Beyond Stock: What Our Designers Have Learned Using AI-Generated Images in Real Client Work

Marketing leaders are watching AI rewire every part of the stack—from copy to media buying to creative production.

The visual side is also changing rapidly. After two years of putting AI-generated images to work on real B2B client projects, here is what we have learned as a design team, and what it means for the brands we serve. (For the broader view, see our take on the AI tools reshaping B2B marketing in 2026.)

Stock photography is on its way out

Generic stock has always been a compromise.

The same handshake. The same boardroom. The same vaguely diverse team pointing at a laptop.

For B2B marketing, that compromise is getting harder to defend—buyers tune it out, and it does nothing to differentiate the brand.

AI is replacing stock photography because it lets us build custom brand imagery on demand. We can match a specific industry, a specific moment, a specific tone—without scrolling through 400 nearly identical photos.

Even Getty and Shutterstock are leaning into generative tools, which tells you where the category is headed.

How we are actually using AI in client work

Most of our work with graphic design AI falls into three buckets:

• Extending existing photography. Expanding backgrounds, reformatting for new ad sizes, or generating variations from a small set of brand-approved photos.
• Building conceptual visuals. Abstract campaign imagery, mood pieces, and composites that would have required a full photoshoot three years ago.
• Accelerating ideation. Pitch concepts and early mockups that move a client from “I think I want this” to “yes, that” in a fraction of the time.

For marketing teams, the impact of AI image generation and AI-assisted design shows up where it matters: faster turnarounds, lower production costs, and creative that actually reflects the brand instead of stock-photo shorthand. That changes the math on how marketing budgets get allocated.

The work is in the human direction

Here is the part the AI hype tends to skip. Designer AI tools do not replace designers—they raise the value of designers who can direct them.

Random prompts make random outputs.

A real creative AI process starts with art direction and brand fluency—someone shaping the input, refining the output, and integrating it with everything else on the page.

We rarely send a raw AI image. We retouch it, layer typography, adjust composition, and pull it into the client’s palette and visual system. That step is what produces genuinely brand-focused visuals, and it is the same instinct behind our broader definition of UX: every visual choice is part of the experience.

What we have learned about AI-generated images

• Specificity wins. The better the prompt, the closer the first output. Reference styles and context matter as much as the subject.
• Mix the tools. Pairing AI image prompting with Photoshop, Illustrator, and generative engines like ChatGPT, Claude Design, and Gemini produces better results than relying on a single platform.
• Iteration is non-negotiable. First outputs are starting points, not deliverables.
• Brand standards are the test. If a visual cannot pass for something the brand would have commissioned, it’s not done.

Those lessons sound simple. They are also what separates useful generative AI for designers from the slop flooding LinkedIn feeds.

What the shift means for marketing leaders

If you run marketing at a B2B company, two shifts are worth tracking. First, AI in creative workflows is making custom visuals affordable at a scale that used to require a much larger budget—which has real implications for how you build your funnel.

Second, the teams winning with AI for marketing design are the ones treating it as a craft, not a shortcut. McKinsey’s 2025 State of AI report shows marketing is one of the fastest-adopting functions—the gap between teams who use it well and teams who do not is widening fast.

The strongest results we see still come from human creativity leading the work. AI can speed up execution, but it still needs clear thinking behind it.