Loading summary
Transcript61 lines
- [00:00]
A
In AI today, skills and building skill files are one of the most important things you can do to up level the results you're getting from AI. And today we're giving you a $20,000 AI transformation consultant in the form of a skill. We're going to walk you through how to use it, how to get the most out of it, and bonus, at the end of the show, we're actually going to give you a master skill template that's going to make it really easy for you to build any new skills you want for yourself or for your team. If you want the files we talked about and can click the link below, we'll get them right to you. Let's get to today's show. Here's a quick word from HubSpot. HubSpot helped Tumblr solve a big problem. They needed to move fast to produce trending content. But their marketing team was stuck waiting on engineers to code every single email campaign. Now they use HubSpot's customer platform to email real time trending content to millions of users in just seconds. The impact, three times more engagement, double the content creation. Want to move faster like Tumblr? Visit HubSpot.com. Okay, so Kieran, recently on the show we talked about some new data from Anthropic that there's this big gap in what's possible with AI technology and what people are actually doing. We kind of teased that we wanted to help people solve that. And, and I'm here to say that I have spent probably 10 hours and I don't even know how many credits of compute to try to come up with the best solution to that problem. And I want to go through that with you in real time today, get your feedback and then it'll be available in the link below for everybody to go and try this out themselves. But we've been talking a lot about on the show skills and how skills are a real unlock of how you use modern AI assistants and agents. And, and I've spent a lot of time building what I call an AI transformation skill that helps anybody take the work they're doing and transform it to an AI first way of doing it. And you kind of talked about this. I think you have a dream scenario of like how this should work. So tell me what you think the dream scenario is and let's, let's see if I can deliver on it.
- [02:16]
B
Yeah, I think the way this is going to work for most teams is that you can record yourself just doing work and narrating your work and you can give that to an agent, and that agent will build skills that help you replicate the kind of workflows you're doing on a day to day basis. Then it's the easiest way for the average person.
- [02:36]
A
Okay, I agree with that. So here's what I did just for the sake of the show, Kieran. I've posted in here a transcript. It's two minutes long. I did a very short video recording of myself walking through something I have to do all the time, which is review and synthesize information in slides, collate that feedback, and then send that feedback in an email. Something you do all the time. Something I do all the time. Something I think everybody watching the show probably does all the time. Right, right. So that's what you're looking at here is literally just a transcript and what I've done. And this is in Claude. I've attached the skill file as well as the examples and reference data. So if you're doing this, you just literally get like a zip file, a folder of these markdown files and you can just upload them to Claude, to chat G, to perplexity, whatever you're using. Right. And you can tell it to use these skills. So it's literally, as you're seeing, that's all I've done. Right. I literally said run the skill against this transcript.
- [03:34]
B
So I have not seen this done this way before. You're just adding the files right in the chat and telling that to use this stuff in the chat.
- [03:41]
A
So the way I'm showing this done right now, Kieran, is I would say we call the basic of basic ways you can go and install this skill in Claude. But what I've done is I've made a fairly complex skill where I have a skill file and I have different reference files that all come together in one folder. And so if you're using Claude cowork or Claude code or Perplexity Computer, it's easy to just upload and access that whole file. So for the purpose of this, I'm trying to do it the basic way, which is like you just upload the files in that folder to a chat with whatever LLM you're doing and you put the task that you want to do it. This is not like, oh, I'm going to do this a hundred times. I have it perfectly installed in Claude. This is a. I'm not sure what I'm doing and I want to get started really quickly and really simply. Does that make sense? Right.
- [04:28]
B
Yeah.
