So What? Marketing Analytics and Insights Live
airs every Thursday at 1 pm EST.
You can watch on YouTube Live. Be sure to subscribe and follow so you never miss an episode!
In this episode of So What? The Trust Insights weekly livestream, you’ll learn to revamp case studies with the power of generative AI. You discover how to transform existing content into compelling narratives that showcase your expertise. You learn a streamlined process for creating new case studies efficiently. You also gain insights into leveraging AI to analyze and improve your marketing assets, ensuring your revamp case studies resonate with your audience.
Watch the video here:
Can’t see anything? Watch it on YouTube here.
In this episode you’ll learn:
- Best practices for creating case studies
- What data you already have for your case studies
- How to use generative AI to revamp case studies
Transcript:
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.
Katie Robbert – 00:36
Well, hey everyone. Happy Thursday. Welcome to So What, the Marketing Analytics and Insights live show. I’m Katie, joined by Chris and John. How’s it going, fellas? We even got the on-location high-five.
John Wall – 00:50
We made it work.
Katie Robbert – 00:52
This week, we are continuing our summer makeover series where we are taking things that we already have that we just haven’t gotten to. And this is true of, you know, a lot of companies. We put ourselves last and we’re using generative AI to overhaul some stuff that’s been on our list. This week, we’re talking about case studies. So this came up last week when John and I were using generative AI to overhaul the Trust Insights services page. And if you missed that, you can go to Trust Insights AI YouTube where we basically created a scoring rubric to align the current services pages with our ICP and figure out how we needed to improve them. And one of those things that came up over and over again was a lack of a clear case study where we could demonstrate here’s the work that we’ve done.
Katie Robbert – 01:41
And it sort of—the conversation we realized needed to be its own thing. Because especially when you work at a consultancy and you do, you know, you’re in the service industry, it’s harder, at least in my opinion, to demonstrate results because it’s not, “We increased sales by 30%” or “We did this thing.” And that’s where I want to start the conversation. But I also want to, you know, take an opportunity to look at our existing case studies and see if we can use generative AI, not to only revamp them, but come up with a more efficient process.
Katie Robbert – 02:16
Because, you know, as we’ve stated, we put ourselves last and therefore doing things like putting together a case study, which should be part of our process offboarding a client or like once a quarter when we review contracts and see what we’ve done, we should just spin up a new case study, put it on the website. Because, John, one of the things that you mentioned in our livestream last week was eventually the case studies become the services pages. Like, you sort of, you point people there versus pointing them to the services because you want to demonstrate your capabilities. So all of that today to say, today we’re talking about using generative AI to revamp case studies.
John Wall – 03:00
That’s definitely where we want to be. The idea is you get to that level where it’s not just a list of services, but people can actually say, “Hey, I have these problems XYZ,” and the map is there for people to see that we know how to do it. But yeah, it’s a ton of work to get there. It’s not a simple process at all. It takes time. And even just, silly as getting customers who are willing to talk—like, we face this all the time. We have customers where we’re doing some really cool stuff. They don’t want to tell anybody what they’re doing. The last thing they want is word to get out. So, yeah, there’s a ton of things to juggle, but I’m interested in, you know, having a clear process for it is always a great thing.
Katie Robbert – 03:38
So the way that we’ve started, the way we started to structure our case studies that we have on our site to keep it simple was, unsurprisingly, using the 5P framework. So that’s typically where we start: What is the purpose? What was the problem we’re solving? Who are the people? Who was the client? Who was the stakeholder? Even if we have to anonymize it, what was the process we took? What were the platforms we used? And what was the performance? What was the outcome? It’s a very simplistic framework in terms of how to structure a case study, but it still feels heavy because you have to think about what were all the things we did. So, Chris, this is where I turn to you because you always have some really interesting ideas on how to use generative AI to make things more efficient or cleaner.
Katie Robbert – 04:24
So where would you like to start today?
Christopher Penn – 04:27
I will start. You already covered the 5P, so that’s all squared away. One of the things, the places that I typically start with an inquiry like this is our deep research prompt framework. Because generally speaking, none of us are walking around with encyclopedias in our heads. We don’t know everything that’s been happening in the marketplace, in the market space. We don’t know what new research has come out, what has changed with buyers. And the other place that I would start, and you have to go back probably about two months now in the Trust Insights live stream archives, is the sales playbook. So to what John was saying, if we scroll down in the sales playbook here to the different services, basically, what’s for sale? The ideal end state, except for things like Google Analytics, which doesn’t exist anymore.
Christopher Penn – 05:19
But the ideal end state is there’s some kind of case study for each of the things that are for sale. And if there’s not, then the question becomes, should it be for sale? Or do we have a case study that we’ve just forgotten about for that particular thing? So if we look at something along the lines of like, custom model development, have we built an actual model? Like not a GPT, not an agent, not a workflow, but an actual model for a customer? The answer is no, we have not. Have we done it for ourselves? The answer is yes, we have. And so a case study there would, might be.
