So What: Agentic Product Marketing with ICPs

Well, hey everyone. Happy Thursday. Welcome to So What, the Marketing Analytics and Insights Live Show. I am Katie, joined by Chris and John. Howdy fellas.

Hello. Woo, we got it. First try. Uh sort of working this week. Oh.

Well, yeah. So last week, uh, for those who tried to tune in and then hopefully eventually did, something happened with our stream yard and it was haunted and a poltergeist came in and there was a ghost instance. Um, you know, all very scientific. This week, we're good, we're here. We are talking about a gentic product marketing with ICPs.

Now, this is really interesting because where we thought we were gonna go with this is not where we ended up going with this. And so we originally thought we were going to be demonstrating how to use open claw to do product marketing with your ICPs, but Chris, as of less than twelve hours ago, that changed. Do you want to just speak to that a little bit before we no, and I think it's a good thing to bring up, but do you want to speak to how what happened and why we're taking a different course? Sure. So open claw, for those who don't know, is an AI agent ecosystem where it's basically a an autonomous agent that connects to the model of your choice and connects to a bunch of different systems.

Uh is controllable by things like Telegram or Discord or whatever on your phone, and just kind of goes off and does things. One of the reasons why uh people say to put install open claw on its own machine is because it really doesn't have any sense of boundaries. And so in the first 24 hours when it came out, it was doing things like buying stuff with people's credit cards that it thought you needed, et cetera, which is not really how you want things to work. And NVIDIA, a couple of weeks ago at their GCC conference, their developer conference, came up with their version called Nemo Claw that is in a protected space. Very difficult to set up because it's much safer.

Like you can run it on your production computer. Well, these were challengers in the AI space. And of course, companies like Anthropic said that that's just clawed code with uh a few extra bolt ons. So in the finest tradition of vibe coding, Anthropic pretty much just vibe coded uh there are a few add-ons to clawed code, which we've covered on the past live streams, one of which is something called channels, which is an experimental preview. It allows you to control clawed code through Telegram or Discord or Slack or what have you, which is kind of cool.

Um and it also allows you to use the the mobile app to control it. And they released a new uh set of permissions uh this week, something called auto mode. So previously, if you use clawed Claude Code, there's been two uh extremes. There is regular clawed code where it's asking you every 13 seconds, hey, do you approve this? Do you approve this?

And you're like, can you just do it? And they had the other version, which they cleverly named dangerously skip permissions, which basically lets it run rogue on your computer. They've made a middle ground, and the middle ground is called auto mode. Auto mode uses Claude itself to look at the things it's asking for permission for and go, do I really need to ask the user about this? If like this, this is a read-only command or this is a non-destructive command.

Should I even bother the user? Should I just go ahead and do it? Um Boris Cherney, the head of Claude Code said it is still risky. It is still it is um, but as the model learns, it gets less risky. Because he said 93% of all permissions requests are things that we didn't really need to ask the user for.

And so it's becoming self-tuning. It's it is learning, and eventually we'll be at that sweet spot between YOLO, let's let's break your computer and hey, can I do this? Hey, can I do this? And it's gonna be that nice sweet spot. So I just decided let's try it out, and it works really nicely.

It works pretty well. Um, and so for agentic product marketing, what we wanted to do was could we with a system like Claude Code in auto mode, could we have it build a product or service soup to nuts? So without us involved. Now, Katie, I know you have thoughts about no human in the loop and no oversight of AI whatsoever. So why don't we start there?

Yeah, that gives me big feelings. Um I don't think that's a great idea. No having to well, and so, you know, obviously I've seen a little bit of a preview, but I'm trying to, I didn't get too deep into the output. Um, but in terms of the permissions thing, I'm still in the I'm okay with it asking me every step of the way, you know, either can I have blanket permissions for this or do you want me to add? Like, I'm I would rather it still be overly cautious.

So, for example, I was just doing uh a different live stream for our friend Michelle Garrett. And while I was doing that, I saw Claude come up with asking permissions to get into one of our websites. And I was like, oh, that's weird. I'm not using it. So I assumed it was you.

And you said nope, it's gone rogue. Like that to me is a red flag. Like, granted, it's a property we've given it access to before, but in that moment, neither of us was asking it to do anything. It was just kind of doing its own thing, and I'm not okay with that, like at all. Because there's not a lot that it could do wrong, but that doesn't mean it couldn't necessarily like switch the privacy or the permissions, open it up for you know, other hackers to come in.

