Loading summary
Transcript66 lines
- [00:00]
A
Hey, we all thought getting people to use AI was going to be the hard part. It turns out that we were completely wrong. The harder problem is getting them to use AI correctly. We're now flooded with AI slop. One person pasting something into Claude and then sending that to the other person. It's becoming an epidemic in your company and every company out there. Today we're joined by Hilary Gridley. She built an amazing system at WOOP and turned all this AI slop into repeatable quality. She's showing us how to build real tools today that can fix this problem. You want to stick around? She's taking us behind the scenes. This is going to up level the quality of AI work within your company dramatically. Let's get to today's show. Hillary, welcome to the show. Today we're really excited to have you on marketing at the Grain.
- [00:43]
B
Thank you. I'm so excited to be here.
- [00:44]
A
We're talking about a fun topic that I think literally applies to every single person in the world and especially every single marketer or builder in the world. And that topic is we've all started to use AI and, and AI has become a big part of our lives and now we are flooded with AI slop. I'm on a group chat and this is senior leaders at all like brands that you all have heard of and know is like, I feel like so much of my day is reading people basically typing something into Claude. Taking that response from Claude goes to another person and Claude's summarizing it for the other person and it's like going back and forth. And basically the tagline of this channel chat is like, people are outsourcing their taste and their judgment to a decision making to AI and it's creating this huge wave of crap and slop out there.
- [01:35]
B
Yes. And we have to stop the slop. We cannot let this tidal wave of slop wash us all away because we will, we will not be happy campers if that happens. Yeah.
- [01:44]
A
And so you were obviously you were a builder at whoop. You saw this firsthand. You were, you were here through kind of the, this first wave of AI which is like, oh, we all need to start using AI now. We're past that and we actually need to actually get into a place where it's like, oh, we're using AI to scale our taste and judgment, not to outsource our taste and judgment. And so maybe like from the front lines, take us behind the scenes. What did you see and what have you seen that is like the big reasons that this problem Exists, Sure.
- [02:15]
B
So I've seen this from a few vantage points, as you said. I saw it as I was leading teams at WOOP and I was working to get adoption in the early days of we could see people working differently with AI. We were trying to get this out into coding teams, into non technical teams across the entire organization and figure out how to do it in the right way. And then from that I actually started teaching a course called how to be a Super Manager with AI. And so through that course I have talked with hundreds of managers at other companies who have also been on this trajectory. And I have seen how the pain points that managers and that independent contributors have have changed over the past two years. So when I started this course, I was still at Woop, and this was in February of 2025. And the most common thing I was getting asked was how do we get people to actually use these tools? And this was something that at woop, I think we did a very good job of. And we had some really smart folks on the internal AI that were really focused on finding power users, getting them equipped with the tools and then letting that sort of trickle out to the rest of the organization. And so that's what a lot of people were doing. And it was really great. It was fun. You had these AI super users who were really just supercharging the way they worked and the way that their peers worked. And then I noticed something started to shift, which is, wait, everybody's kind of doing this in their own way.
- [03:46]
A
Problem one, it's all decentralized and completely decentralized. Nobody knows if people are using this stuff the right way or the wrong way. I totally agree.
- [03:53]
B
Exactly. And it kind of has to be in some ways because you need the domain experts to be the ones who are actually figuring out how to do this work. Right. You need a marketing expert to figure out how you make a really cool AI powered marketing workflow. And just pulling in somebody from an engineering team who has no domain expertise around marketing is not going to do a very good job of that. So it kind of has to start that way. But everyone's doing their own thing. They're all pulling from different contexts, right? They're all using different sources of truth. Some of them are outdated, some of them are conflicting with each other. Some people are connecting their tools, some people are not connecting their tools. So that's problem two is there's no kind of canon, right? There's no central brain. Although many companies are kind of attempting to do this, it's easier said than done. And then the third problem is nobody is really articulating what good looks like here. And this is the problem that I care the most about as somebody who's led high performing teams, as somebody who teaches people how to lead high performing teams in this AI era, is you really have to be clear about what good looks like. And if you are not setting that quality bar as a leader or as a manager, you can't be surprised when that quality starts slipping. And when people find out that there's an easy button that they can press to do their job for them. And maybe it does like an okay job and maybe in some cases okay is fine, but then you notice it starts kind of seeping into more and more things. And that's when I think you get this dynamic that you were just talking about where you kind of have this feeling that you're just like talking to a person who's just an intermediary for Claude, where it's like, I could go talk to Claude if I wanted to, like I'm talking to you. And I think that that is something that I'm hearing constantly from managers now is like, what are we going to do about this?