- [04:29]
A
Okay, so that's where we're at here. That's all we're looking at. And so I have the skill file, I have the reference files and I have a transcript. One of the things that I think is really important, Kieran, is I've scoped this skill to understand the type of work the person's asking to do. And I basically have it. Is this a task transformation, a deeper like individual level transformation, a team level transformation, or a department level transformation? And so you'll see here that I have it change the output based on the type of work it is. In this case this is a task, so I have a much shorter output. I have a two to four page output for versus like if it's a team level transformation, you could have like a 40 page output. So it's basically reviewing back to the skill in real time to understand what type of work it is and scope the support that it needs to give. So what's really important the next thing here is that it's very flexible depending upon the type and scope of the work you're doing. Okay, then I also have it checking for current tool pricing to make sure that it's giving you like reasonable, accurate pricing in the recommendations. It then goes and goes through all the steps and writes the report. And here is the output of the AI task transformation plan. We're giving you an amazing skill that you can use to transform any work you're doing to be AI. First you can just talk to it, type it, upload work files, record what you're doing and it will tell you how to do it in a better AI first way, what it's going to cost, how long it's going to take and how to actually do it. And you're going to stick around to the end of the show. We're going to give you a bonus skill as well. You can find everything in the link below. Just give that a click and you can get those files. So this is an output that you're going to get and no matter what you put through this skill, you're going to get a report like this, which is basically what the heck am I trying to do? Here's the task that I'm currently doing. It understands the tools that are currently used to do this task and the current pain points of doing the task. Then it gives you a before and after of what are you doing currently and how long it takes versus using AI how long they think it's going to take. And so this is open deck. Manually hold evaluation criteria in your head while assessing each item. Right. And then it's showing you in this case, what Claude could do in that same way. And it basically then is showing you how it's thinking and working through the problem and how much time it thinks. It's saying, hey, if you do this I as a CMO am we're going to recover one hour to two and a half hours of time per week is its estimate. And it's going to tell you if it's just time savings or if it thinks it has an improvement in quality in addition to like the time and efficiency. It's going to give you tools recommendation. This is like the most basic task. So it just has Claude as a recommended tool in here. It's not like a very complex multi step task and then it gives you the next step to go and try and build it out for the week and long term. So literally you can just record yourself and it will tell you how to do what you're doing in an AI first way.
- [07:41]
C
Thoughts?
- [07:42]
B
So it's not building new skills, it's recommending like a workflow. An AI workflow.
- [07:46]
A
It is recommending an AI workflow either for a human to do with an assistant, which is this example, or if you're talking about a much more complex example, it will show you actually, you know what agents you should build the skills you need. All of the more in depth answers. Like for example, if you have a problem that you want to just run through, we can also put a more complex task in here and kind of go through and see what's happening, what it recommends.
- [08:12]
B
So in the report, can you bring up the AI workflow, What did it recommend?
- [08:17]
A
Okay, this is the AI transformed workflow.
- [08:20]
B
Yeah.
- [08:20]
A
So this is saying this is what you're currently doing and it takes 25 to 45 minutes. This is AI assisted, takes 8 to 15 minutes. And this is the steps it's telling you to follow specifically. And get this is a very basic task. This is literally just trying to show people that you can do this with anything.
- [08:37]
B
When does it get the current time from that 25 to 45 minutes. How does it calculate that number?
- [08:44]
A
Yeah, it's estimated because I go through in my voiceover of it and say what I'm doing.
- [08:50]
B
Okay.
- [08:51]
A
It's like I'm going through each slide manually and so it's an estimated time right. Now obviously if you are going through and just doing something in real time, Right. If you just record an hour long video of yourself doing something then it's going to know it takes you an hour.
- [09:04]
B
Yeah, yeah, yeah, yeah.
- [09:06]
A
So here's a different Example Kieran. But I took one of Rich's looms, like the three hour one.
- [09:11]
B
Yeah.
- [09:11]
A
Where he's just working in Claude code, and I put that in and it did a full breakdown of what he was doing and, and the technology. So like it says it's like, hey, it's three hours because that's how long the video is. And it breaks down all the current steps he's doing to do data analysis, Right?
- [09:29]
B
Yep.
- [09:30]
A
And then it's actually telling him all the things from an infrastructure standpoint he should do to make it faster and better. So like run project, bootstrap script, create directories, import docs, converts to md. Like basically automate a lot of the upfront work to basically change his workflow. Right. And you could literally then just take anything from this report and put it into Claude Perplexity Chat, GPT, what have you and have it build you a more detailed plan. Right. Like you can keep building on any of this. This is meant to basically give you an initial breakdown of like you're doing this thing and you could do it in a different and better way.
- [10:09]
B
Yeah.
- [10:11]
C
Before we continue, let me tell you about a podcast I love Marketing School. Marketing legends Neil Patel and Eric Su bring you daily actionable digital marketing lessons learned from years in the trenches. Marketing School delivers bite sized marketing wisdom you can implement immediately whether you have a new website or you're an established business. You'll learn the latest in SEO, content marketing, social media, email marketing, conversion optimization and and general online marketing strategies that work. Today. They just did a great episode called 10 SEO Lessons that Still Work in the AI Era. Listen to marketing school wherever you get your podcasts.
- [10:54]
A
What don't you like about it? What would you do differently?
- [10:57]
B
I think it's really good. I think when you look at it, it's quite overwhelming the amount of information for a single workflow. Like having a whole report for a single workflow is a lot.