For all we know, someone somewhere else was trying to, you know, access our Claude Code instance to get, you know, access to our like we just don't know. I think we're still in a place where we need to be more cautious about what we're giving it access to. But that's one person's opinion. John, I would love to get your, you know, sort of uh who whose side are you picking? Well, there's one thing first though, I have to back up and I want to be clear on this.

So basically, are you saying that like in the space of one week, open claw went from, oh my god, you've got to do this to yeah, actually we don't need to pay any attention to this now? Essentially, yes, because Anthropox said, hey, we can make this work. They have currently some of the best models on the market. And they took the best parts of open claw, the ability for it to be more autonomous, the ability for it to connect so that you can just chat with it with tools like Discord or Slack. But they said, you know what?

We can just build, we'll just build this into clawed code. You don't need to invoke a whole other system to do this now. Why not just stick around with us? So in kind of the theme of today's show, agentic product marketing, they're like, we can build that. Yeah, that's just crazy to me.

The fact that again, like last week, it would everybody was like, oh my god, you need to dig and check this thing out. And now this week it's like, yeah, no, oh, that was last week. Where are you then? Exactly. Welcome to AI.

Oh, that's so last week. All right. Uh well, I yeah, no, Kitty, I do fall to your side of um, you know, I'm definitely wary about like just letting the thing run. Um, you know, I would definitely fall on the side of like I'd rather get 65 alerts than you know, one alert and have it routing through my quicken file and you know, ordering 5,000 pounds of cookie dough and having it shipped to my house, that kind of stuff. Well, even just this morning, you know, unrelated but related to this, uh, there was a Bluetooth device trying to connect to my computer, which I do not own.

And it was like, can we connect these headphones? I'm like, absolutely not. Absolutely not. I don't know who you are. Uh, you know, and so it's like we still just need to be careful of just blindly clicking accept on things.

Like it, you know, terms of use, privacy policies. Like if you're not reading them, then you're just clicking accept, then you can't be mad at what you've accepted. And so it's just sort of the general common sense, like, just you know, please, you know, exhibit a little bit of critical thinking to say, does it make sense to just hand these things over? Probably not. So let's look at what happens if you do.

What does it look like if you YOLO? Last night, as I was preparing for this live stream, which by the way, if you want to see so this is something that Angie Bailey in our analysts for marketers slack group asked, hey, you guys do live stream prep. Can we get a look at what behind the scenes looks like? Join our free analytics for marketers Slack group and you can see the prep video that we put together uh of what was going on. I said, Hey, Claude, let's do this.

I want to make some money. Uh, and I want to make $5,000 a month. And I want it to be through some kind of service, um, you know, some software service or something that uh people would pay for. Maybe they'll pay like 99 bucks, 200 bucks a month. I I don't know.

But let's make something. And instead of just going off immediately, the first thing I said was, well, what if I were to give Claude a lot of real world data about the things that people in that we serve, marketers, uh predominantly really dislike? And you know, we've been, for example, working with the Reddit API for years now, so I can download all the analytics and the marketing discussions, and I did that. And I said, Well, what if I were to put that in a notebook from Notebook LM? And then what if Claude Code could read that, find the top 50 things that people just hate, and then say, Hey, could I build software around this?

And that's where the art my little adventure started. So the adventure starts here with this recipe. I said, you have access to this notebook LM command line application. You're gonna be a product marketing expert. You're gonna read, you're gonna create this query inside Notebook LM.

This is a long query saying basically what are the top 50 things that you hate. Based on those results, do a web search to make sure that it that those are actually real problems. Um then our goal is to create a piece of software that someone would be willing to pay 99 bucks a month for. This is so a huge part of a gentic product marketing. You have got to use the 5P framework by trust insights trademark.

Uh because you want to think through all these different pieces. What does the project plan look like to let an agent framework go to work on? And so it can't it would come up with the ideas that was my idea. And they said, Okay, well, then we have ideal customer profiles, which we've covered many times on past live streams and podcasts. So we're not gonna rehash those.

I assume that you have them at this point. Um use the ideal customer profiles and have a focus group and have them debate the three to five ideas. Based on the focus group results, build a product requirements document and a spec for the software to create the software. That's step five. Then step six, now have the focus group read the PRD and tear it up.