- [05:39]
A
And so if you're interested in everything we covered today, you're going to want to scan the QR code, click the link in the description below and we've got a ton of resources to take you to the next level. I have felt that what you've said deeply and I've talked to a lot of other leaders who have also experienced what you're talking about. There's a few things I would add. Like, my hottest take here is that you had two types of managers in the pre AI world. You had managers who were deep craftspeople and that might be a amazing email marketer, a great creative director, might be a great UX designer or pm. And then you had what I'd call context carriers, that their whole job was to move information around. And like they didn't really have deep craft expertise, but they were kind of a human API. They were a glue that could move context around. And AI is kind of obsoleted. That second group of people, and there's still a lot of those groups of people across all these companies. And because companies are in the midst of making these transitions and I find that, that, that the slop is more prevalent with the context carriers because they can't do exactly what you just said. They don't know the standards and how to set them and they're not deep enough in the technology. And the application of the technology to actually teach their teams what they're good at is moving information around. And AI makes it really easy to move information around. And that's what creates all the slop is you get in this information regurgitation on like an endless cycle.
- [07:12]
B
I think that's exactly right. And I would even say for the information carriers and the context carriers, there is still a role for people like that to play. And you can be really good at it. Right. There are ways to be not very effective at it. And I think a lot of times, like a lot of these problems come down to as a manager, if you're not really sure what you do or what value you bring, then that is going to be made very apparent in the AI era if it was not before. Because if you're just saying to people like, hey, run this by Claude, what value are you bringing to that conversation? Right? But I do think for somebody who is like a context thinker, who has a genius for that, it is easier said than done for a company to figure out what context is real and to keep it updated. And I think, especially now, like this maybe gets solved in six months. I don't know. But I think what you're saying about the people who are real craftspeople, they have a completely different opportunity, which is, how can I shape these tools so that they are making my team better at what they do? Even if you took the tools away tomorrow? And this is, again, what I teach in my class is like, you want to build these tools in such a way where they are teaching your team how to do things, how to do a good job and what good looks like. But they are teaching it in such a way that if you kicked the crutch out tomorrow, they wouldn't be like, oh, no, I haven't learned anything. And I think that is why it's so important for leaders to be really involved in these implementation questions of how we design these AI systems, because there's so many choices to make along the way. What does the human do, what does the AI do and how do they interface? That really do have an impact on the cognitive ability of your team and whether your teams are getting better. And if you can make teams that get better, those teams will make the systems get better, which then make the people get better. And it creates this virtuous cycle as opposed to if you just have this cognitive rot happening where everyone's outsourcing their judgment to the AI, nobody's questioning anything that's coming out of it, the people get worse, that makes the systems worse. And then you get into this sort of slop doom loop.
- [09:16]
A
Yeah. Once the slop starts, it's really hard to break that slop doom loop. As you're saying. I have a couple of things I want to follow up there. But before you use the word context many times, and context is a word that's getting thrown around by everybody out in the world, me included. I would love for you to define for everybody watching today, when you say context, what do you mean?
- [09:37]
B
I mean it in its most kind of pure sense, which is what is the information that a person or an AI agent needs in order to do a job? Well, if I go to a new hire and I say plan this launch for this feature that we have coming out, if that's all the context they have, they're not going to be able to do a good job. But if I tell them, here's everything you need to know about the feature, and by the way, here's everything you need to know about launches that we've run in the past. Here's what worked, here's what didn't work, here's what we tried with that context. A smart person can really build on that and do a really excellent job planning this launch. But me as the manager, how I think about the information that I give to them in order to do that, right? Like, you don't want to give them way, way, way, way, way too much. If I say, like, here's every A B test that we have ever run in the history of this company, knock yourself out, right? Like, that's not helpful. But likewise, it is not helpful to give them no information. And so how you decide what information you give them is actually like a core challenge of being a good manager. And that, as I said, is true whether you're managing humans or whether you're managing agents, because an AI agent is going to have the same problem.
