Sapient Slingshot wins Technology Breakthrough of the Year.
Sapient Bodhi ranked #1 for deep research.
Meet Sapient Sustain: context-aware AI for complex IT operations.
3M lines of COBOL → clear specs in 8 weeks.
October 12, 2026
Loading AI-generated summary...
Finding an image sounds like a small part of the creative process—until it takes hours. For a long time, creating imagery for everyday projects meant searching approved stock libraries for something distinctive enough to feel like our brand and then editing it to make it work.
Even then, there was a risk that the same royalty-free image could appear somewhere else, including in a competitor’s content. Generative AI offered another option, but it struggled to consistently create images that looked and felt like Publicis Sapient.
So I built a reusable AI workflow around our brand with a custom GPT.
My teammates and I experimented with several AI image tools, including Adobe Firefly and ChatGPT. The biggest challenge was consistency. Different prompts could produce dramatically different results, and recreating the right brand look meant repeatedly explaining the same requirements. We were effectively starting over each time.
I saw an opportunity to centralize that knowledge. I created a custom GPT informed by our updated brand guidelines, mood boards and photography direction. Instead of asking every designer to independently prompt their way toward the brand, the tool gave us a shared starting point for generating imagery with the right lighting, color and overall visual feel.
The goal was to make our creative direction reusable so people don't have to recreate it every time.
The idea really took shape while multiple designers were working on a website project at the same time. We were experimenting with AI imagery separately, but we were essentially trying to solve the same problem over and over. The custom GPT gave us one place to capture what we were learning.
It did not work perfectly at first. Some early images looked unmistakably AI-generated. Others were unrealistic or disconnected from what I had requested. I refined the rules and instructions, adding more precise language around qualities such as realism and candid photography.
The process became iterative: generate, evaluate, adjust, try again. Over time, what started as an experiment became a repeatable creative workflow.
Making image generation faster does not make creative expertise less important. I have found almost the opposite. Ask AI for a car on a road, and it can generate one. But that does not mean it will be the image you actually need. Give it more precise art direction—such as a bird’s-eye view, a wide-angle lens or a specific composition—and the output gets much closer to the intended vision. That means knowing how to describe an image technically and knowing how to evaluate what comes back still matter. Not every generated image should be used. I still need to decide whether the lighting works, the composition makes sense, the details feel believable and the result is appropriate for the project. AI can generate, but I still have to direct and decide.
Using generative AI meant we didn’t have to rely on stock libraries as much. We estimate that the time spent searching for and creating imagery has decreased by roughly 85 percent. Purchased imagery has also decreased 71.4 percent compared with 2025.
More people started using it, too. The team started more than 180 chats with the custom GPT. Because each chat can generate multiple images, the actual volume of imagery created is higher.
For me, the bigger benefit is what those numbers give back. We can spend less time looking for assets and more time actually designing.
Designers were already experimenting with generative AI and trying to move outputs closer to the brand, but we were doing that work independently. By putting shared rules, creative direction and brand knowledge into one reusable workflow, I was able to turn individual experimentation into something the wider team could use.
That lesson is bigger than the tool itself. AI platforms will keep changing. What matters is recognizing when people are repeatedly giving AI the same context, applying the same rules or correcting the same mistakes. That is usually a sign that the knowledge can be turned into a reusable workflow.
What I built for an internal creative need reflects a larger challenge facing brands. Generating more visual content is becoming easier, but generating it quickly while maintaining brand standards isn’t.
Publicis Sapient is addressing that challenge for customers through Sapient Bodhi, our agentic platform, and its Visual Brand Assistance capability. Beyond simple image generation, its value comes from surrounding that capability with the rules, context and human expertise needed to make the output useful. The solution can help turn photography guidelines, governance rules and regional requirements into a reusable brand profile, support teams in finding approved imagery, generate new, brand-informed campaign visuals through Adobe Firefly and score imagery against brand rules before approval.
For one global consumer goods company, Bodhi supported content workflows spanning campaign concepting, copy, SEO, product content, lifestyle imagery, video scripts and asset resizing. The organization created more than 700 assets in two months, achieved 60 percent reuse across brands and reduced production cycles from weeks to days.
The scale is different, but the idea is similar: capture the knowledge that matters, make it reusable and use AI to speed up production without removing human judgment.
What I built did not eliminate the creative process; it just removed some of the friction around it. The brand guidelines provide the boundaries while AI provides a faster starting point and I still supply the art direction, evaluate the result and decide what is good enough to use.
That combination has helped us reduce our reliance on purchased imagery, dramatically cut the time spent searching for and creating images and give designers more time for the work that requires judgment. For me, that’s the real value. AI may make that process faster, but creativity still depends on the person directing it.