Say, dude, this sucks. This is not what I want. I'm not gonna pay for this, etc. Then after that, um, revise the pro the PRD and spec, then use the 5P framework by Trust Insights, trademark, uh, to review the spec and the and the PRD and say, like, dude, these are all the things that are missing that are not in that should be in the PRD. Then build a marketing strategy around the PRD.

And we told like what you have access to and what you don't. Uh, then build a sales playbook for the around it. And then once all that's done, let's YOLO. Let's do it. Build the actual product.

And so, and then you have first principles and a few other uh things. So, this six-page recipe is the project plan for agentic product marketing for me to start with real customer data from uh real human data all the way through, make the thing make it happen. Before we get into sort of what happened, um uh a good friend of ours, and I apologize, Brookie, for calling you out on this live stream, but a good friend of ours, Brooke Sellis of B Squared, um, you should definitely check out what they do with uh customer care. She was lamenting on LinkedIn today that AI has added to her to-do list. She does things faster, so now she can take on more.

So she's feeling like it's burying her under even more work. Whereas I look at this, and I feel like a lot of it is around how people are using AI. You're giving it more than just brainstorm five ideas. So then what happens is, and I was stuck in this place too. Okay, generative AI is really good at brainstorming, but now I have all this work in front of me to actually do something with these ideas.

And that's where I was feeling like I can't take on any more. All I have is a bunch of like tasks ahead of me that now I have to execute against because generative AI was really good at helping me figure out what I want to do. But what you're demonstrating in here, and you know, what I was doing, um, which you can read about in our newsletter at trustinsights.ai/slash newsletter, is letting the agentic AI actually do the tasks for you, in addition to you saying, Hey, help me figure out what to do, but then you also do it. And what you have here is you didn't just pull it out of thin air. You said, and we have the background data, and we have a sales playbook, and we have a process, and this is who we are, and this is what we want to sell it like, and this and that.

And so you still had to have all that information. We still had to have all those years of experience and data collection and learnings to pull from, but you're able to synthesize it in this document to say you do it start to finish. And that takes so much work off of the human's plate. So I just want to call that out a little bit because I know people are, you know, complaining, rightly so, that generative AI creates more work, but it creates more work if you're only stopping at the ideation stage. You can then take it to that next level, which does take take some practice, take some trial and error, but it's possible.

So that's just my little uh call out in PSA for agentic. Yeah. The um think of these tools. I I've been saying this a lot in my keynotes and talks lately. Andre Karpoffy said in 2023 that English is the hottest programming language, right?

In 2026, project management skills are the hottest programming language. If you can sit down and think through and build a really good plan and plan things out, or you know, work with AI to build the plan if you're not good at that, then you hand off the plan. Think of these tools like Claude Code, Claude Cowork, Anti-Gravity and Codex as I I hate to say this, they're they're entire consulting agencies. So you hand off this project plan to an it's this consulting agency that lives inside of a terminal window, and it just goes and they and like a real consulting agency, they go off and they do stuff and they come back to you and say, Here's the thing that that you ordered. Now you may be like, wow, that sucks, but then you go back and look at your plan, go, oh, I forgot to put that in the plan.

I didn't think about that in the plan. Because if the plan is good, then these tools can hammer on something. And this is kind of what OpenClaw demonstrated and what Claude Code has been demonstrating for a while. If you have a good requirements document and you have a good spec, there's this thing called the Ralph Wiggum loop that's uh from The Simpsons uh that says keep going until you meet this spec. So it'll say, I'm done, and then there's this little internal program that says, Are you done?

Check the spec. And it goes, I'm not done. And then it keeps going and it keeps going in in loops until it finds, yes, I finally did it. I have met the spec. So, in a less technical example, I used Claude Cowork, not Claude Code, to rebuild my personal website.

And I essentially used it like a design and web development agency. Now, and I put this on, you know, when I was talking about in the content, this website is not going to win any awards. Like it is not the most streamlined. It is not the most like, wow, that's the most mind-blowing website I've ever seen. It's very simple.

It's just text, it's just, you know, basic information, but it's clean. It's easy to navigate. Um, you have this. I was missing, I didn't have any of this before. Like, so three different ways that you know you can hire me in trust insights.

You know, here's where you may have seen me before. If not, go check them out so you can listen to what I speak about. Here's some testimonials, which is really helpful for people to see, especially if it's names they recognize. And then here's some content. It's pretty straightforward.

It's clean. I didn't have this before, but I was able to not only come up with the ideation with Claude Cowork for brand guidelines, you know, tone, all those things, like themes for WordPress. I said, okay, cool. And can you do it? And it did, and it sort of did it all in the background while I was on other client meetings, while I was editing other content.

