5 Real Examples of Enterprise AI Image Generation Use Cases

Moving beyond hype to examine how brands are actually using image generators.

Image via Poplar Studios.

One critical component of all hype cycles – including our current generative AI moment –  are promises and use cases based on what’s theoretically possible. However tantalizing, possible use cases do not necessarily meet enterprise adoption needs, and that’s true of AI image generators, too. Rather than talk theory, let’s examine how real enterprise marketing departments and practitioners are actually using AI image generators. 

This article will not compare platforms offered by Midjourney, Stable Diffusion, DALL-E, Firefly, Getty, Amazon, Synthesis X, and others. Instead it will examine five real use case examples of AI imagery as executed by brands and agencies. 

Hopefully, reviewing real examples will allow the industry to focus on proven opportunities. Use cases that have moved from ideation to public implementation offer a different perspective to brands weighing possible AI pilots. Rather than merely considering what's theoretically possible, brands can take their cues from real world, practical examples.

This is particularly true of brands new to AI who are weighing their first enterprise AI implementations. Experienced brands will still want to identify and deploy AI use cases that meet their business needs rather than replicate or evolve already existing public examples. After all, AI technologies work best when they serve an enterprise’s strategy. Tech for tech’s sake is not a sustainable approach to adoption.

In most of the use cases, risks presented by AI image generation such as copyright infringement, data security, ethics and bias have been addressed or thought through. However, brands should not blindly engage in AI pilots without researching legal indemnification for intellectual property infringement, data security mechanisms, and human biases found in training data or use of the AI tools.

In that vein the following examples are just that, examples. They are not endorsements of the “correct” use case or the best AI vendors. Public use cases rarely show the metrics that are used to weigh their impact. Instead, view them as examples of AI pilots that made it to the public eye for discussion.

Use Case 1: Storefront Imagery

Image via AdAge/UnderArmour.

Virtual storefront and other advertising uses of generative AI help brands sell their products. To help them achieve new sales, several brands including UnderArmour, Walmart, and Levis have used generative AI to replicate variants of clothing and product lines in different available colors for their storefront.

The immediate cost benefit is obvious. Instead of having to orchestrate a photoshoot for each and every product variant, a brand can hire a singular or a handful of models, capture initial images, and then use AI to “extend” the creative experience to the entire suite of product options. 

In the case of UnderArmour, they hired creative company Tool’s AI service to train on the brand's "visual DNA" encapsulating their clothing’s style. UnderArmour used generative images to create realistic new product and model images, reducing the need for photoshoots. With some quick touch-ups, the AI-generated assets are deployed across the company’s website and advertising channels, allowing UnderArmour to iterate on their visual assets.

Levis deployed visual AI for a similar purpose. Unfortunately, Levis touted its AI use as a means to offer more racial and cultural diversity in its models. The brand quickly came under fire for not simply hiring a more diverse set of models, in effect taking money out of the pockets of an underrepresented population. This ethics gaffe was a strategic public relations mistake rather than a technological one, but nevertheless demonstrates how bias can tank an AI implementation.

Amazon announced the availability of similar AI capabilities for its store users, empowering thousands of small and medium businesses to better represent their products on the retailer’s online marketplace. Amazon touts its Ad Console image generation service as a means to deliver lifestyle- and brand-themed images. 

We like Amazon’s concept in the sense that it democratizes generative AI for small and medium retail enterprises, making the technology available to many brands who probably could not easily find, train and deploy these tools. Ad Console GenAI is demonstrative of using technology to serve an enterprise’s business model (online retail marketplace) rather than building tech for tech’s sake.

Additional examples of generative AI in retail include virtual try and buy tools that transpose a product onto a buyer, in their buildings, or other physical being. A wide swath of retail brands are adding this to their virtual storefront including Decker Brands’ Hoka, eyeglass brands provider Roka, Ulta cosmetics, West Elm room designs and Walmart retail sharing with friends. 

Use Case 2: UGC Consumer Activation Campaigns

Image via Dress X.

User generated campaigns like Doritos' Super Bowl Ads have been a hallmark of brand campaigns since the launch of social media. Image generators have further empowered consumers to generate new brand content. Coca-Cola and Heinz have deployed two of the most well cited UGC image generator campaigns out there. But they are not the only ones.  

Tommy Hilfiger deployed a similar campaign for 2023 AI Fashion Week, with consumers and fashion aficionados designing their own outfits. After determining the winning design, Hilfiger rewarded the creator and their design with a place on the online virtual fashion platform Dress X, where consumers could try on outfits, an AI effort from end to end. AI Fashion Week encourages designers to incorporate AI into their design and production processes.

Engaging customers and and larger communities in contests and challenges for unique AI imagery is smart. A small percentage of any user community has its artists and some of them are surely using AI tools. However, these images do need to be vetted for copyright issues, and submissions must include licensing rights. We discuss copyright issues more in depth after use case three.

Full Disclosure: CognitivePath offers a workshop and advisory services to help brands develop their use cases. Learn more on the CognitivePath website.

Use Case 3: Agencies and Services

Image via Hive3

AI has triggered innovation amongst third-party service providers like agencies and graphic design marketplaces like Fiverr, that offer AI-enhanced vendors. Some unique service providers have launched to offer these AI design services as a part of their core DNA or in some cases as the primary offering.

Hive3 is a new contest marketplace where brands can deploy challenges to the site’s “league” of AI designers. Brands use Hive3 to sponsor image and video design challenges. AI design aficionados offer their best campaign ideas, and 1st, 2nd and 3rd place winners are awarded accordingly. Hive3 offers brands a lower risk way of engaging in AI design projects (listen to the No Brainer podcast with Hive3 Founders Adam Brotman and Andy Sack).

AIimagination takes the AI image generation concept a step further with a “machine art generated” agency. The creative boutique, launched by international advertising agency Fred & Farid, uses AI image and video generation to cater to CMO needs for “speed, cost, and efficiency”. Clients include The North Face in China, cognac brand Louis XIII in France and climate group Fridays for Future in the U.S. 

Whether Hive3 and AIimagination represent the creative model of the future remains to be seen.  Agencies and service providers are likely to offer differing degrees of hybridization of human/machine creative services. Large agency conglomerates like Publicis are also investing millions of dollars into hybridized AI offerings. As with every technology cycle, we see These new offerings do offer brands a lower-risk means to experiment with AI.

Use Case 4: Product Packaging and Design

Image via RetailDetail.

Offering unique variants of package design and magazine covers are a marketing tool used by brands and publishers for decades. The age of AI has made creating variants easier and scalable. AI Image generator tools have been used with pre-programmed fonts to offer designers the ability to generate dozens of logos and images with one click of a button. Brands like Beck’s and Coke both have succeeded in using AI packaging (ANA will soon offer its members access to whole library of AI case studies, including Coke and Beck’s, prepared by CognitivePath Analyst Greg Verdino).

Nutella took package variants a step further than the norm with its Unica project. Parent company Ferrero used an algorithm trained on patterns and color palettes to generate 7 million unique packaging designs for Nutella jars.  Each jar had its own authentication code and was called a piece of art. 

Ad agency partner Ogilvy & Mather Italia worked with Nutella to deploy an advertising campaign selling the product throughout the country. While the initiative generated quite a bit of publicity for its use of AI, it's illustrative on how AI can help brands create varying and special themed package designs for marketing purposes. As we have seen with brands like Stanley, limited release color ways and packaging help create consumer demand.

Use Case 5: Stock Image Generation and Licensing

An Adobe Stock image by By IBEX.Media created with AI.

Perhaps the easiest and most obvious use of image generators is to create stock images for a wide variety of use cases. Indemnification and copyright issues, ethics, quality and bias have prevented some brands from using images. While these are serious issues that must be accounted for in every AI use case, indemnification, copyright, ethics and bias can be resolved with strong governance and procurement policies. 

Quality issues such as humans with too many fingers have been a constant detractor against the diffusion models used to create AI images. Various workarounds have addressed many of these issues, from touch ups by humans to algorithm improvements to designers who have mastered interfacing with the tools to coax the desired results from the AI image generator. Quality is expected to improve with each new diffusion algorithm model deployed.

Several stock vendors have expanded to include AI images, resolving brands of the concerns presented by AI. Adobe Firefly, Getty Images, and Shutterstock trained their image generators using their massive portfolios of stock imagery. These images provide the training data necessary to fuel AI diffusion engines that will produce images without copyright issues. 

While the results have been criticized for looking too much like stock photography, the results can be impressive. In fact Adobe, Getty and Shutterstock offer AI images created on their engines for license. Further, brands who use the services are also indemnified from copyright infringement. 

Adobe has integrated Firefly into its full creative suite allowing graphic designers and photographers to use the service as a tool within their larger software apps. And all of the large cloud providers offer brands the opportunity to create private instances of image generators.  Using AI to scale creative isn't the domain of third-party providers, but it does require resources and training.

Adobe, Getty Images and Shutterstock are not the only image generators and stock image providers indemnifying brands for using them. For example, Bria and Google joined them in offering indemnity-free AI images.

Client indemnification is becoming the standard for all image and video generators; however, not all image generators have subscribed to this. Most notably, Midjourney does not indemnify users from third party’s rights when it comes to images. Midjourney is often at the center of image copyright issues related to others’ past works. 

Part of a brands’ AI governance and procurement initiative should vet diffusion models and service providers for indemnification and ethical use of imagery. Internal usage guidelines should also make sure quality controls are in place to vet images that may look like others’ works . If the image looks too close to a prior original work for it to truly be unique, then it probably isn’t and should not be used.

Conclusion 

One common theme emerging in the various use cases is scaling creativity. From the ability to offer image examples of all products to scaling product design to producing wide portfolios of stock imagery, AI allows creative teams to iterate and re-iterate content on a scale unimaginable to the design teams of even five years ago.

Time to produce these images is another critical factor brands. AI allows companies and creatives to develop professional-grade imagery within days and even hours. In a world where product launches can take months due to the need to not only design and manufacture, but also build marketing campaigns, AI imager can help companies shave weeks off their production schedule.

Much has been said about AI content lowering the quality of content, including imagery. There are real imagery issues with AI. However, it is clear professional artists can use AI tools to develop professional works worthy of top brands. It’s how the tools are used.

Brands who incorporate generative imagery into their creative works need to consider legal implications, too. Whether internally or through acquisition, companies need to vet partner and software contracts to indemnify the brand. Usage guidelines for internal employees should center on proofing work for qualities and to protect artists’ intellectual property rights and the brand’s reputation.