• by Sam Lewis , January 21, 2025

When it comes to creating with AI, there’s a certain magic in seeing a simple text prompt come to life in the form of, say, a collegial video of Presidents Biden and Trump enjoying ice cream together.

Yet the more you use it, the more you realize that for many tasks AI doesn’t quite get you all the way there. An AI-produced Coke bottle that’s 90% right is still 100% wrong from a brand perspective. Like many “last mile” problems, it’s solving for that last 10% that requires the most investment and human brainpower.

Despite these limitations, AI remains a remarkable creative and business instrument. But to get the most out of it, brands must approach with the right mindset. Following are tips for getting started:

Matching use case to business advantage

AI presents an astonishing number of use cases. A global CPG company could use the technology to instantly focus-group campaigns against synthetic audience panels, saving the time and expense of more traditional research. Or a luxury brand could reduce reliance on expensive location shoots by using AI to recreate far-flung locales.

Central to any potential application is ensuring that gains on one side of the equation aren’t offset somewhere else -- for instance, using AI-generated video to reduce production time, only to end up investing more on the post-production side making manual adjustments to meet quality standards.

Accurately assessing where this balance lies will become an increasingly important skillset, requiring practitioners who can blend business, creative and tech expertise to make strategic AI decisions.

Train, Train, Train

AI is only as good as the data you feed it. While off-the-shelf models can produce impressive, almost-there results, the secret behind ready-for-prime-time work is training. This involves first uploading all relevant data -- brand guidelines, audience definitions, creative concepts -- into the platform and then providing feedback on the outputs it’s producing, shaping responses to consistently align with specific goals and values.

Even with identical training, different models can produce outputs with significant variation. For a recent activation, we found that Claude was consistently funnier, while ChatGPT was more brand-safe. Understanding these distinctions enables marketers to choose the best tool depending on assignment.

Cultivating the AI / human interface

Like managing an eager but inexperienced intern, successful AI implementation requires workers with the know-how to optimize its energies to tasks where it will be strongest. That means training the humans in the room on effective AI stewardship -- with the difference between poor- and high-quality users magnified by the power of the platforms.

With similar AI tools broadly available, this human factor is a major difference-maker. Cultivating this expertise starts at the organizational level, with investment in AI literacy programs and an emphasis on making AI the purview of everyone, not just a select group of internal experts.

AI as a new creative medium

Finally, AI isn’t just about completing existing tasks more quickly and cheaply. It’s finding ways to accomplish things we haven’t been able to do before

Where AI can be particularly useful is in the realm of real-time hyper-personalization. For example, a fitness company could dynamically tailor its video workouts in response to a person’s heart rate, blood pressure and oxygen saturation. To support a new movie, a studio could create an AI clone of one of the characters for fans to interact with or receive personalized video messages from, deepening engagement and creating potentially millions of pieces of distinct, shareable content.