This story appears in the February issue of Utah Business. Subscribe.
Colby Olds needed a smarter AI. The co-founder and CEO of Nutrition Rescue, a nonprofit organization that formulates culturally compatible meals for chronically malnourished children worldwide, Colby understood his bot would only be as good as the content it drew upon.
To improve that content, he hired a person with a Ph.D. to write several academically referenced and peer-reviewed essays and loaded them into his bot. Now, what it’s able to produce is atypical of marketing fluff. All created materials draw from information-rich, scalable content and little else.
Colby isn’t the only business leader who needed a GPT model tailored to his industry. Within just two months of OpenAI announcing GPTs in January 2024, users created over 3 million custom versions of ChatGPT. If your company isn’t yet one of them, it’s time to face the music: vanilla ChatGPT just isn’t cool anymore.
AI that speaks your language
According to April 2024 research from the University of Chicago’s Becker Friedman Institute for Economics, the industries that have most widely adopted ChatGPT include the technical sector, education and advertising — for software development, personalized learning and campaign optimization, respectively. But the wide range of custom GPTs offered by OpenAI is tailored to a wide variety of needs, including everything from outdoor trail recommendations to design tools to even providing what book to read next.
Colby’s son Cameron Olds, president and co-founder of Protagonist Communications, offers customized GPTs for brand messaging, allowing companies a better chance of being seen in a crowded marketplace.
“We asked ourselves, ‘How can we take these messages that last, necessary mile? How can we make sure people are being heard?’” Cameron says. “That’s what was hopefully going to generate the ROI for them and the results they were after.”
After looking closely at large language models (LLMs) with generative AI and ultimately creating their own fine-tuned bots, Protagonist Communications team members offer data specific to certain messaging parameters that are highly sensitive to unique information. A trained bot makes it easier for their clients to execute the work they do, and the bot generates quicker, more usable results.
[That’s] what the mind should be ultimately used for — creating all that hasn’t been created yet.
— Austin Larsen
Changing the model’s DNA makes it much better at writing on-message copy than normal ChatGPT. Outputs are more usable and customized to a company’s brand. What would take ChatGPT 5-10 prompts is accomplished in a single prompt. Clients can train their bot using specific parameters, allowing it to create and react to whatever is deemed necessary, Cameron says.
Boutique data sets allow teams to make small but strategic leaps that are more tailored to their industry or brand. Cameron believes this indicates a larger shift in attitudes about information gathering.
“We’re seeing a turn away from the need for lots of data to maybe not needing as much,” he says.
The efficiency equation
An early adopter of these custom models, Colby says using a customized bot has catapulted his company’s productivity forward without needing to increase his headcount. It’s revolutionized what his team has been able to do. Tasks that have taken weeks in the past are now completed in 90 minutes, and he’s seeing actionable results in days rather than months.
However, efficiency is only one reason to buy into this evolved way of messaging. As a former EVP of marketing, Colby says it was always difficult for his team to present original ideas from start to finish. From the time the project is dreamed up to the time it is finally created, communication often breaks down, and original plans end up diluted — not unlike the telephone game of yesteryear, where original messages are passed down from one person to another, often distorting them from what they were at the onset.
“It’s darn near impossible to keep the integrity of the idea intact,” Colby says.
While efficiency is important, consistency is monumental. The bot Colby uses for Nutrition Rescue allows him to keep all content created around his brand within a tightly guarded space, providing the exact positioning he’s trying to utilize. For him, that’s a game-changer.
“Now I’ve got a bulletproof document capable of standing up against scrutiny, and all of my marketing collateral gets pulled from it,” Colby says. “It’s a night-and-day difference.”
Outsourcing the ordinary
Another main benefit of leaning on a customized GPT model is that it allows business owners to use their imagination more, trading drudgery for regular creativity.
Rivala co-founder and CEO Austin Larsen is still training his bot to communicate how he wants it to, but he knows it will make his forthcoming startup highly efficient. Larsen sees this process as an alternative to onboarding and training a person or group of people to speak about his brand, products and services. It will make his work days more productive by taking over the parts of his business he’d prefer not to concern himself with and allowing him to focus on what’s essential.
“Most customer service requests are 90 percent the same: asking if a service will do one thing or another, how it does it and so on,” Larsen says. If his GPT model is trained on responses to repetitive questions, he believes it’ll free up his creative power to come up with new ideas designed to push his business forward.
“[That’s] what the mind should be ultimately used for — creating all that hasn’t been created yet,” Larsen says, underscoring the primary benefit of using a custom GPT. “Let’s create new stuff and outsource the rest.”