- [10:52]
A
I think there's a third level to it and it's the thing that I believe the most in, which is it's the clarity and decision making foundation that leads to all the work that you need to get done. If you show me a team that's really executing really well in general, they're very focused and they're super clear on what the strategy is, what their priorities are, who they're serving, how they're adding value to them, and it's easy to just lose sight of all those things and just ask Claude and have a hundred different people asking Claude a hundred different ways what the answers to those questions are, you know, Kieran and I have a book coming out called Loop, it's coming out in September. And one of the ideas we have in the book is what we call a taste profile. And it is all your context that you need to build your marketing. And it is essentially all of the stuff that isn't in an ideal customer profile, non demographic data. So like what that customer believes, what they feel, what pushes them too far, all of that. Because that is actually like the instructions that AI needs. They need a deep emotional understanding of the customer. And the other part of it is all the stories you as a brand, you as a company are trying to tell. You know, like what is the core product narrative you have, what is the core emotional brand story, the thing you want people to feel. And if everybody on your team has access to that same taste profile in this case or same context file and what you're saying, the quality of work automatically goes up, the consistency automatically goes up. And you don't have this slop cycle because you're going to reduce so much of that because the judgment making by both AI and the people gets a lot more focused.
- [12:35]
B
And what I think is so funny about that. So I totally agree. I think that's amazing. I love that concept. And what is so funny to me about it is even if we had no AI, if you as a marketing leader took the time to, to assemble something like that and disseminate it to your people, everything would improve immediately, everything would improve dramatically.
- [12:55]
A
Right. And that, and it has happened in some ways before, like with creative briefs and there, there are other artifacts of this. But now in the new world you actually need more depth. And I think that, you know, slop is a function of a lack of depth is my argument. Right. Like you don't have contacts, you don't have depth in what you're trying to do. And so you stay at this surface level. I find that the people who are really great at working with AI have a ton of beliefs and point of view and they'll go back and iterate a hundred times on something. It's not just asking Claude to spit out something one off. But I know you have a bunch of lessons that you have learned and advice for folks who are in this slop cycle and how to get out of it. And I think, I think you got a couple things you want to share with the audience today. So I want to kick it over to you to kind of kick us off in the solution stage of if you are on a team, you're in a company, you founded a company, and you're just like, I'm just surrounded by endless words and images that are kind of sloppy and crappy and don't actually have any value. What the heck do you do? Sure, when a buyer asks AI for a solution like yours, do you know if you're showing up or not? At HubSpot, we've got a great product that actually helps you get discovered in AI search. HubSpot AEO helps you show up in these moments with the right answers buyers are looking for before the first click, before the first fill. That's the moment HubSpot AEO is built for. Go check out HubSpot AEO. You can trial it for free. And it's going to help you understand how you're showing up in AI Search and how you can show up even more.
- [14:32]
B
The broad theme of what I'm going to talk through is that you have to make it super clear for people what good looks like. And that shows up in a few different ways. I'm going to show a few different tools and ways to think about doing that. But if you take nothing else away from this, I think that is the most important thing. What does good look like? What does good work look like? What does a good version of this job look like in an AI era? And if you don't have clarity on that, you cannot expect anyone on your team to have clarity on that. And if they don't have clarity on that, you can't expect them to meet that bar. And so much like you were saying, if you take the time and effort to codify all the context that people need to do a good job, whether AI has access to that or not, people will become better at their jobs for having that. It's the same with this. I laugh with all this AI stuff
- [15:22]
A
because it's like just good leadership and management.
- [15:25]
B
Exactly. Yeah. It's like you need to bring the clarity for expectations and for quality. And if you can do that and you can help people meet that bar, you're going to be happy no matter what the technology you're using.
- [15:37]
A
Can you talk about setting a quality bar real quick? Because I think humans have a hard time one setting a quality bar for themselves. And depending on the manager, sometimes they're scared to set the quality bar, and sometimes they're setting the quality bar too aggressively and too high. So maybe like, help us unpack what it means to do that.