Christopher Penn – 05:59
Be more focused in. When we look at the five Ps about the purpose and the process, but we might not have the people, or we might have a performance in the sense of this is the outcome we got, but we can’t say this is what the customer did with it. And to your point, Katie, very early on, in a lot of ways, if you think about, we’re kind of like, you know, a physician. I can tell you to eat less and exercise more all I want, but if you don’t actually do the thing you, then I don’t have an outcome I can point to that says, like, my advice helped you become healthier.
Katie Robbert – 06:34
Well, and I think that’s, you know, so that’s what John and I were talking about last week. That’s one of the challenges with case studies is you don’t always have that concrete outcome. And I feel like, you know, depending on the direction you’re taking your company, if you’re looking to try new things or break into new arenas, you may have only done stuff for yourself. I, I think that still should have a case study because you still did the thing. You just sort of disclose up front, like, “We did this for ourselves, but this is what we did. This is how we did it.” I don’t think that makes it an invalid case study because you still had to do the work.
Katie Robbert – 07:12
And I think that’s where I know, you know, I’m a little, I was trained more traditionally on a lot of these things. And that’s where I kind of struggle to break out of the, you know, “But this is how it’s done” because I was a little bit more formally trained on these things. And so I’m interested to see how we can creatively look at the work that we’ve done for ourselves, turn those into case studies, because we did the thing, we’re using it, we are our own client, and I think we need to look at it that way.
Christopher Penn – 07:43
So the first place I went, unsurprisingly, was Google’s Gemini. And I gave it a long, long prompt of how to, how should, how do we get good case studies? And the deep research, I fed it our sales playbook, which contains five different ICPs, which contains what’s for sale, which contains all of our marketing. And Gemini came up with a 28-page report based on sales. Some academic research it did, because I told it I want you looking at things like behavioral psychology, I want you looking at things like graphic design, I want you to look at peer-reviewed research and come up with a guide for how do I get a case study? And I gave it the five Ps where the purpose of the case study is to get someone to say, “I have that problem, I should hire Trust Insights.”
Christopher Penn – 08:34
And so this 28-page guide now gives me the basic foundation, which we are, in many ways, like you said, we already know the, the best practice in B2B marketing is called before, after, bridge. Like, you know, before and after. Our colleague and friend Jenny Dietrich calls it heaven and hell. We call it purpose and performance. Whatever you want to call it, that’s the basis of it. But now we have all the guides for design, how to appeal to the ICP. One of the things that came out of here was you might have case studies per different ICP, because a case study that appeals to a CMO is going to be different than a case study that appeals to a mid-level manager.
Christopher Penn – 09:10
Or a case study that appeals to a Fortune 500 is going to be different than the case study that appeals to mid-market.
Katie Robbert – 09:17
Makes sense. And I think that’s where we need to rethink the rules for ourselves in terms of approaching a case study. Especially if we’re not doing work that has that very tangible outcome of, you know, “We’ve done that work before. You know, we fixed up your SEO and we saw traffic increase from this to this.” Like, that’s a very tangible case study. But we don’t do solely and exclusively that kind of work for our clients. We do a lot of consulting work and consulting, to your point, as much like, you know, a doctor can give you the advice of what you should do, but you still have to do it. And so the doctor can’t claim that they had, like, they were the ones who got the result. Like, you, the human still have to.
Christopher Penn – 10:04
Do the work, right? So in those cases, there’s two different ways to approach it. One is doing what we’re doing right now, which is instead of a case study, you have something like a walkthrough or a live stream that demonstrates that you have the knowledge and the skills to do the thing. And other people can watch and go, “Yeah, he knows, she knows what she’s doing.” Or in other cases, you synthesize the case study based on results you’ve actually gotten, but, but in a more compact format. So let’s tackle, let’s go into this list here. One of the ones that we’ve done a lot with in the past is email list hygiene alert and deliverability consulting. And there’s actually a very specific use case for a client that we did. We just never wrote it down.
Christopher Penn – 10:51
And so we should probably try and transform that into a case study because I know exactly which client this is, I know what the details are, I know what the results were that we got, and now we just have to do it. So the first place that I would recommend starting with this is to pull open your phone, go into your voice memos app on your phone, and if you know the company or you know the situation, just start talking about it and say, “Here’s what was going on.” So anything you want to add before I start foaming at the mouth?
Katie Robbert – 11:28
No, I’m interested to see where this is going to go.
