AI LEARNING
AI LEARNING
I've been exploring AI's capabilities and applications, from brand strategy to building AI tools
Tasked with helping designers bolster their strategy skills using AI, I created a workshop that taught participants how to turn any competent Large Language Model into a strategic partner. My question was: could AI tools, when properly prompted, provide the value of a trained strategist?
Strategists and Account Planners have historically played two important roles in creative advertising and design: informing work and selling it. I created step-by-step prompts to guide an AI through both processes. I also provided content and training to help workshop attendees understand these tools and how they are used. The idea was to democratize strategic frameworks (e.g. the 4 C's), thereby empowering anyone and everyone to benefit from the tools of strategy. I then walked through this approach with a group of designers signed up for an AI design certificate course at the Pratt Institute.
Not surprisingly, replacing a trained strategist with Gen AI wasn't entirely successful. Although the content the prompts returned was often generic, this wasn't the main problem. Instead, it was the inability of those untrained in strategy to identify the value and potential in what was shared back. While hardly surprising, this points to a paradox that arises when trying to replace experts with AI. AI may be able to generate amazing suggestions and ideas, but only experts are able to properly identify and assess their value.
The competition between the big tech companies behind Gen AI captures the headlines. But there is another battle playing out in the minds of everyday people. This is the battle of which AI tools they will come to trust and rely on. In the article below I explain why I see this as a brand battle and what it means for our future.
Historically, innovation is fueled when ideas from one domain are applied to another. Because Large Language Models have read every research article published online, they have exceptional expertise, broad and deep across disciplines. What if there was a way to leverage this exceptional insight to help create fresh and inspiring ideas?
Using the GPT builder in ChatGPT, I built a step-by-step process that first asks the user to share a problem with the AI and then prompts the AI to identify unexpected approaches taken from other areas of human knowledge and achievement. One persistent challenge was getting the AI to produce unexpected connections. Despite pulling from excitingly different areas of human knowledge, the AI seemed eager to make connections that would reliably work--suggesting solutions that the user may already be aware of.
The GPT I created, which I named Breakthrough, was good at pulling ideas from unlikely domains. It may be able to help product designers, consultants, and advertising creatives who are looking for new and creative ways to solve problems.
I love reading. But I don't have much time to do it. So I need every book I pick up to be a great choice. Because ChatGPT has read all books published online, it is the ultimate librarian. My question was how I could leverage its vast knowledge to provide book recommendations--which are deeply personal and shift with our interests.
I built a recommendation process based on 3 user inputs. Knowing that we're not always the best at describing our own tastes and desires--I wanted to start by giving the AI a lot of relevant data. So, I prompted the AI to ask the user a question: what book would you like to read again, for the first time. To ensure that the recommendation captures what the reader loved about this book, I added two more questions: about story (content) and about style (prose, diction, etc). The idea is simple. If a book is about something interesting and written in a way that keeps you engaged, one sentence at a time, it's likely to be a winner.
The GPT I created, which I named Book Muse, has been an easy and effective tool to identify prospects for my next great read. Unlike other GPT book recommenders, it follows a set 3-step process, ensuring the GPT has the info it needs before recommending. Try and see if it works for you!
Brands are built on meaningful differences. In the case of AI tools built on ChatGPT, the difference is easy to define. All GPTs (as they are called) use the same underlying Large Language Model (ChatGPT). They differ because built-in prompts influence their output. What if there was an simple way to brand these AI tools by laddering differences in output to a meaningful user benefit?
After many failed attempts, I created a sequential brand-building process. The GPT acts as a guide. Employing a 'laddering approach,' the GPT gathers information and asks users to make choices about key branding elements: e.g. what differentiates the AI's output, who is the target user, and how the tool benefits this user, functionally and emotionally.
While not perfect, the AI tool I created, called GPT Brand Builder, simplifies the brand identity process. The output (a brand name, a logo concept, a tagline, and a value proposition) includes most of what a user of ChatGPT would need to brand an AI tool they created for this platform. I've. used it to help brand the AI tools I have built on ChatGPT's platform.
Developing a brand strategy isn't a fixed process. It's messy, involving a lot of back and forth from creatives, clients, and the broader team until everyone aligns on an exciting approach. I wondered, where was AI most useful to create better brand strategies and creative briefs?
I started by focusing on the key results of a successful brand strategy (the creative is inspired, the client buys the work, and the campaign is effective) and used these to build a simplified process that incorporated best practices (the 4Cs, the creative brief, executive storytelling). I then attempted to use AI with any and every step along the way. Without a client problem to work from, I chose a brand that was trying to revive past glory with a recent rebrand: Jell-O.
Using AI was like having several excellent junior and mid-level strategists at one's disposal. While the results were compelling, what impressed me most was the additional time to focus on areas of interest and make decisions. AI boosted my confidence. I was able to gain a compelling understanding of the brand and category, and gather stats to support my case without sacrificing time needed to choose the strategy that I thought best.
I wanted to test the creative ability of AI tools. So I decided to use a brief that I had used with human creatives and evaluate the results. My question: how would AI respond to a brief designed for humans?
While in graduate school, I developed a creative brief for Payless Shoes that was based on a powerful emotional insight. Payless was facing a brand stigma that it couldn't easily avoid: it was perceived as being "cheap". We discovered a powerful benefit to "cheap": lower cost and lower perceived value meant lower risk for shoppers looking to try new styles that they didn't feel confident about.
I told the AI to act as top creative director, asked it to generate creative concepts, and then cut and paste the existing creative brief into Claude, ChatGPT, and Google Gemini.
While the creative concepts weren't earth shattering--they were on-brief and felt safe enough that clients might buy them. I was surprised to receive one creative concept that felt unexpected and compelling. Google Gemini delivered a concept called "Fail Forward" that embraced the idea of the fashion fail. Seeking to normalize "fashion trial and error" it suggested "partnering with relatable figures (minor celebrities, everyday people) to share hilarious fashion misstep moments," calling these "Own Your Oops Stories." I wondered if such a campaign could find a foothold in culture.
The process left me impressed. The AI had no trouble working from a traditional brief. Even without iteration, it delivered concepts that were in-line with what I would expect for an initial creative review.