- [15:56]
B
Well, yes. So I want to think about it on two levels, and the first level that I'll Start with is what does the job look like? What does a good version of this job look like? And I think what I mentioned early on, people were giving AI to super users, seeing how they used it and then letting that sort of trickle across the organization, finding examples of good versions of using AI and then piecing those together. And I think that was a great way to start. But at this point, I think we need to go beyond that. And in general, I like to think, where are we trying to get to? And then how do we work our way backward to where we are today to figure out how to get there, rather than how do we start from where we are today and then incrementally change things to try to get to some future state? And so I think we're past the point of starting from where we are today, imagining our jobs today, and then thinking, okay, how can we add some AI here and some AI there? Right. I already am going to write a brief. Like, can AI write that brief for me? Whatever. So instead I would urge leaders to say, like, think about a year, two years, five years in the future. Five years is maybe hard to predict. Think about two years in the future.
- [17:07]
A
I can't imagine maybe six months.
- [17:09]
C
I don't know.
- [17:10]
A
The world's moving fast.
- [17:11]
B
That's true, that's true. Every week you're like, oh, gosh, seriously, everything's changed now. So imagine a year in the future, right? Like, what if all your dreams come true and your team is totally AI pilled and they're working in an AI native way? What does a day in their life look like? And how is that different than what a day in their life is like today? And can you show that picture to them so that they have the clarity? They don't feel like they're just throwing spaghetti at the wall, trying random things with AI and seeing what works, but they actually have something they're working toward. Well, so that's the first thing that I.
- [17:44]
A
And I would just interject that I think it's really important for people to understand that that dream state where people are working in AI native way, they could also still be doing the wrong things and not being productive. Right. It's not just working in an AI native way. It is working against a strategy that's going to add value to your customers, to the business. And I think so many of you know, again, I'm on these group chats and it's like, people are doing so many cool things, but over half of those things have like, no value.
- [18:14]
B
Yes.
- [18:15]
A
And that's the like that, which is a different layer of slop. I think when we think about AI slop, we think of like regurgitated Claude text or ChatGPT text. There's a whole other level of slop of like, I built these 10 applications that nobody's ever going to see or use.
- [18:29]
B
Yes.
- [18:30]
A
Because I thought it was cool and only I thought it was cool.
- [18:32]
B
And that as, again, as a manager or as a leader, you should be thinking like all those levels. Like, the top level is like, okay, how are people spending their time? What are they doing? How are they working? Then there's a level below that, which is of the hundred projects they could be working on, what are the 10 they choose to work on? And how do we make sure those are the right ones? Then on the level below that, it's like, for any one project, what is the work that is coming out of it and what is good for that piece of work look like and how are we codifying and systematizing that in some way? And it is, it goes all the way up and down the stack. Like, you can't just focus on one piece of it. You really do have to be managing at all layers of that. So I can show you a bit about how I think about the kind of top layer. And then we're going to go down to a very granular layer for a piece of work. How do we actually think about articulating what good looks like for that?
- [19:22]
A
We should do that. Pull that up while you pull it up. I would just say, to reiterate what you're just saying, I think this AI transformation of work and getting rid of slope is not that dissimilar to any other technology that we've had. And I read this book recently called Inside the Box by David Epstein. David's coming on the show soon and he's got great stories in there about how constraints and focus help you. And that's essentially the thesis, and I think that's especially true in the AI era. But he has the whole story of how Pixar iterated and stayed laser focused to get to the Toy Story moment. Right. And I think that is essentially the exact thing that's happening in every team, is that your job as a leader is to set the constraints and focus. Because you can't have people going off and doing a hundred different things, as you were just saying. You need them actually going really deep on the one to three things that you believe are going to have the highest impact when they're done. In this new AI Native way.
- [20:21]
B
It's so true. And that's why I always kind of laugh when people say, well, is there a reason that we need managers in this AI world? And it's like on one hand, there are a lot of ineffective managers out there. We've never needed ineffective managers. And so if some cleanup of ineffective managers is happening, fine. But to your point, the scope of chaos, of what a single person can do if they are not given that clear context, clear strategy, like if they're not kind of pointed in the right direction, it is just, we're seeing this, we're seeing this play out and what kind of chaos it can create when every single person has their own mental model of here's what I should be doing, here's what matters, here's what good looks like. And there's no kind of orchestration happening between that. And so I actually think having a good manager is more important and higher impact than ever. But yes, if as a manager you're not doing any of those things, then you're only going to add to that chaos.
- [21:17]
A
This is the classic example of the more things change, the more things stay the same. You still need the great fundamentals, truly. All right, walk us through what you got here.