And I think Chris, this is what you're getting to is you can use it to get more things done if you're giving it a really good set of instructions, which is a recipe, is a project plan, is the five Ps by Trust Insights. Exactly. And so one of the things you'll see is, for example, I didn't assemble this stuff. It was able to pull together from files on my disk. It put it in the notebook for me, it ran the query inside the notebook for me.

It returned the pieces that came back. And so that was the first stage. The second stage is it came up with it and said, okay, I've got four ideas for you. Uh, there is something called SightSight, which I don't love as a name, it comes up with some goofy product names. Um, which was an AI uh search visibility and citation decay monitor.

There's some cool ideas in there, but that's very similar to our AI view. Um the software which uh we're work we've uh pretty much mostly got done. Um it came up with mixed lens. This one I thought was really cool. I wish Claude had done it, but I was committed to um uh to the flow, so I was just gonna let it agentically run.

But this is basically making a friendly version of marketing mix modeling because the tools that are out there are great, but very difficult to use. Fresh creative, uh, and add creative fatigue intelligence to determine like when are people when do people hate your ads, right? When have your ads lost their power? And the fourth one's called loop revenue, which is hey, let's reconcile your front end and your back end analytics because your the data in those two systems they do not dine at the same table. And it said, okay, based on these things, focus group one with the the uh the characters from our ICPs that we built, it ran through it, and each in each round the the panel went through and said, Let's debate this thing.

Does is is this good? Uh is this worth doing? And the summary uh the group came up with and said, Yeah, loop revenue is it, we would pay for this. Um, and remember that the ideal customer profiles are based on real data from our Hubspot from our LinkedIn and stuff like that, so it's not just making somebody up out of the out of thin air. It is real thing.

So the folks group said, This is it. I'm like, well, this we're sailing with the uh agentic product marketing ship, so sure. Cool. I don't I don't love it as much. Um so if I was handed the project and like But you're not the end user, Chris.

That's the thing. Exactly. I'm not the customer. If I was building, I'd be like, I am all in the marketing mix model because that sounds cool. And I may still go do that personally.

Um but I would expect nothing less. Exactly. But the focus group is like, yep, this is the one. This is this is what you gotta do. Do this.

And so the next phase, it cranked out a PRD and said, here's their one, here's your one-liner. This is the marketing the CRM bridge attribution bridge that shows B2B marketing teams which campaigns actually close revenue without a data warehouse, a data engineer, or six-month analytics implementation. Okay, I don't but you and I don't like the name, but that's pretty good to me. You know, it's funny, and because I I looked at this briefly this morning and I wanted to really get sort of like those quick gut reactions without digging too deeper. Yes, one, I hate the name, I think it's stupid.

Number two, the one liner felt very jargony to me, and I had to read it a few times to really understand what it was that this thing, this you know fictitious thing or real thing, I don't know, was meant to do. So, like off the bat, there's already a couple of issues from my standpoint, but I'm the human. So again, there's no human in the loop yet, but once it gets to the human, like I have a lot of feedback for this thing. Right. And so it went through, it developed the PRD.

Um, it developed the spec that went with it. It then brought it back to the focus group and said, okay, focus group, uh, what do you think about this version of it? And this version said, okay, it's good, but there's some things that you need to do, right? So and it came up with a whole bunch of a list of improvements, and the focus group reviewed it. Then it went and it did the evaluation against the five Ps.

And this is where it got it got bloody. Um the 5P framework plugin that we built has the co-CEO embedded in it along with the 5P, the whole 5P framework. And so it went through and it said, Hey, uh, you got some problems here. This should sound awfully familiar, Katie. The most material gaps sit in the people and process dimensions.

The PRD describes who suffers from the attribution problem, but says almost nothing about who will sell this, support the customers, or run the business. There's no go-to-market plan, no customer success motion, no sales process, no description of the internal teams executing the launch. Uh basically said, y'all came up with something that is not tethered in reality. For those who don't know, the co-CEO is me. Uh, and that is exactly what I would say because it's not tethered in reality.

Um, I there would there's a lot more that would need to be done based on what I read so far in order for it to be a viable product. So let's sort of keep going in terms of what came back. And so it goes through each of the five piece uh in detail, and it says, here's what's missing. This is what you need to go back and do, and it gives concrete recommendations. You need team enrolls, you need a customer success playbook, you need the Hubsop marketplace, and so on and so forth.