- [11:06]
A
Right.
- [11:06]
B
You're having to read like a four to six page report on how I could do slides better.
- [11:12]
A
Well, you don't have to read it. You could literally just upload that report to Claude and have it give you the prompts of what you need to do. Right. You're creating like a more detailed schema of what the work is really.
- [11:23]
B
Yeah.
- [11:23]
A
I also made it that way because there are a lot of people watching who have to deal with very conservative companies and bosses and just like they're going to need that level of not just detail, but simplicity of detail of like you have to do these Things and do it this specific way, and this is how we think it's going to work so that people are not, like, freaked out.
- [11:42]
B
Yeah. I think I would kill myself if I was in a company where I was getting reports on how to automate my work.
- [11:48]
A
Well, I guess what I'm saying is I don't think I would ever actually read all these reports, but I would use them as inputs.
- [11:55]
B
Yeah.
- [11:55]
A
To then build the agents and build the workflows and build the skills and everything that you need.
- [12:02]
B
Right. Yeah. I think the ability to upload a transcript and get ideas and how you can do that better is valuable for all people at right now. Everyone's trying to work more efficiently. Everyone's trying to work faster. I think this is a good way to kick off some lumen with your teams and get some ideas on how they can go faster.
- [12:22]
A
There are a lot of teams that have very complex workflows. Right. It's like, oh, I'm using, you know, Canva and HubSpot and G Drive and Snowflake and all these things. Right. And actually, one of the things that I learned from you very early on in this AI journey is that most teams are not clear on, like, how the work they do actually happens. Right, Right. And so part of what this is, and I think why you need a more detailed output, is so that everybody's actually clear on, like, this is how the work we do right now actually happens.
- [12:52]
B
Yeah.
- [12:52]
A
And do we agree that that's how it actually happens? And if so, then can we agree on a new and better way to do it?
- [12:59]
B
Yeah, yeah, yeah. It's like having a little work analyst.
- [13:03]
A
Exactly. And again, you can click the link below if you want to check out this skill. The other thing I think we're going to put in there, Kieran, is in part in making this. I made a separate skill that's a meta skill for making new skills. And I know you're like, well, all of the platforms have, like, skill builders, right?
- [13:21]
B
Yeah.
- [13:22]
A
The instruction I gave it was to go and crawl all of the current AI skill directories out there. So, like, what are the most popular skills, the most used skills, and how are they set up? Because, reminder, everybody, when we're talking about these skill files, you are limited to 500 lines in one of these files. That's how long that file can be. And so the challenge of what you're trying to do is pack the most density possible into 500 lines. And so I went out and I actually did a huge bunch of research within perplexity computer in this case and basically learned from GitHub, HashiCorp a bunch of other places what skills were performing, took the best practices and basically consolidated that into one skill template that you can now use to build any skills. Which I think is pretty cool.
- [14:12]
B
That's pretty cool. Yeah. So that is a skill builder.
- [14:15]
A
Yeah, you can see. So this is a skill architect. It builds remarkable agent skills that are discoverable, concise, effective across all major platforms. So it also like makes sure that they work on everything. When to use the skill, create a new skill from scratch for any domain, improve or refactor an existing skill, design a collection of related skills, a skill family. I've been working on a couple skill families, decide between a single file versus multi file skill architecture. So a lot of the stuff like you don't want to take the time to think of this crap, right? No, but it's one of those things where it's like, it's that it's compliant, that it has the right architecture, that it's actually improving what you're already doing. This is something that if you have just you can run a pass on anything you're doing and it's just going to make it a little bit better. And that little bit better is really impactful, especially if it's you're working on building a skill that you're going to use a whole bunch.
- [15:05]
B
Yeah, this is really good. I think this is going to be really helpful for people, which is you can just install this skill on your system of choice and then ask it to build a skill mimicking this template.
- [15:17]
A
Yeah. And it's just going to follow a really good template. It has the right decision trees for architecture and how it should be organized and how it should handle skill families. If it needs a skill family. Right. What I just showed you on the AI transformation was a skill family and folder that it had essentially made. Because I used this skill to iterate and improve on that last skill I showed you, the AI transformation skill was made a lot better by running it through that skill builder, which was super cool.
- [15:43]
B
Yeah, I think that's really cool.