- [21:25]
B
Okay, so what I like to tell people is think about a situation, right, that you are going to encounter in your role, or that somebody on your team is going to encounter in your role. And then how would you map out what a person does today when they're navigating that situation like something has happened? What do they do? And then what? And then what? And then what? And then take a step back and try to reimagine what that would look like if it truly had AI at the center of it. And to have AI at the center of it. I basically think everything's happening proactively. So anything that today you kind of have to notice or be alerted to, or go in and find, imagine that signal is being served to you proactively. And then as you think about every step that happens after that, think through, okay, well, what if AI did the next thing? What if it went even further? What if it went even further? And. And so I have an example of that here. I'll show you one that I have for a marketing team, a situation, right? Your CEO has just asked, why does a competitor's new campaign sound exactly like this positioning that we have?
- [22:26]
A
Sometimes examples feel too real. This is one of them. Says everyone watching at this moment. Uh huh.
- [22:33]
B
Well, I don't know. We interviewed all the same people. I don't know how this happened, but anyway, so I think about how this works today, right, where you kind of, you pull up your competitor's site, you're sort of trying to, you're reactively figure out what happened, right? Like, oh, wait, what is their campaign? Let me go see it, let me go find their social, let me go find their website, let me piece this whole thing together. And then you kind of put this all in a doc and then you kind of look at your messaging and their messaging and you're going back and forth, you call the team together, you kind of whiteboard a bunch of ideas out. Somebody takes pictures of the whiteboard, nobody ever looks at those pictures again. At some point you write a brief, a designer to try a different spin on this. Your CEO meanwhile, is pestering you, saying, what's going on? What's going on? And you're like, ah, we're working on it, Right? Does that sound familiar?
- [23:19]
A
Maybe it sounds too familiar.
- [23:22]
B
So what would this look like, right, if nothing was a surprise and if the AI could always go one step further. So if nothing was a surprise. Right. The system has flagged this somehow. Right. Imagine you have an agent that is constantly scanning your competitors for their competitive messaging and comparing it against your own. And so it should proactively be flagging when something like this happens. But then what would it mean to take that a step further? Right. It's not just flagging for you and saying, oh, hey, we've got a problem. It's proposing three different angles, right? Each with a landing page mockup that you can preview. So it's kind of prototyped out these different responses and then you as the human can kind of use your taste, use your judgment to choose between those. And maybe it pulls in some supporting data, right? Maybe it has a conversion estimate on each, which is pulling from your site data. And so maybe you pick your favorite ones. Your system updates the landing page and then it updates all the trickle down, the ad copy, the email sequences. And all of this happens in, you know, a couple hours, maybe less. Yeah, and then it allows you to then respond to your CEO and say, hey, we're on top of it. We have a new page shipped. This is what we expect. I'll follow up in a week when we have some results from how this new page performed. And so something like this, hopefully this sort of illustrates the difference between a situation. The situations might not really change, but what happens as a result of that situation can change a lot. And so I encourage managers to try to go through an exercise like this and to really think about what are the types of challenges, problems your team is solving today. What would it look like to solve that in a more AI native way and then walk them through that, paint that picture for them and bring it to life with really concrete details, very concrete to the point of what are the data sources that this is pulling from, what is the exact next action that the AI is doing? And what is the role of the human in all of this? I think giving people something like this, it makes them, in my experience at least it gets them a little bit excited because they can see where this is going, they can see themselves in it. It reduces the fear, uncertainty and doubt of just like, I don't know, man. My manager just keeps telling me to use Claude and it doesn't seem like they have a clue. Right. It gives them confidence in you because it seems like you know what you're doing, which is important. And it gives them something to work too. Like what I have on the right here might be a little bit beyond the capabilities of your system or your organization today, but you can get there. You could get there right now if you wanted to.
- [25:50]
A
Yeah. And I think what's interesting is you, you break down the workflow, I think really well. And I would say there's also like kind of a top line constraint. One of the ideas we have in the book is like things have moved from months to minutes, you know, is this really something we believe? And if you just had a, a goal underneath your problem staying up there, which is like you need to get your CEO a prototype which he or she can review within 30 minutes of the new page. Right. That's actually what we're talking about. We're talking about solving a problem in a way that was never before possible. And the steps you have here, essentially that's what they're solving against. They're solving for this thing of like, oh, my CEOs got this crazy problem. Before it would have taken me like weeks to like get him or her something to look at. Now I can get them something like within 30 minutes. And by the way, it can't be slop because I'm not going to send a slop landing page to my CEO. Right. So it's like kind of a really good constraint and forcing function. And if you can do something like that, then you could probably almost do anything that is slop free. Right. Because that, it's a good way to show I actually know how to work in this new world in a way that's valuable to that point.