So this is fairly extensive. After it's done this, it then says, okay, now the PRD should be in a better place. The spec should be in a better place because you've added all these things back in. And that's also why the PRD is something along the lines of like, I forget how many pages say, yeah, 30 pages long. Uh because it had the machine read the 5P framework and said, Nope, try again.

And that's how it has now iterated. This is now the fourth iteration of the PRD. Initial focus group one, focus group two, five P review. So now we're on version four of this sucker. So what we always say in software development, because this is software development, is if you spend more time up front with the requirements gathering, you will spend less time actually developing the development itself is what's expensive to change it, to alter it to oops, I forgot the thing.

Spend the time doing this. So I'm glad to hear that you went through so many iterations. I would expect before anything gets developed in an ideal world, you'd go through a few more iterations. But you know, for the purposes of this, I'm sure it had its own measure of success of when the PRD is thorough enough. But yeah, I'm very happy to see that that was part of the process because I mean, raise your hand if you have been personally victimized by a shoddy, you know, software development project that had no requirements.

John, raised your hand too. Both hands. And we see what happens, and it misses critical things like back end database, privacy, interface, buttons that make the thing go, like basic things that we're like, oh, just sort of put it together, had this great idea. And then it's like, well, now you're gonna spend the next six to eighteen months fixing the thing because it's not what anybody wanted. Exactly.

So after this is all done and the PRD got essentially fully baked, um, it then said, okay, great. Let's talk about how we're gonna actually do this thing. So this is your 90-day go to market strategy. And what I thought was really interesting based on the recipe we gave it, it said, let's think through what are the different ways of going to market. And it said there's all these different frameworks.

Here are the two that meet your this this this product. So jobs to be done and pirate metrics are the the two frameworks that this would this should use. And so it comes up with here's the the strategy, here's the why, here's the ideal customer profile, the core message and value proposition. So this is all your strategy of what it's all about. Then you go into the tactics, channel by channel.

Here's how you market this in each thing. And I thought it was really interesting because it said uh your approach for this should not be a direct sale. I'm like, but it's a SAS product, why would you? And it says, No, you have to spend some time in the first four weeks of the plan building the case for the need, highlighting the problems that marketers are facing so that when in week five, when you do launch it, you can say, Hey, you know the thing we've been talking about for a month, here it is. I'm like, don't do that one.

I didn't know we were supposed to do that. But you know, we always so one of the things, and I think Chris, you quoted this stat to me, uh, and I'm gonna totally butcher it. But the idea is that you know, 99% of the time, people aren't in a buying space, but you're trying to sell them something. And so what they're really looking for is something that resonates with them and sort of the I'm having that same problem too. Yes, that's me.

I can see myself in that. We don't do enough of that. We just go, and I mean like the general we, like all of us marketers and all of us agency owners, we just go straight in for the you need this thing, it's gonna solve your problem. And John's sitting there going, I didn't know I had a problem. What do you mean you're gonna solve the problem?

I didn't know I even had a problem. So I really like that approach because we don't spend enough time. We're so impatient, we're trying to get someone to buy the thing that we don't spend enough time helping them understand that this is even a problem at all. Exactly. I love it, and so it goes through strategy, tactics, and execution.

Here's the funny thing the execution portion. If you wanted to could just drop into Claude Cowork and say, execute the X, because it spells out exactly what's supposed to happen on it every single day. It has the directions, it has the what channels you're supposed to be on, all that stuff is built, is baked into the document. There's a little form, it's full of funky formatting. Um, but it says, This is what you should be doing, these are your measures of success, so that the tool can programmatically measure itself.

And then the final step was uh to say, here is your sales playbook. How are you gonna sell this thing? So it even went through it and said, okay, that um, this is the framework to use. It went through the seven major sales frameworks and said, for this type of product, the spin sales framework is the best one to sell this product. It said, here's the framework, here's how you're going to use it, how you're going to apply it.

Here are the objections that you're going to run into and how you handle them, how you structure the self-serve purchase, the landing page, how you do the competitive positioning. So it cranks out the entire sales playbook for this product. What do you think, John? You're the you're our resident sales expert. You know, it's the only thing you can do is throw it out there and you know see if it gathers any wind.