- [15:45]
A
Yeah, I think there's a couple of things that you need to do when you are just working on building in this AI world. And some of it is just like having some meta level processes and skills. And so a skill to build skills I think is something that should be in anybody's toolbox 100%. Okay. Now when it comes to AI transformation, we talked about this skill. We talked about the easiest thing to do is Literally just record yourself on screen, what you're doing, use the AI transformation skill to go through it after that's happened. Kieran, you're like, hey, that report is really overwhelming. What would you do with the synthesized data of that report? Like, if you were recording yourself doing something for an hour, like, how would you then go and use that to make it super actionable for you? You get this really detailed readout of what all that work is.
- [16:34]
B
It depends what you're trying to do in terms of your AI transformation. There's a multitude of different things. So if you're trying to get folks to use a certain tool, let's say you wanted to get all your team using CLAUDE or all your team using OpenAI, then you would skew all that towards what are all the things that they could have done in ChatGPT or Claude that they're doing manually right now? If you really want to get your team using skills in whatever platform they're using, then you would say, well, use this to create skills that teams can share to automate this workflow. Or if you want to start to make AI part of your employee onboarding, you could create onboarding docs of how we work and that would actually really help. I think it's going to be very nuanced or different depending upon what you are trying to get your team to do. I think it does start with trying to understand ways we work today and then how AI can start to help. I think the big part that AI is going to struggle to help with is behavioral change.
- [17:29]
A
Yeah, talk about that.
- [17:30]
B
I think a lot of folks can kind of maybe know what to do, but changing your behaviors and habits is probably the thing that humans are struggling with. Most like, you kind of just default to the way you do things because that's the way you do things.
- [17:42]
A
And what have you seen working there in the. The work that you're doing that's helping people change, or is it really hard?
- [17:50]
B
Yeah, I think it's really hard.
- [17:51]
A
Right.
- [17:51]
B
I think where it's changing faster is actually where people are doing things that they weren't able to do before, like people actually coding where they were never able to code before. People do an analysis where they were never able to do analysis before. People like doing the new things, like applying AI to do new things that they were not able to do before. I think it's a little less exciting to use AI to do something you already do, even though it could help there. It's like shiny object syndrome. Yeah. I think starting small as well don't try to get overwhelmed and do all the things. Just try to use it in a certain number of small ways and make them part of your habits and then evolve from there.
- [18:25]
A
The framework I've been using here, Kieran, is I think in a pre AI world, the average person was more likely to be like, lower agency, so less dependent, less just trying things, less building things, and higher tolerance, which is like, ah, I gotta go to this meeting, or, yeah, this thing isn't that great, but I don't have time to do it any better. And I think all the people that we're seeing make this transformation are the exact opposite. They're super high agency, they're curious, they're out there building. They're like, hey, I couldn't do this before, so now I want to go and do it because, like, now I can. And that's awesome. And now they're also now very low tolerance. They're like, hey, this stuff stinks. And I can make it way better. And I can make it way better in, like, minutes or hours, not days or weeks. And, like, those two things combined are really, really powerful. I would encourage anybody to, like, look around at who you're working with, and I suspect the best people you're working with fall into that, like, high agency, low tolerance bucket.
- [19:22]
B
Yeah, Right. People that have high agency and just want to make things better.
- [19:26]
A
Yeah. This is a time for winners. Transforming with AI is very hard. I think, Kieran, you and I believe it is one of the biggest business opportunities in the world is just transforming to a more AI first method of working. We've tried today to package up why that's important, how to actually do it. Gave you a couple of skills to go and hopefully do that a little bit faster and a little bit better. And we'll be with you real soon. On the next episode, Marketing against the grain. Hey, everyone. You know, Kieran and I have been doing the podcast for a while now. We've been at this for a couple years. We love it. We could not be happier to be doing this, but we wanted to take things to the next level. We want to level up the impact we're having with Marketing against the Grain. So the next step of our journey is something we're really, really excited about. We're going to launch the Marketing it's the Grain newsletter. And Marketing against the Grain newsletter is going to be amazing. If you are a marketing leader, practitioner, you're in the trenches doing marketing every day. This is for you. We're going to deliver right to your email inbox and you're going to get all the behind the scenes frameworks, practices, tutorials from us, from guests we have on the show, and from people even beyond the podcast that we think are going to be helpful and really have an impact on your day to day, week to week doing marketing. You're going to love it. It is something we've been talking about for a while. We're really excited to have it out in the world. We've already got a hundred thousand marketers who are on this newsletter. Please join. It's completely free. We'd love to have you as part of the Marketing against the Grain community. And it's easy. You can click the link in the description below, or you can head to marketing against the grain.com subscribe.