Christopher Penn – 11:32
Today, we’re going to do a case study on email list hygiene and deliverability consulting. We did this for one of our big B2C clients. We got a huge data export out of their Salesforce marketing cloud. Email marketing system was a database of 6 million emails that they had no idea whether or not they were good, but they did know that the deliverability was in the toilet. So what we did was we first validated their infrastructure. So we checked out their domain name, their DNS, we checked out everything at the MX toolbox. We ran through Google Postmaster tools, and we figured out that the reason that they were having issues was that like a third of their list was basically garbage.
Christopher Penn – 12:11
And so using some deliverability tools, both built-in that we made for ourselves plus commercial services, we took the whole 6 million person list, ran through it, pulled out as much junk as we could based on error messages, based on clearly nonsensical email addresses like asdf.asdfdf.com and email validation service. The one we were using at the time was called Million Verifier. And we cleaned up their list. Their list shrank from 6 million to 4 million. So they, we found 2 million bad email addresses, just garbage that they were sending to. And then when we gave it back to them and they re-implemented, they suddenly found that their complaint rates dropped like a rock, like almost to zero. Their, the bounces that they were seeing in their reporting console dropped to almost zero because they were no longer sending to nonsensical addresses that didn’t exist.
Christopher Penn – 13:07
And the net result was that their marketing campaigns were much more successful. We don’t have an exact number because we couldn’t see that in the system, but we know from their feedback that the project was a massive success because they were no longer sending to garbage. So when we approach this as a consulting firm, this is how we tackle it. Now, I’ll turn off my voice memo. That’s the. The, you know, the dime-store version of it. In any system you use, there’s probably a way to get the transcript out of the voice memory you just made. I’m going to go into my transcripts and let’s see where did it go? What’s today? I just transcribed the wrong one. I agree. 116. Yes. There we go. Let’s do that again. For anyone curious, this is using an app called Parakeet mlx. There we go.
Christopher Penn – 14:07
There’s my transcript of what I just foamed at the mouth at for almost two minutes. Now we’re going to go into. Let’s pick the generative AI tool of your choice. It does not matter which one. Claude, ChatGPT, Gemini, the works, they all work about the same. For this purpose, I’m going to attach my sales playbook, and I’m going to attach my customer, my custom research. I’m going to say from the attached transcript, following the best practices guide and thinking through the appropriate ICP from our sales playbook, let’s construct a case study narrative in the canvas, from the transcript.
Katie Robbert – 15:02
Before you hit go on that, wouldn’t you—or is it jumping ahead? Wouldn’t you use our case study template as well?
Christopher Penn – 15:10
It is jumping ahead. We want to get the story down first because there’s a very good chance that it’s going to create something that we need to proof to make sure it’s not lying. Because at least for us, that’s a core value.
Katie Robbert – 15:23
It is. We had an extensive conversation about that earlier.
Christopher Penn – 15:27
Exactly. I specified in the canvas because in ChatGPT, in Claude, and in Gemini, the canvas is editable and you can make modifications to the output right in the canvas if you want. With something like this, we always want to use the smartest model possible, reasoning model. So Gemini 2.5 Pro, ChatGPT03, Claude Opus 4, if you have it, Sonnet 4 if you don’t. What we’ve got here is a long, large amount of narrative. That’s a beefy amount of it. Looks good in terms of the language it’s using, before and after. Now we should go to our case studies page on our website and get an actual case study template that we can use. I’m going to use the Google Analytics one here.
Christopher Penn – 16:28
I’m just going to download this as the PDF and go back to Gemini and say, “This is great. However, it doesn’t fit our current template.” Question: I’m going to attach our current template based on the best practices guide. “How effectively does the template design and style meet best practices? Disregard the content, we’re focused on design and layout.” And I’m going to load the PDF, which is here from our website, and we’re going to get some feedback. Turn off deep research because I don’t know why that’s on. Oh, I’m in the wrong chat. Duh.
Katie Robbert – 17:32
Could you create a scoring rubric for how good your case studies are against?
Christopher Penn – 17:39
We sure can.
Katie Robbert – 17:40
Okay, I’m not saying we have today, but I’m just sort of wondering in terms of like how to make these things scalable for larger teams. You know, how does someone get more real-time feedback without having to like make like the director of the team the bottleneck?
Christopher Penn – 18:02
There’s a couple of different ways we could do that, and let’s tackle that in just a bit. When I retype this prompt that I was just working here, “Half of the design, meet current best practice standards, disregard content and focus on the structure and design.” Now, while it’s thinking, you absolutely can build automated workflows that can do that. So you could take an existing, say, narrative or voice memo or a conference call recording from a client, you know, if you have an off-boarding call and wire it through a set of prompts like we’re doing right now, but automated that would then spit out the individual pieces. The overall assessment of our current template, it serves as a basic project summary, falls short of the modern high-conversion best practices outlined in the guide.
