policies around, you know, who can connect to what, and it can have better observability of this sort of keep the um you know, sort of better manage risk. So, they can say, "This this group of people has access to this tools. This group of people have access to this

[27:04] tool. These agents can connect to this. They can can can send email. These agents can write it you know, can can read email." And that kind of like permissioning policy stuff is like so important at any organization because these agents are really powerful and you don't want them just running willy-nilly. We had a situation in our company where somebody had deleted all the tasks on our Notion task board and we couldn't figure out who it was. But, we're all trying different apps and tools and agents and they don't identify themselves well enough. And so, we couldn't figure out who even is connecting this. Is this a

[27:37] live connection or is it something that we connected a while back when we were playing with something and stopped using it? All righty. So, here we go. We've got the uh we've got the recap here. So, redacted for demo. Um strongest themes uh reliability and incident response. Uh AI agent strategy kept coming up across org design discussions, resource allocation, customer prospect carry forward. You know, so, follow up with exact team member. This will end up in my outbox. Review this. This will end up in my to-do. Send or preserve notes from tonight's customer prospect dinner. So, this is a thing I got to go remember to

[28:09] do tonight. And then uh then it's just basically, "Hey, you got some high-leverage weekly task." Before I got uh um Travel. I I got some travel vacation next week that I'm doing. What's the What's the part where it's going to do the work for you? So, basically uh these will the emails it goes in drafts

[28:29] for me. Let me see the email that it's drafting. So, to Slack. Uh it will I Sorry to interrupt, but you're saying the follow-up with this executive team member, it's doing for you on this timeline thing that you need to carry for someone else in some initiative. It's It's obviously it's hiding it because this is demo mode, but it's

[28:46] doing this follow up for you. Well, so for Slack related ones, I have them draft them for me and then I just copy and paste into Slack. I could have it send them directly. I haven't done that yet because you know, I mentioned the Gmail one was like 80% good. The Slack one is more like 60% good.

[29:07] Okay. And so I find that I'm doing a lot more editing there for the internal Slack ones. But the way we do it is it put it right here in a way that you can copy and paste out to the team. So generally what it does is it generates a a live recap for me that I can go just copy and paste from the doc where I need

[29:26] to go for them. And then the send or preserve notes from tonight's customer prospect dinner, that is that something that it would do for you automatically? Well, so this it's going to just put in my to-do list for me. Okay. All right. We didn't get anything that it would do for you automatically here. Give me just a an example of something that ordinarily would get done for you

[29:48] automatically. So like if you and I were chatting here and it said, "Hey, you know, if you had said, 'Wade, I really want you to have I really want to have this skill to to share.'" And I said, "Sure, Andrew, I'll go ahead and send that skill over to you later tonight." It would have caught that in the transcript and been like, "Great, let's go tee up an email for Wade to send to Andrew that has that

[30:08] skill." Let's do that right now. Let's just say that everything that is a skill here that can be shared, we're going to give out to the Are you running Granola on Riverside, which is what we're using to record? Uh Granola, let me just check. Oh, no. Is Granola not running on this? It doesn't always kick up on this.

[30:24] it's not running on this one. Ah, bummer. Okay, I will get you I will get you a transcript then so that it can read it and it can send it out to us and do the action for us. Everything will be here for people to try. It's going to be in the description below and I'm excited about this. Frankly, the problem that I have is that I'm not remembering to do this. I'm not

[30:43] thinking to do this often enough. Um it's one last thing. Is there a way to get whatever agent I'm using to keep reminding me that I should be doing this? That if it if like I'm using the same if I'm having the same problem with a PDF that it should say to me, "Hey, why don't you create a skill to stop

[30:58] this?" Well, so this is where I come back to that second demo where it's like look over all the things that we've done every single week. Right. And uh propose ideas for me. And so, if you just have that one running on a schedule for you once a week, once a day, whatever, it's going to catch that stuff and say, "Hey, you need to build a

[31:15] PDF chip reader or whatever." All right. Right on. I'm going to be using that. And that skill by the way will work now in Claude, the desktop app in um VS Code which I use in Cursor which you use. Thanks. Let Let me know everyone what you use. See you in the comments and bye everyone. Hey, my agent says that if you watch this far, you're going to want to subscribe. And Google thinks that if you watch this far, you'll want

[31:35] to watch that video.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt