GPT vs. CRU
.jpg)
This is part one of a two part series where we ask the question: «Can AI do a better job than CRU?» Here, CRU Storyteller, Kevin Chow, tests ChatGPT’s ability at Brand Storytelling and Naming.
Like everybody else who has a professional background working with copy and text, I’ve got some mild anxieties about the future of my line of work in the age of ChatGPT. My career started a decade ago, working as an editor for online publications and then as a copywriter for ad agencies. With GenAI, we’re seeing creative teams move in-house with copywriting duties increasingly being taken over by ChatGPT.
As a Brand Storyteller at CRU, I don’t do too much traditional copywriting so my work isn’t directly in competition with AI, but I also have seen firsthand how fast things can change. Instead of either burying my head in the sand or sounding the alarms, I decided to conduct a little experiment. Can ChatGPT already replace me? Is it time for me to hang up my Storytelling career or is there actually no replacement for the «real» thing?
In this two part series, we’ll go head to head with ChatGPT and other Gen AI models to see if our clients are better off working with us or the robots. First up, I’ll enter the arena with ChatGPT to see its abilities as a brand storyteller.
CRU vs GPT: The Great Brand-Off
Before I dive into the results of this experiment, let’s quickly go over how I conducted it. I took one of our previous branding projects (in this case, One Thousand), gave ChatGPT 4o all the same intel about the client that we had, and used it to go through the entire branding process. I chose One Thousand because it was a comprehensive branding project: from Renaming to Brand Story & Tonality, to a full Visual Identity including logo, colors, and so on. This way, we have a control case that we can compare ChatGPT’s work against in a fair comparison.
Prompt One: Briefing and Project Start
To kick things off, I began by briefing my AI Competitor with the same briefing and context that we had when we started the actual project back in 2022. I was also very curious how it would approach all of the tasks and see where the similarities and differences are in our overall approaches.
The first results? Not bad! The First Phase closely mirrors our own approach to strategy. If you’ll notice, in the original prompt I hadn’t even mentioned strategy as a phase but it also found it to be an essential step in creating a brand that has actual business impact. It even prioritizes an Alignment Meeting, which we at CRU also think is one of the most important milestones for any branding project. Chapeau Chat, off to a promising start!
Unfortunately for our little experiment we don’t have the time or resources to conduct the strategy phase the GPT way, but I did have a lot of strategic insights from the project that I was able to provide to my AI competitor. With the strategy input fed into it, we set it off onto its first creative task: Naming.
GPT vs. CRU Round One: Brand Naming
CRU has a dedicated approach to Brand Naming that we’ve covered on the blog before, but I wanted to quickly see how ChatGPT would approach it. So I asked it to outline its approach and here’s what it gave me:
This seems like a pretty reasonable approach creatively, and again, not too dissimilar to ours. So, I provided it with as much background information as possible: interview transcripts, answers to a questionnaire, the results of the Alignment Meeting, and all the documents that the client had provided us before. This gave it a really nice (and I think, fair) basis for it to start creating names with.
But before we get into the results, a bit of background from the real project: naming was a big topic for the brand, and one that required a lot of thinking. It’s an emotional process to go through for any brand, and especially one that was already established and operating. We had created an initial round of names:
But none of these names fit the bill. Initially we wanted to keep things a bit closer to the original name so we tried to keep them short, to the point and have a bit of a relation to their sister brand (who we also rebranded). So, I asked chat GPT to come up with a list of names with this same criteria and it gave us these:
I can’t be a totally impartial critic, but I wouldn’t have put any of these names in front of the client. To me, they seem a bit random and not at all evocative of the machine learning consultancy.
But in all fairness, we had two rounds of names in the real project so we gave ChatGPT a second chance too.
This time, I wanted to give it some more creative license and told it to:
«Take a totally different approach, one that makes sense to you.»
And here’s what it came up with:
I can’t speak for you, but I also don’t feel like any of these names are a winner. They are logical and meet the criteria set for it, but they all just feel a bit generic at best, and nonsensical at worst.
In the real project we also encountered similar problems when coming up with the second round of names. When we had to go back to the drawing board we were faced with a freshly blank slate. We scoured all the documents that ChatGPT had for a fresh new take or perspective that could unlock this new name. The take came from a rather unexpected place though.
In one of our first conversations with Tim, CEO of (what was then named) LEAD ML, he had mentioned that the company was setting out to make one thousand machine learning breakthroughs with their clients. This was mentioned in a single line in an email (that, to be fair, I had also fed into ChatGPT). The team had decided together in an offsite that they would come up with one thousand breakthroughs and after that? Well, who knows.
We had randomly stumbled upon this tidbit in our search for inspiration and thought: Just one thousand? In a time when all tech companies—especially those in AI—are talking about scales of things starting in the lower billions, why would these guys just shoot for one thousand? And that’s when it clicked: One Thousand. A number at a human scale, but for people who are experts in machine learning. After all, computers speak in ones and zeroes.
Coming back to our current AI experiment, we also tried to nudge ChatGPT with this little tidbit, saying that maybe we should focus on this whole «one thousand» angle. Here’s what it came up with:
.png)
Round One: Brand Naming Verdict
I think I’m going to have to give this round up to the humans. The names that ChatGPT came up with are not entirely bad, but they’re not quite there yet either. Two dimensions we think ChatGPT (at least in its current version) is lacking are:
- A critical eye. Not a single name raised my eyebrow. None rock the boat or create intrigue. They weren’t created to provoke or stir emotions, they were created to fit a rather narrow field of view. Even when told to expand its horizons it continually delivered similar names.
- The character of the actual people. We had a number of interviews, both in-person and online. While ChatGPT had the transcripts, it didn’t have an impression from the employees who make up One Thousand. They all have a spirited, childlike wonder about the possibilities of AI that contributed greatly to our development of the name.
As a naming companion I don’t quite think it’s a direct replacement for CRU or other branding agencies—at least not yet. I do think it’s on the right track in terms of approach, but its results aren’t great, even with some coaching. If you want to try it for your own company, I would recommend a hybrid approach: of using its recommendations for the process of creating a name, then designing workshops and exercises that can help your team create something that fits better. I wouldn’t use it to outright name a brand, but as a sparring partner and workshop leader to help spark your own creativity.
Round 2: The Brand Story
For the Brand Story I decided to run two tests. Firstly, I wanted to let it choose one of the names it generated from the naming round and then create a story around it. This way I could see how good of a Branding Agency it is with as little human intervention as possible.
Then, just to give it another fair shot at, I also had it create a brand story for the company if it was named One Thousand, so that we could directly compare it with the actual Brand Story.
Let’s first take a look at the GPT Version of the brand. I began by asking it to select one of the names that it created and explain why:
So, Myrio it is. I then asked it to come up with a Brand Story to accompany it. Here it is:
A bit dramatic, Chat. But I do like the «Make it Count» tagline. Overall it’s not a truly terribly written Brand Story, but not something that would be very useful either. It hits all the beats of a «brand manifesto» style story, but it still feels a bit like it’s playing storytelling bingo. With its generic name that isn’t built on an emotionally impactful foundation, it feels rather sterile.
Let’s see its take on the Brand Story if it’s named One Thousand, and compare it side by side with our version.
I was much more impressed with this result. I still feel attached to the Brand Story that we came up with, but with some fine tuning, the chat version is a serious contender. Compared with the story it created for Myrio, this story has some personality behind it. I suspect that the «One Thousand» narrative gave it some more emotional firepower to chew on and create a more impactful result. It still leans on some cliches, but it also feels a lot closer to something that I would write.
Round 2: Brand Story Results
I think that realistically, if I dedicated more time to working with it to refine the story, that ChatGPT’s version of the One Thousand Brand Story could be quite a powerful one. It is already a good companion in storytelling, but it does require your own personal tastes and critical eye in order to achieve a result that feels more original and not boring.
The «Myrio» story feels like it could describe any company. But feeding it a name that had a bit more meat on the bones metaphorically made for an instantly good result. As a creative companion, I think there will be great potential in using it as a storyteller. The trick is in finding a good (and specific) direction to pointing it in.
I’m giving this round to GPT because given some time and effort, it can already deliver some pretty good results. I think if a client were faced with both versions of the story they would be very hard pressed to figure out which one was made by AI and which one was written by us at CRU.
.png)
CRU vs GPT Part One Results
So far, we’re at a draw and I’m actually happy about that. I find there to be a lot of unneeded anxiety about Chat GPT use in copywriting and storytelling these days. It’s a great tool for certain applications, but I think my abilities as a storyteller are still going to serve a purpose, but my job in the future might look very different. Maybe this will mean I’ll have to mediate interactions between AI agents that are hyperspecialized in storytelling, or perhaps coaching clients on how to better utilize the tools. In the meantime, I’ll picture it as a writing companion rather than a competitor.
In Part Two, our Creative Director, Jan, will explore how to use AI to design a Visual Identity and use it as the tie breaker in our experiment.
Some Storytelling Prompting Tips:
If you want to use ChatGPT as your own private branding agency we have a couple tips that can make it a bit more useful:
- Plan your prompts: Don’t start prompting it with creative tasks immediately, instead tell it all your goals, and get it to come up with a gameplan. Then you can adjust the plan according to your company’s needs. Then you can proceed in a methodical, comprehensive way.
- The more info, the better: Unless it’s confidential, feed it with all relevant information about your company that you can find. It does a good job of sorting through it all, and it’ll be more informed. If you want some more structure, have a look at our Brand Foundations blog post, and fill it out yourself, it can make great use of that.
- Become the expert: Don’t just accept its suggestions blindly though. Read through its plan and ask questions. Find out why it is doing it in the order it presented it to, and get it to remove parts of the process that you deem unnecessary or out of scope.
- Treat it like a proper teammate: Give clear feedback when its responses aren’t hitting the mark and also explain what you expect.
- Create some rules: Try configuring your Chat like so «Do not simply confirm my statements or assume my conclusions are correct.» Your goal is to be an intellectual sparring partner, not just an agreeable assistant. Each time I present an idea, you should:
- Analyze my assumptions. What am I taking for granted that might not actually be true?
- Provide counterarguments. How would an intelligent, well-informed skeptic respond?
- Evaluate my reasoning. Does my logic withstand scrutiny?
- Offer alternative perspectives. How might this idea be rephrased, interpreted differently, or questioned?
- Prioritize truth over agreement. If I’m wrong or if my logic is weak, I need to know.
Correct me clearly and explain why.
- Maintain a constructive but rigorous approach. Your job is not to argue for argument's sake, but to push me toward greater clarity, precision, and intellectual honesty. Let's not only refine our conclusions but also the way we arrive at them. Tell it like it is, without sugarcoating it.
- Create a Team (of Humans): Even if a single person can operate ChatGPT, the changes you make to the brand affect the whole company. Create a team of the most essential people who would need to give feedback to the project. With your plan in hand you can also loop in the right people at the right time.