Christopher Penn – 18:57
It reads more like an internal report than a persuasive marketing asset. Point-by-point comparison narrative framework. The template uses the five Ps. This is very different from the before-and-after bridge. While problem performance loosely matched before and after, the structure is more of a technical report and lacks the clear, compelling story of transformation. Trophy case is missing. This the guide identifies as a critical executive element for executive skimmers. What is the most impressive result? Powerful testimonial missing. This is a major weakness. Client snapshot incomplete. Call to action missing. Visual hierarchy missing. The headline is there, but it needs more. More pizzazz. Visual flow. The hierarchy is flat and lacks a clear visual path section. Typography is okay, but it does—has a poor hierarchy and has a bland use of white space. Data visualization successfully shows up as a basic visualization.
Christopher Penn – 19:52
So, based on the general best practices, the Gemini version of this is not fond of it.
Katie Robbert – 20:01
I don’t know, John. I’m having a rough couple of weeks. Last week, all of the services pages that I created scored super low. And now this week, the case study template that I created is just not good. So I’m about like one more live stream away from resigning from my position.
John Wall – 20:23
Well, you got to keep in mind that these large language models are trained on the whole web. And you know, there’s companies that have a team of 20 that are going to work on this page. You know, you’re getting judged against that. So.
Katie Robbert – 20:33
No, I know it’s. I’m. But I think one of the things that, you know, I like about this process that we’re doing, the summer makeover series is because we are a small company and because we tend to wear many hats, I tend to wear the CMO hat. But I’m not a CMO. It just needs to get done. And so it’s helpful for me, as someone who is not a CMO, is more on the CEO COO side of things, to understand like, where is this falling short? How can we do better? Because if my job primarily is growth, then I’m not doing it. So what can I do to, you know, get better growth? So I actually see it as a real opportunity for learning, provided I’m open to it, which I am, you know, so how can we do better?
Katie Robbert – 21:26
And so, Chris, one of the questions that I was asking you that, you know, we can get into, it’s not so much the process for creating, it’s the process for reviewing the case study. So a lot of times in a team, you know, the person who heads up the team or the CMO is responsible for like that final sign-off. And that’s where a lot of bottlenecks can happen. And so that’s why I’m wondering, like, could you put together a scoring rubric to help whoever’s putting together the case study get it to a certain point? Like, don’t put it in front of the final signatory until it’s like a 95 out of 100, otherwise you’re wasting their time. So, like, use the machine to keep refining it until you get to that high score and then ask for sign-off.
Christopher Penn – 22:13
And magically, Gemini has created a case study evaluation rubric, which we are now going to put in Google Docs just to make sure. Pro tip, by the way, folks, if you are using Canvas in Gemini, the export button has an export to Docs.
Christopher Penn – 22:58
Yes, this is what works in 2025 for case studies. However, I would recommend that if you’re going to do this in production, please do spend more than 15 minutes reading the output because 28 pages is a lot. So we’ve got the scoring rubric, and now we can say, “Great, using the scoring rubric, score the existing AAA case study PDF I loaded.” So we can say we can do exactly what you said, Katie, which is run a score analysis on to see how it comes out.
Katie Robbert – 23:33
I feel like these are really handy. I wouldn’t even call them shortcuts. They’re just really good tools in your overall arsenal, especially when you have a larger team and everybody’s sort of fighting for that attention of like, “I need feedback on this” and “I need feedback on this.” Putting together these rubrics are really good way to cut down on that bottleneck of one person or two people being responsible for reviewing. You know, that by the time it hits your desk, it’s already in really good shape, and so the back and forth should be, you know, minimal.
Christopher Penn – 24:07
So this case study scores a 47 out of 100. The big gaps, client is the hero and addresses emotional anchors. Those two things are fall short in that category. And the persuasion components, benefit-driven headline, trophy case of key results, powerful client testimonial, persuasion, persona-driven call to action. So we did specify persona in the original case study, which is something that the rubric does want us to also be focused on and addressing cognitive biases. So thinking about the cognitive biases for the reason why people read case studies in the first place. And so it says it’s got a strong foundation, but it needs more.
Katie Robbert – 24:45
So does this tell us the next step? Like, okay, what do we do?
Christopher Penn – 24:51
Yes. So the next step would be to say, great. Based on. On the scoring rubric, do a new draft of the AAA case study in the canvas using HTML, CSS, D3JS, JS or equivalent from CDN JS Tailwind that keeps the. What is the. Is a page 4 by 3 or 16 by 9? Actually, this is 16 by 9. The 16 by 9 aspect ratio, but transforms the case study to align with the best practices guidelines and the Trust Insights sales playbook.
Katie Robbert – 25:49
What’s interesting is you’re having it redo the entire thing, whereas I would have thought to just redo the content and then repaste it back into the template.