- [27:02]
B
When you are making things concrete like this, you can have real conversations about what a person's role is. And that includes, like, the accountability. Like I tell people, the job has never really been about the work. It's always been about accountability for the work. Right? In theory, you could have contracted your work out to some third party and surreptitiously had them do the work and then send that to your boss. And the only reason anyone would ever know if that happened is if the work wasn't good. And then most people would say, well, it would be impossible for this random third party to do the work secretly without having any information as good as I could, so I'm just going to do it myself. But now anyone kind of has that third party and it doesn't have to be a secret. And it turns out it's actually pretty good. But you're still accountable. You're still accountable for the quality of your work. And so you can say something like this. To your point, your job is to get a prototype to the CEO and success looks like it's a good prototype. If I look at it and I'm like, no thought has been put into this. And this is just like, clearly, AI slop that you typed in, hey, make a prototype for me, and then sent out to the CEO, you're not meeting the bar. But I see a lot of confusion too. Like, well, I told my team to use AI and now they're using it, but they're making all this bad stuff. And I think that's too focused on how to do the work, which is not actually what you want to have the conversation about. You want to have the conversation about what you're accountable for and what good
- [28:24]
A
looks like, exactly what the goals, the constraints, the timelines are. And that forces you to work in this way. Like, one of the things I believe is that I believe some, some and very few people are really good blank page people. That they're like, oh, I know I got this blank page and I can just whip up the right thing. I think people, most people are great editors, myself included. I'm a terrible blank page person, but I'm a great editor. And what I love with the CI world is I can get all these options that you outline, right? It's like, hey, I need help with this. Give me three different angles and then I can edit and move and say, oh, actually, I don't like any of those three. I like this other thing. And that is, I think, a boon and a tailwind to Anybody out there who. I think it's most people who are just really good editors, they just need to use AI in that way versus just totally outsourcing the thinking and the editing to AI itself.
- [29:17]
B
Totally. I think there's almost too much focus, especially in non technical fields around automation, which implies it starts the job, it finishes the job. And for me, the way I think about it is I still want to do the things I love doing about my job. I just don't want to be starting from zero. There are things where me starting from zero, I'm able to bring some kind of generative seed into that, which then the AI can sort of supercharge. But there's so many things we have to do over the course of the day where you just really. The cognitive energy it takes to get from 0 to 80% of it is a lot. And it's not that differentiated. Right. The differentiation comes in like the you exercising your judgment and taste to really get it over the line and get it from good to great. And so if you can have just good land on your desk and then your job becomes getting everything to great, I think that's a lot of fun. I think it's a great way to work.
- [30:10]
A
Yeah. In a world where AI slop is so prevalent now, like the remarkable, the OMG decisions and content and products get way more rewarded. And there's only come from people willing to make hard choices and having a real clear vision of what they're trying to achieve. Okay. I think you walked us through a really good perspective of how work actually should get done and how it's transformed and changed a lot. What else should people be doing to get the slop out of their own work as well as like the people they're working with.
- [30:43]
B
So the other thing that I like to teach managers about is how do you make tools that give feedback on people's work so that they can understand where it is, not meeting the bar and how to get to the bar. And part of why I think this is important is because it forces a more iterative way of working where you are doing a shot at something, you're trying it, you're getting feedback, you're improving it, you're taking another shot, you're getting feedback, you're improving it. Historically, that would be really slow to do if every single time they had to wait for your feedback. Right. It's like piling up in your inbox. And you as a manager, you just like, nobody wants to be the kind of manager that's like, everything has to go through me in order to move forward. But if, presumably, if you have some level of craft, if you have some level of expertise, that's actually quite helpful for your team. And so you want to give that to them as a way for them to make their work better, not as like a process hoop they have to jump through, but again, as a way to just keep getting better and better and better. And by the way, this is something to your point around. If you feel like you are getting slop and you're just like, it doesn't feel like you put work into this, one very simple non AI approach to that is just say to them, this is a good first start, but it still feels too AI generated, so keep going. And I think a lot of times people get stuck in this like it's either bad or it's good, right? I either did a bad job or
- [32:11]
A
I did a good job.