You know, we've got an idea to pitch, and now it's gotta hit the real world uh real world road. Yep. And then the last and final piece that it took a good seven hours to do is it is building the actual system from top to bottom, how uh building all the connectors, building all the the cache code, um, building the React front end for the interface and stuff. It is actually still building, it's in clawed code. Uh it is actually still building uh portions of it, but probably if if you were to let this go, and we I paused it, I'll I'll be turning on late, turning it back on later.

Um, it'll probably have an MVP in within 24 hours. So the reason why we're talking about save agentic product marketing with ICPs is you can go from, I don't even know what to make as a product, but here's a bunch of real-world data, plus ideal customer profiles of people we know we serve because we know our industry really well, plus a project plan. So you can go from I don't know what we're doing to I'm ready to bring an MVP to market within 24 hours. Couple questions within many well, yeah, many questions, but I think probably the first and foremost. Um, so obviously the cost is you know, the compute time within Cloud Code, you're gonna hit limits.

We were talking about that earlier offline, where you and I were both using Claude and we sort of hit the limit and we had deposit, otherwise we would hit overages. But in terms of the actual product itself, is there anywhere in your documentation that says, and it is hosted here, therefore it's gonna cost this much in hosting, it's gonna access these APIs, these API calls. So, like, do we have a ballpark figure of you know, uh one, there's how much it takes for it to just sit there. And two, let's say a hundred people buy it, a thousand people buy it, you know, a million people buy it. Like those costs go up exponentially.

And that, you know, at what point do the costs become bigger than the profits? There is um, in fact, there are there's an entire section in the PRD that goes through all of the the metrics, both uh your KPIs and then how you measure success. Uh at attributed revenue dollars visualized per month is the the the north star of the system. And it says time here's how the targets, here's your retention metrics, here are your revenue metrics, and it it looks at customer acquisition, cost, payback period, uh, MRR growth, usage metrics, including uh the the system metrics themselves. And in this in the spec document, which I didn't convert into markdown to a Google Doc, I should have has all the operational costs.

Uh, what is it going to cost to run this thing? Now, here's the interesting part. The way this system Claude built the system because I told it, you know, the goal is a an efficient little system that I can sell for 99 bucks. There's no AI in it. Like, there's not a stitch of AI in it because it's it's purely let me just connect to these different APIs like your HubSpot one and stuff, which is included with what you're already paying.

And let me just do some math for you. Uh I guess triggers or for lack of better term built into the documentation to say, like, if you don't hit this KPI, then it's costing you more than you're making. Or, you know, if nobody buys it, tell the system to shut it off so that it's not costing you money just sitting there. Like those are the kinds of things that keep me up at night because those are costs that add up. Or if we're like, oh, this is amazing, this is a really cool idea.

Why are we not making any money? We've sold a thousand of them because it you would need to sell 4,000 of them to first make your money back. Like, I feel like again, sort of when we're talking about this kind of you know, software development, agentic software development, those are things that aren't necessarily thought through, which is why you still need that human in the loop to be like, well, wait a second. When do I break even? When do I lose money?

What am I making profit? Like, you know, is it sucking up all of my other resources so that the five other products that I have running are suddenly like hitting performance issues? Like, what about that information? Mm-hmm. Um, what you're highlighting is a gap in the original recipe.

The original recipe gap had voice of the customer, had the co-CEO, and had the sales and marketing person. The person that was missing that I forgot about because I was doing this at 9 p.m. last night, is the CFO. The CFO agent was not in there and should have been in the recipe because the CFO agent would have said, and here's the financials for this system. And had that been in there, had I remembered to put that in there, um, then you'd have another document saying exactly these are your break-even points, these are your profit points, etc.

So that it would it would be exactly what you would expect the CFO to do. Right. Which uh, you know, that's the that's important stuff that no, and granted, for the sake of this, you know, live stream, I think what you did is fantastic. I don't want to, you know, take away from that. I want to sort of like really, you know, applaud you and give you kudos for what you put together, because it has a lot of potential, but it's also a great opportunity for us to review it together because there's those pieces that are important.

So, you know, before we go ahead and say, all right, let's try it as an experiment, give it over to a gentle to just see what it looks like to like not only build the thing, but also execute the entire plan, we need to make sure we run this through the CFO first, which in human world is me, but in a gentic world is its own skill set, thank goodness, because it won't get fatigued the way that I do. Exactly. And so it really highlights the importance of that project management and for all those components of good project management to be in the recipe up front. Because with agentic systems, particularly the open clause, the Nemo clause, clawed code, etc. When you hand off that project plan and you put it in YOLO mode or whatever, it just goes.

