By 2025, three out of four CMOs say their company will be using generative AI for content creation, according to a recent study from IBM’s Institute for Business Value (IBV).
Now, after a year of experimenting and working in beta, IBM itself is publicly releasing its case study for using Adobe’s Firefly generative AI platform in its marketing and advertising content.
For its 2023 “Let’s Create” campaign, IBM put Firefly directly into its work process, using simple text prompts to generate 200 unique advertising assets and over 1,000 marketing variations for the campaign that took moments rather than months. More impressively, the campaign performed well above IBM’s benchmark, driving 26 times higher engagement, and reaching highly valued audiences (20 percent of campaign respondents identified as C-level decision-makers).
“It was a very high performing campaign, and a great use case of the technology because it played to the positive characteristics of Gen AI, and what it’s uniquely capable of doing, which would take a designer a lot of time manually to composite,” says Billy Seabrook, IBM Consulting’s global chief design officer.
Seabrook says that when embedded into other Adobe tools like Photoshop and Illustrator, Firefly was quickly able to dramatically accelerate productivity by speeding up early creative processes like sketching, prototyping ideas, storyboarding concepts, and actually expanding how IBM designers and creatives were brainstorming.
“We saw immediately, from an internal workflow standpoint, how it unleashed the volume of creative ideas that can be rendered quickly, and the acceleration of tedious production tasks like retouching and resizing,” says Seabrook. “And then, it actually expanded the design and ability to create visual work to a much broader audience, equipping a copywriter, for example, to play more in the visual creative process.”
Perhaps the most significant difference between Firefly and other generative AI tools is that it only uses Adobe’s stock image library and appropriate open source material. That’s a limitation compared to tools like ChatGPT and Midjourney, but much more legally sound when it comes to copyright.
“Having the content credentials, the identification, checks a big box for us in terms of legal comfort and using the tool,” says Seabrook. “Our next bit of concern is just the ethics and the bias around the content it generates, and that’s an evolving story. But we think that’s critical for brands to have an appreciation for before they go and launch something, that there needs to be a really good governance model in place to check for ethics and bias.”
In that same IBM IBV report, more than 42% of CMOs said scaling hyper-personalization is a marketing priority, and 64% said they expect to use generative AI for content personalization in the next year or two.
Seabrook sees that potential with Firefly, both in how it integrates and enhances his team’s current way of working and in how it’s evolving in terms of the type and quality of content it’s producing. “We’re creating the building blocks to make that a reality once we feel comfortable with the brand safety around the content,” he says. “I would argue within the year, you’re going to see a lot more campaigns with quality content going out that have been sort of curated properly.”