- [32:12]
B
And if you say this is too AI generated, it means you did a bad job, which means you did something wrong. And I think when you're kind of introducing this new way of working, which is like the AI is helping getting you further and then you have to keep going, it's not like, hey, this is bad. Like, you made slop, you're a bad person. It's like, okay, this is maybe like a first take at this. It doesn't seem like you've put a ton of thought into it. So I'm not going to put a ton of thought into it. Why don't you take another pass and sort of put your spin on this and then I'll sit down and put my spin on this. And so really working and doing that change management for teams to work in a more iterative kind of continuing to tweak and iterate and make better and better and better way of working. And so I have a lot of tools I've made for how to do this. And what I teach is basically how to take expertise that's in your head and turn that into some kind of tool. So it could be a custom GPT, which is just a very simple way to have people upload some piece of work and then get feedback on that. And the feedback is not just what ChatGPT is saying, but it's actually following a set of instructions that you have written that are based on what you evaluate when you look at a piece of work and help them understand if they've met the bar or not. But there's a lot of different ways you could do this as a skill. If you're in Copilot, you can do this as an agent, or if you're in Gemini, you can do it as a gem. So I made a tool called the Executive Editor. And basically this is a tool that anyone on my team could upload an email that they were going to send to an executive and would give them feedback on how to make that email basically more likely to just get a yes, a green light, continue doing what you're doing. Because I saw that so many times my team would put all this work into something they were working on, and then they would think of as kind of a formality, like, oh, we're going to have to change the launch date. We better give the CEO a heads up. And then they would send this email and the CEO would respond, be like, what are you talking about? We have to change the launch date. Absolutely not. Like, not acceptable.
- [34:08]
A
Literally, one of the things the CEO never wants to do is change the launch date. Just FYI, free. Free tip for everybody out there.
- [34:15]
B
Yeah. You know, I always laugh when I give this example because people are like, well, you know, is our email like, do we really need to be spending that much time working on our internal emails to each other? And I'm like, oh, have you never had your day blown up? Because, you know, you, like, thought you were going to do something that was going to be relatively straightforward. And then you're like, hey, is it cool if we move the launch date back a month? And the CEOs like, no, absolutely not.
- [34:40]
C
No, it is not cool.
- [34:41]
A
Not even a little bit.
- [34:43]
B
So I made this tool basically to help guide them through improving this. And again, this isn't something I wanted to spend my time on. I don't want to be editing people's emails for them. So. So this tool does it for them. But importantly, it is based on the principles that I bring when I'm thinking about what good executive communication and managing up looks like. This is different than if you were to just like me as a manager. Saying, hey, can you run your emails through Claude before you send them? Is like, there's no me in that equation.
- [35:09]
A
Yeah, break down for us. Like, how did you, like, what's the logic? How'd you actually build this? Somebody's like, you know what? Yeah, I don't want to look like an idiot to my CEO anymore. I want the same thing. Like, what'd you do?
- [35:20]
B
Yes. So this is a custom GPT, so somebody would paste in their draft email and then it would give them feedback, basically, like, pass, fail, and here's what to Improve. And here are some suggested rewrites to extract your standards. So what are the ways? Like, what are the things that I look for when I'm evaluating an email? Now, that's kind of hard to do, right? Like if I asked you, what are the five things that you look for when you are, you know, reading somebody's email that they're going to send to a CEO?
- [35:49]
A
Yeah, I don't think I could tell you at this moment. Yeah, but it's hard, right? Yeah, it's hard.