Christopher Penn – 26:01
I forgot. Brand standards. You must use the following brand standards. Now we hit update.
Katie Robbert – 26:16
And if you haven’t already created like a knowledge block for your brand standards, do it. It’s going to be a pain in the butt to keep looking it up every single time. And where did I put that PDF? Like, wherever you keep your generative AI information, your knowledge blocks, your prompts, whatever it is, create a knowledge block of your brand standards, even if it is, as Chris pasted in your fonts and your colors.
Christopher Penn – 26:41
Exactly. We’re going to see what this comes up with. Because these tools cannot natively create slides, which is functionally what our case studies are. They’re single-page slides. They just do that poorly. They’re getting better, but they’re still not great ads. But they can create web pages. We basically said, “Make it into a web page that is 16 by 9.” Now let’s see as this thing renders. So it made the chart separate from. I don’t know that it needs to be quite that literal in the before. The version. The after. Yeah, that’s. That’s taking it a bit literally. But we do have a nicely remixed graph, which is. I like that little project start thing right there. It says, “Hey, this is where things were.” That’s. That’s kind of nice. And I feel like it’s a bit on the wordy side.
Katie Robbert – 27:41
It is. And you know, and this is John, where I sort of ask you. And this question came up in our Slack community today, which if you’re not a part of, you can go to Trust Insights AI slash analytics marketers, and, you know, question around, like, the templates. Like, in your experience, you know, is it better to have a long version that really gets into, like, “Here’s exactly how we did it” and all the things, or do people just want like those quick headlines take away? Especially the sales team who’s just like, “Let me just show what we can do.” Like, is there room for both, or is it one or the other?
John Wall – 28:17
Yeah, there’s definitely room for both. It’s, it’s a matter of the use case, right? It’s like, what is it going to get used to? It’s, I mean, in general, most of this stuff is marketing stuff. So yeah, it can never be too short, and it can never be too simple. You’re dealing with, you know, room-temperature IQ at best. So it’s gotta be like that. But there is still space for, you know, academic white papers and case studies. Like, there are some people that really want to dig into the full details and find out exactly what’s going on and how it’s working. But the thing you’ll find too is that, you know, those are way down at the bottom of the funnel and this stuff tends to be top-of-funnel stuff. So, yeah, you know, it’s gotta be super simple.
John Wall – 28:54
But there is a lot to be said for, like, I was at a company that had a software development tool, and we had a couple of white papers that were 10 or 12 page documents on, you know, how process works and how your development process should come together. And those were, we did those white papers, and those would pull for seven or eight years. You know, those became standard resources that it was like, “Okay, if this is your first day on the job, go get that white paper and read it.” And now, it didn’t close any deals right away, but two years later, when the company got big enough that they were having problems, they would remember that were the ones that did that. But yeah, I think unfortunately, this is way on the marketing side here.
John Wall – 29:31
It’s got to be big, bold, and the fonts can always be a little.
Christopher Penn – 29:35
Bit larger. Make the logo bigger.
Katie Robbert – 29:39
Fair enough.
Christopher Penn – 29:40
So it’s going to now have to think through the design, but I gave feedback, I said this needs to be a single page in one logical section. The chart dangling underneath the narrative is ugly. The headlines are good, but too literal. We don’t need to tell people the before, bridge, and after. The text is also very wordy. Think about the ICP and how you’d word this for maximum impact, but word economy. And so now it has to go back and take that feedback and incorporate it. So that’s an improvement. That’s, that’s already better. It’s now more of a portrait size rather than a landscape size. That’s something that we might want to think through. But pretty clearly, when you look at just what’s in the preview, go, “Oh, 55% less unattributed traffic. Cool. I have that problem.”
Christopher Penn – 30:22
And so this makes a lot of sense to me narratively. If I was an outsider saying, “I’ve got a lot of unattributed web traffic, who can help me?”
Katie Robbert – 30:29
Right? And you know, the portrait versus landscape, that’s not something I’m going to like, put up a fight over. Like, if portrait works better than landscape, great, let’s do that instead. I don’t really care. I just want the case study that’s going to tell people, “Here’s what we did, here’s how we solve your problem.” So if this is a better format, amazing. Let’s put all of our existing case studies through it and then create some kind of a process where we can do what you did, Chris, and just put together a narrative and have the system just churn these out one after another.
Christopher Penn – 31:06
This is the revision of a case study. So this takes an existing case study and aligns it to the best practices and then makes. Makes some decent design out of it and stuff, which is good. We good?
Katie Robbert – 31:20
Before we move on from it, could we take what we just did and ask again? I might be jumping ahead, so I apologize. I don’t want to keep stealing your thunder, but could we take what we just did and ask the system, “Can you turn this, what we just did, this process into a prompt or system instructions that we can reuse?” Like, is that. How is this where we would do that? Is that even something we could do?