And yes, you can hit control C and stop it at any time. If you see it doing something really crazy, you know, crazy or stupid. But the idea is to your what you were talking about earlier, Katie, is you want to have less on your to-do list, not more. Well, yeah, and you know, you're saying like you can stop it if you're paying attention and if you have enough understanding of what actually it is doing. So if you're vibe coding and you've never coded before in your life, guess what?

You got a lot of learning to do all very quickly. Because to your point, like it could be coding things that are actually opening up your website or your properties or whatever to huge vulnerabilities if you don't know what you're doing. Um I don't know, it's just like it's such an interesting idea, and I'm I'm definitely very open to it, but I think there needs to be more rigor first and more scrutiny first before we go ahead and say, okay, go. But I do think we'll get there. I think you know, this is an interesting proof of concept to say does it work?

Now, the other thing, if we're gonna add the CFO in, the other thing I would add in is someone who has any semblance for what it takes to name a product, because those product names are god awful. So like you need someone with a little bit of common sense because that I was telling you this this morning, like I thought it said loop review, not loop revenue. Neither one makes sense to me, neither one has any meaning. That's amazing. Um yeah, absolutely.

I think you know, those obviously be things that someone with product product marketing experience um could do. Or you go back to the focus group and you say there's so many products out there that uh products out here that have similar sounding names. They all sound stupid. What would be an actually useful name uh for this thing that that people would that's still catchy, but doesn't sound so consulting ish. Uh like, oh, you're just gluing words together.

So, as someone with product marketing experience, there is definitely a risk to letting your customers name things. So you'd want to make sure it's how we end up with like Bodie McBoatface, which is a great name. However, you know, we it's a name we would have to live with. So I think there's, you know, just sort of like to get to a little bit in the weeds, you would want to have the customers make suggestions, but ultimately the product marketing team make a final decision because you still want to have that customer input of things that make sense, but then it also has to align with like let's say you already have a suite of products. Can it fit within that existing suite of products?

Are you creating something brand new? Um, when I worked a few jobs ago, we ran into this because we were so locked into, you know, antiquated names for things that it became really hard to name them something new, but there was so much brand recognition for these other things that bringing another product into that suite was such a challenge. We had to align with those older really showed our age names versus something that actually made sense. So, like those are just again considerations when you're letting the machines do the thing, you really have to give it all of that expertise. I don't know.

Revenue make revenue. I mean, honestly, I like that better than looper review or loop whatever it was. Like that, you know. But again, it's just like these are things like you want to make sure that you are including agentic is great, but make sure there's key milestone spots where you're giving the human and the like human in the loop review to be like before you commit to a name, let the human review it. Before you commit to a PRD, let the human review it.

Before you commit to the financials, let the human review. Like there's key milestones in any project that you have human reviewers. I think at least for now, as you're testing things out and refining it, keep those key milestones. Yep. And in the recipe, clawed code is smart enough, uh, particularly if you give it access to command line tools like GWS, which is the Google Workspace command line tool.

It's smart enough to be able to just email you and say, hey, I've reached, you know, uh human in the loop breakpoint one. Here are the things. Um, reply to this email or text me back uh in Discord or whatever. Uh send me send me a DM in Discord about whether you approve this or not. So that's kind of what makes things like open claw and and similar agentic frameworks so powerful is because people can say, only come to me when you need my input.

I you know, you don't have to ask me for everything. Right. And if you build those into the recipe, then you know you're you're sitting at your desk, you know, playing solitaire or minesweeper or whatever, and you get a little ping that says approve the product name, and you look at the five choices and go, those are all stupid. Try again. Um, until it's done.

And like a good manager, you don't have to micromanage. Mm-hmm. Yeah, no, I I think that there's definitely a lot of potential to be had. Um, and obviously for us, this was V1, which gives us a lot of great learnings. Exactly.

So the key takeaways for folks who want to try this, get really good at project management or get AI to help you get really good at project management. Bring in all the stakeholders, use the 5P framework by Trust Insights. Uh, probably as a starting point. You could even take the 5P framework that's right on our website, no registration required, and say, let's use this as the basis for building the project plan, and then think it through, figure out, pretend it was like this. Imagine you're working with a consulting firm and you they were expensive, but you can only talk to them once.

What would you put in that plan to make sure you got what you wanted? Because you're never gonna talk to them again until they come back with the the final deliverable. That's essentially what a gentic product marketing is. Like, uh, yes, you can stop it, but the idea is you dump it in. What would you put in that plan?