- [35:53]
B
And so there's some ways that AI can help with this. My favorite one is like giving it examples. So I literally had a document and in one column I had draft emails that people had sent me. And then in the other column I had my revisions and I put that document, I just uploaded it into ChatGPT and I said, what is the difference between column A and column B? What are you noticing about the edits that I make over and over and over again? And can you help me spot the patterns? AI is very good at spotting patterns. And it said, yeah, well, sure you are looking for, does it lead with the message? Like in the very first sentence, is it actionable? Is the tone right? Is it clear? Like, is every single word adding clarity rather than ambiguity? And so then I say, okay, can you turn those rules into criteria? Right? Give me five criteria and then write out in plain English, what does passing versus failing that criteria look like? And then you have a rubric and you can run anything against a rubric. And so what you do, you can say, like, write that rubric as a prompt that I can paste into a custom GPT or into a skill, whatever you want to do. It's just English, right? It's just going to make an English document. You copy it, it says, step one, evaluate the email across these criteria. Tell the user whether anything has passed or failed, and then give suggestions for how to improve. And then you just copy that, paste that in the system prompt into custom GPT and then anyone can use it. Anyone has access to your brain. And the thing that I think is important about this is I've seen people who sort of make these second brains for their team. I think it's really important to do it in a purpose driven way. You'll notice that I was not like, I'm going to make a second Hillary, which is just a whole bunch of documents about me that anyone can upload whatever they want and get feedback. I made this as specific as an editor for emails that you are sending to an executive that you need to get a yes on. And so I've made dozens of these and they're all that specific because I really, I mean again, this is sort of a silly example. It's like internal emails. But I really do care about what good looks like, whether it is, you know, a product requirements doc, whether it's even a problem statement, the design of an AB test. Like I have opinions about how you should do that, what good looks like. And so that's what I'm trying to encode in these tools to give my team basically on demand access to, to my own thinking models rather than just saying like, yeah, I don't know, go talk to Claude.
- [38:28]
A
It's kind of like the AI equivalent of the old John Wooden story where, you know, the first day of basketball practice at UCLA when John Wooden, the famous coach, he'd start with having everybody put their socks on. You know, it's like the most fundamental, basic thing, like you need to do, like you can't play basketball without socks. Right. And what you're basically saying is that there are certain touch points and every person's day and how they do work that are foundational. And one way to avoid slop, get better outcomes as well and just be more successful in your job or your career is to have little very focused tools for those moments. You showed us a really good example of them. Everybody probably has six to eight that they probably should build that will actually set that foundation and set that bar to actually make them position for success.
- [39:18]
B
And like everything else, like I get so much clarity by going through this as a manager. It's actually really good to sit down and ask yourself, what do I think success looks like? You know what, like what, what, what is a level that if my team is doing this on a consistent basis, I will be happy and how can I articulate that? And even if you don't go and make any tools out of this, even if you just have conversations where you say that to people like you're going to be a better manager.
- [39:45]
A
Yeah, the best AI work has nothing to do with AI. It's getting clarity, understanding standards, getting clear on who you're serving and how you help them. All of those decisions which you can be hand wavy and ignore really easily, but when you set them really helps you transform. Hilary, this has been an awesome, awesome show today. I have taken away so much from our conversation. I think you've taken people behind the scenes to actually understand what it's like to work in a really high performing team that isn't just slopping AI text back and forth to each other. Where can everybody find out more about you? If they want to take courses, do everything else. Like where can. Where can folks find you?
- [40:22]
B
Great. I have a substack where I post a newsletter that's probably the best way to stay in touch with me. It's called hills.substack.com, h I L S just one L. And then as I said, I teach a course called how to Be a Super Manager with AI on Maven so you can find me there. And yeah, I'm around the Internet talking to managers and sharing what I hear from them.
- [40:43]
A
Well Hillary, this was incredible. Can't wait to have you back on in the future with more lessons learned. And thank you for joining us on Marketing. It's the Green.
- [40:50]
B
Thank you so much for having me.
- [40:52]
C
Let me tell you about a great podcast, Demand Decoded. From the team at blend, a B2B website and demand generation agency, comes Demand Decoded, the podcast that breaks the mold of traditional B2B marketing tactics. Hosts Dan and Phil deliver actionable insights that transform your marketing into a revenue generating machine. Whether you're struggling with LinkedIn ads that aren't converting, want to understand the difference between lead generation and demand generation, or need to optimize your conversion rates, Demand Decoded gives you strategies that work today. Tune in for weekly episodes featuring expert analysis and hard earned tips that'll help you become a better B2B marketer. They recently did a great episode titled how to Fix the AI Inconsistency Problem Wrecking your Go to Market. Listen to Demand Decoded wherever you get your podcasts.
- [41:41]
A
This data is wrong every freaking time.
- [41:44]
B
Have you heard of HubSpot? HubSpot is a CRM platform where everything is fully integrated. Whoa.
- [41:50]
A
I can see the client's whole history. Calls, support, tickets, emails, and here's a task from three days ago I totally
- [41:57]
B
missed HubSpot Grow Better.