Christopher Penn – 31:46
You could, but we’ve already written a good amount of it, so it’s more just some copy pasting.
Katie Robbert – 31:51
Okay.
Christopher Penn – 31:54
So let’s take. In fact, you know what, let’s do that just as, as good practice, let’s turn off campus and say, “Okay, create system instructions for an LLM like Google Gemini. To repeat the process of evaluating a case study and then generating a new version using the canvas.” The instruction was going to say, think through: 1. The intent of the prompt, 2. The downstream or second order effects, 3. The appropriate prompt engineering techniques, especially those after 2024 01014, the background knowledge necessary, then create the system instructions. What we’re doing is having a. Do a bit of reflection and think about what is my intent, what am I trying to do? And then using its own background knowledge to build the thing. Let’s see what it comes up with. Because what we might want to do. I’m going to see if this will work.
Christopher Penn – 33:18
While it’s thinking, let’s see if we can get into this on our client services account. We can. Excellent. Google Opal works. For those who are not familiar, Opal is like a. A Google-centric, cloud-based super simple version of N8N. It’s N8N without all the extra bells and whistles, and it only works inside the Google ecosystem. So if you’re not a Google shop, this is going to be of limited utility. However, we are a Google Workspace company, so that’s not a terrible thing to have for us. So it says we’re going to need our sales playbook. We got our best practices study guide. Yes. Good. This all looks good. Let’s export this to Docs, and we’re going to call this “System Structures Case Study Revitalization Workflow.” This is the giant prompt, which is nice.
Christopher Penn – 34:12
Let’s go into Opal, create a new application, and let’s go ahead and add assets. We are going to. Let me think about this. We want a user input of an existing case study. Add the existing case study. You want to revitalize. That’s our input. Next we need a. We’re going to need an asset block. We’re going to go from Google Drive. We’re going to go to
Katie Robbert – 35:42
Does the system instructions include our brand guidelines, or do we need to add that as well?
Christopher Penn – 35:47
We need to add that in. There’s our case study. And let’s add in. We should just make that new Google Doc. Trust Insights 2025 Trust Insights brand style guidelines knowledge block. And we will say this is: “You must use the following brand standards.” There they are. Good save and close. Add in our brand standards, My Drive 25 brand standards, and move this block over and draw its little connector. Now we should be able to have it do Gemini 2.5 Pro. Let’s think about this. Let’s think about order. Execute the system instructions in and move the system instructions block up using these background documents. Read the user case study in. We’ve basically created a set of meta prompts here joined together. Now, the outcome of this, we have to think about what we want as the output from this.
Christopher Penn – 37:32
We probably want, “How do I make my toolbar going?” That’s annoying. There we go. We probably want this to spit out a nice web page with auto layout. That’d be nice. We might want to also have it spit out something that can go to a Google Doc. And just for fun, let’s see if it can spit out something that can go to a Google Slide deck.
Katie Robbert – 37:59
That would be lovely.
Christopher Penn – 38:01
It’s gonna suck. I’ve tested this. I was testing this yesterday, and it was just appalling. It was terrible. But it’s not bad. I mean, it’s better than nothing. So we’re gonna call this case study revitalization. Now, I do want to caution Opal is it’s not even in beta. Like, it’s in alpha about 50% of the time. It just crashes. So just being aware of that. Let’s see. Also, we need. Oh, that’s. Let’s see. Do I have an user input block? All right, let’s try this. See what happens here. Case study revitalization, start. Let’s take the existing case study. We want to revitalize. Let’s take our AAA on upload from device, desktop. Revitalize.
Katie Robbert – 39:00
I’m curious. I think Opal is an interesting way to approach it. Why would you do this versus taking the system instructions and background knowledge and not make a gem, for example? So therefore, all you would have to do is say basically the same thing. You know, you take the system instructions, you look at all the things, and then you just say, “Here’s my document. Revitalize it.”
Christopher Penn – 39:26
You absolutely could. Opal offers you a few more bells and whistles, particularly like the Google Doc, the Google Slides, and stuff like that, assuming it runs.
Katie Robbert – 39:35
But if you were just looking for like an updated narrative, and again, sort of like what I had said, like where you just want to get the updated content and you can copy and paste into your template, that still takes you a really far away.
Christopher Penn – 39:49
Absolutely. So, for example, exactly what you’re saying. We could take the existing contents, and we can start a gem, and there’s our system instructions. And. And now from our knowledge base, we add our individual files. So we have our brand style guidelines, we have our case study evaluation rubric, we have our case study best practices. We have a master sales playbook, and all of that can get plugged right in as documents. And now it has. We’ve essentially built exactly the same thing.
Katie Robbert – 40:21
Which I think it really depends on your organization. One may be easier to approach than the other. Or to your point about Opal, you may have more things that you need to do with it.