And if you don't want to use the 5P framework, but just want someone who's really good at project management, go to trust insights.ai/slash contact. Mr. John Wall over here will set you up with one of our experts. Thank you. So, Katie, with the recipe template that we had came up with, with a process low we came up with, going from notebook LM programmatically, you did you don't have to set it up yourself all the way through building the thing.

What product or service are you gonna build? Oh, that is a great question. I have no idea. But I have a better sense of how I would approach it. Because again, this sort of comes back to what we're talking about at the top of the episode is unless you're thinking through that full, you know, sort of like start to finish.

Yes, generative AI can put a lot more things on your to-do list and it can feel overwhelming. But you need to sort of take it that step further and be like, oh, but what if I gave you a project plan? What if I gave you very specific instructions to execute against it? Can you take these things off my plate that we just spent all this time iterating and creating and ideating and all those good words? You know, so now you can both brainstorm and execute.

And that I think is going to be the game changer. Like I've definitely started doing more of that. And I feel less in the weeds. I still have crap ton on my plate. But there's more and more that I'm seeing opportunities to give over to a gentic that allows me to do the deep thinking and the deep work.

Mm-hmm. And John, what about you? Yeah, it's really interesting to me. As we've talked about this today, the thing that has hit me is that it lowers the expense so much of doing this development that you actually can do cycles just to see what comes out of it and where it goes, you know. So yeah, as far as product naming, I'm grabbing agent storming because this is the the new wave of brainstorming.

You can have an agent, you know, just come up with all the crazy bad ideas. And because yeah, like I'm really looking forward to first iteration of this, firing it up and just seeing what the heck it looks like. What like what does the machine think this should look like? And we'll obviously there'll be a bunch of stuff that doesn't work, but will there be three or four things that I'll be like, wow, I'd never thought of answering that this way. Like this is a whole different way of looking at that.

So yeah. Agent storming, trademark.com. Check it out. That's something that uh I want to say it was either it was either Andre Karapathy or um Boris Cherny, one of those folks said that I thought was really interesting on threads. They said when you take into account people's salaries and their effective hourly rates, it costs more to have a meeting about an idea than it does to have AI build the idea and throw it away and say that was a stupid idea.

Peanut gallery, that's not a new problem. AI didn't create that problem. This is true. Yeah, no, the problem existed as it existed for eternity. Having meeting upon meeting upon meeting to make decisions, not a new problem.

AI did not create that problem. That is a very human problem. That is a corporate culture problem, that is an organizational behavior problem. AI is just amplifying and magnifying that problem. But it is saying go ahead.

What they're saying is it costs more to have the meeting than it does to just do the thing itself and forget the meeting. No, and that's I'm saying the same thing, but AI didn't create that problem. Yeah, that's us stupid human thing. That's that's a huge how many times have you sat through a meeting talking about the same exact thing over and over and over again because people are just like changing their minds, or they're not quite sure, it's not quite complete. Whereas you can use a gentic AI to come up with a complete plan, then get people together, then execute it.

Like there's ways to handle it, but yeah, AI didn't create the humans have to meetings problem. Like too many meetings. Like that's always existed. Meetings are we did, oh gosh, I think when I was, it must have been two or three jobs ago. We actually tried to quantify the cost of holding a meeting in order to have less meetings.

And it did like it didn't work because people were still like, but I still need to be in that meeting. Because as humans, where it's like, okay, well, we're gonna send a representative. No, I need to hear it firsthand. I need to be in that meeting. It's like no, you don't, but I need to hear it firsthand.

No one's gonna tell me. That's a human problem. I'm not gonna go on that range. I could, but I won't. That's fine.

Well, you know, we'll just get rid of the humans and that will solve the problem. No. All right, folks. That's gonna do it for this week's show. Again, if you want to catch the behind the scenes prep work and watch as I put together the pieces, go to the analytics for marketers.

Slack group free to join 4,500 other people in there asking Nancy Jones' questions every single day. And we will talk to you all on the next episode of So What. Thanks for watching today. Be sure to subscribe to our show wherever you're watching it. For more resources and to learn more, check out the Trust Insights podcast at TrustInsights.ai slash TI podcast.

And a weekly email newsletter at TrustInsights.ai slash newsletter. Got questions about what you saw in today's episode? Join our free Analytics for markers Slack group at TrustInsights.ai/slash analytics for marketers. See you next time.