Christopher Penn – 40:33
Exactly. If we run this, it should do functionally the same thing. The one thing I do like about Opal is that there’s a console where you can see what it’s doing, particularly on multi-step things like doing deep research. That part I really appreciate. But again, you can just hit show thinking and have it do exactly the same process. That would be the ways to revitalize an existing case study. A gem. Copy and paste the prompts yourself if you want, or use workflow management of some kind to glue it all together.
Katie Robbert – 41:10
Now, what do we want to do with the new case study that you recorded the audio for?
Christopher Penn – 41:18
We should probably have it try and make something out of that.
Christopher Penn – 41:40
A net new MM based on the transcript I provided earlier, referencing our sales playbook, our best practices for case studies. Our ICPs render a new document in the canvas that using HTML, CSS, Tailwind, D3JS, CDN JS conforming to our case study frameworks and formats for the email list hygiene and deliverability case. Let’s see if it can take that. Two minutes of foaming at the mouth, and obviously this was a toy demo. You would want to invest more than two minutes talking about what you did for the customer. But let’s see what it comes up within the canvas to see if it can actually put together a moderately compelling case study.
Christopher Penn – 42:49
One of the things that is difficult to overstate is we spent so much time and effort on a sales playbook, on ICPs, and on all these things that this is making it look easy. If you don’t have those prerequisites, you need them because this will not work without it.
Katie Robbert – 43:09
So we did a lot of those on the live stream, which you can get on our So What playlist at Trust Insights AI YouTube. And if you want help, honestly, just reach out Trust Insights AI Contact. And John, the man, the myth, the legend, will be happy to, you know, point you in the right direction.
Christopher Penn – 43:30
Oh, I love that headline. “From 6 million problems to 4 million prospects.”
Katie Robbert – 43:36
Such a marketing headline that is, but I love it.
John Wall – 43:41
We’ve got 6 million problems.
Christopher Penn – 44:07
Exactly. And that looks original list, recovered garbage. Yeah, of course. You know, surgical data restoration, DNS, MX records, million verifier. Yep. By room 2 million bed. I mean, that looks fantastic for two minutes of foaming at the mouth.
John Wall – 44:07
Yeah, I like it. It’s tight.
Katie Robbert – 44:10
Can you use the existing gem system instructions to create net new? Because we called it revitalization, but it sounds like you can use the exact same to take the transcript and create something new.
Christopher Penn – 44:23
I would generate a new prompt for that. I would say, “Let’s now create system instructions for a net new case study. The user will provide a transcript or other background material. You will use the sales playbook, ICPs, case study best practices, scoring rubric to build a net new case study in the canvas, adhering to the brand standards. Convert this into system instructions as we did before.”
Katie Robbert – 45:11
And for those who can’t tell, like we are legitimately doing the summer makeover of our own stuff because these are things now post-livestream that we’re going to start implementing, which is why I’m asking Chris, well, what about this version of it? What about this version? Come tomorrow morning, we’ll be using this for ourselves because this is the kind of thing that you would want to have, especially in a small company or even a large company, to help just make things more efficient. So it’s not shortcutting the work that you have to do for the client or your customers. It’s just making it more efficient to get the word out there that you did some really cool stuff.
Christopher Penn – 45:49
Exactly. So let’s. We’ll export this to documents to save it. We’ll also just copy it. And then let’s go back to where were we earlier? Let’s see. This was the gem. Was this the artifact? All right. So it did not use the canvas in the gem. That’s not a big deal. You can just copy and paste because that’s not hard to do. Let’s see if the gem did what it’s supposed to do. Hey, not bad at all. I like that the logo is broken, but that’s easy enough to fix. All right, so let’s take this. Go into Explorer gems, create a new gem, and we’re going to have a new case study generator gem. And we’re going to take our system instructions, copy, paste, plop them in, take our files from drive. We’re going to use our system instruction. We don’t want that one.
Christopher Penn – 46:56
We want case evaluation, best practices, sales playbook. Those are all going to go in and they’re going to finish rendering here.
Katie Robbert – 47:09
This includes that scoring rubric that we created.
Christopher Penn – 47:11
Sure does.
Katie Robbert – 47:12
Amazing.
Christopher Penn – 47:14
We now got our new gem ready to go to build a new case study. Let’s give it a test. We will give it the foaming at the mouth, make a new case study.
Katie Robbert – 47:34
And while that’s in 45 minutes, 47 minutes, you’ve created two new processes for the sales and marketing team, which is amazing.
Christopher Penn – 47:49
Let’s see how we did in here. It did some rendering. Final critique. Let me check to see if it in fact did put anything in Google Drive or it just favorized, which is entirely possible. Oh, that’s ugly.
Katie Robbert – 48:12
Yeah.
Christopher Penn – 48:13
However, it did something. Yes. Continue. Go away. Stop. Crashed Chrome is what it did.
Katie Robbert – 48:24
That’s again, it’s something.
Christopher Penn – 48:26
There we go. Finally. It basically turned our thing into a slide deck. Now, this is not great, but.
Katie Robbert – 48:35
I can see that there’s a use case for something like this, especially with capabilities decks or pitch decks. That thing I’m interested to know. I know we scrolled through everything pretty quickly, but with the new versions of the case studies, does it give you what it’s scoring itself?
Christopher Penn – 48:58
Let’s take a look. So it scores the original, it does the rebuild, and it does a reflection. It doesn’t apply the scoring critique again. It only applied the rubric the first time around. So we’d have to change the system instructions. Right.
Katie Robbert – 49:17
But that sounds like it’s easy enough to fix.
Christopher Penn – 49:22
Exactly.
Katie Robbert – 49:22
It’d probably just be adding a line like, “Okay, and rescore the new version.”
Christopher Penn – 49:27
Right. So let’s try and see how the GEM did that. We built on its own output. I like the other headline better. “From 6 million problems to 4 million prospects.” I think that was more clever writing.
Katie Robbert – 49:44
I think the other headline is better, but I think that this one aesthetically looks better.
Christopher Penn – 49:52
Yep. And so now that GEM can be used by anybody because it’s on our on the client services account. So any of the folks on the team can use it to create new cases. Particularly if we go back to where we started this live stream and look at the. Look at this case study and gosh, I know. We did a voice of the customer project. When did we do that? Oh, that’s 2022. I can sit down and narrate it and just hand off the narration to the gem directly or to someone on the team and say, “Turn this into a new case study.”
Katie Robbert – 50:25
Yeah. I think one of the things, you know, this is for us to do offline, but, Chris, you and I can sit down and go through these one by one and just record what we know about each of them and hand a pile of transcripts to the GEM and say, “Great, go make a bunch of case studies,” and then ask our, you know, handy web developer intern to get them up on the website. As we discussed last week with the services pages, I am the web developer intern as well as the CEO.
Christopher Penn – 50:58
Exactly. But we’ve built a better. We’ve built an engine. Doing this and that really is the point of all of this is generative AI isn’t just about prompting. It is about building solid process, repeatable process with best practices. Which, as much as people love to crap on best practices, the reality is, I think it is like, here’s the bar of minimum competence. And we know because we are a small startup and some of these case studies, like three years old now, were below the bar of minimum competence. So now we can bring ourselves up to the bar of minimum competence.
Katie Robbert – 51:35
Well, and like I said at the top of the episode, I mean, as the person who had been creating a lot of the case studies, it’s not my core function, but it had to get done. And so I’m excited to learn better best practices of what should actually be in there versus what I think should be in there. And I think that’s a lot of what we’re learning through a lot of these processes. Like creating the ICP, it takes our personal opinions out of it. Therefore, it’s more suited for what the customer wants and not just what we think they want.
Christopher Penn – 52:05
Exactly. So that’s the summer makeover on case studies. We now have two apps that we can use anytime we want to refresh an existing case study or to make a brand new one.
Katie Robbert – 52:17
I’m pretty gosh darn excited about that legitimately.
Christopher Penn – 52:22
John, any final parting words?
John Wall – 52:24
No, I love it. This is just the thing you want. Because now it’s just like you said, Katie, it’s totally. That case study where some prospect comes in is like, “Hey, have you ever done?” You’re like, “Oh, I know we did that sometime on something.” And to be able to just make a neat little PDF that they can send around to their team, like, that’s huge. That makes you look like a much bigger company.
Katie Robbert – 52:43
I mean, I feel like, John, you know, that’s going to be part of our new process is when a new prospect comes in, what are they asking for? And do we have a case study? If not, can we go ahead and just put that up really quickly? Someone’s asking, can we share the Slack link again? Our free Slack group Trust Insights AI analytics for marketers. Totally free. I. I suggest you join personally.
John Wall – 53:08
There’s no downside to being there.
Katie Robbert – 53:10
There’s no downsides. No.
Christopher Penn – 53:11
Exactly. And there’s plenty of upside, including the Friday Questions of the Day. That’s going to do it for this episode, folks. We will talk to you all on the next one. 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 Trust Insights AI TI podcast and a weekly email newsletter at trustinsights.ai newsletter. Got questions about what you saw in today’s episode? Join our free analytics for marketers Slack group at Trust Insights AI analytics for marketers. See you next time.
|
Need help with your marketing AI and analytics? |
You might also enjoy:
|
|
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, INBOX INSIGHTS. Subscribe now for free; new issues every Wednesday! |
Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new episodes every Wednesday. |
Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.
Newsletter
Podcast
