Generative AI is changing where the creative work happens in video production. Tools that draft scripts, generate references, and render shots in seconds compress parts of a project that once took weeks, putting more weight on the choices made before any model runs. The teams getting the best work from AI are the ones giving it the clearest direction.
Roel Leijten is the Studio Manager at SAP, where he leads video production at the company's London headquarters. His team produces corporate broadcasts, webinars, and on-location content for one of the world's largest enterprise software organizations. Before joining SAP five years ago, Leijten spent more than a decade as an independent filmmaker, and in 2020 wrote, directed, and distributed his own feature film on Amazon Prime. That dual vantage shapes how he reads AI's effect on creative work.
"Agents are going to be your new kind of workforce. They'll work like a colleague, but they still need human direction," says Leijten. His view echoes a change that creative leaders across video formats are starting to feel. As execution compresses into prompts, the ideation, references, and briefs feeding the tools carry more of the eventual result.
For Leijten, large language models have become a starting point for early creative work. He uses them to draft scripts and develop ideas, and he encourages collaborators to do the same before bringing him into a project, giving everyone a clearer picture of what the team wants to make.
Visual references benefit, too. Leijten points out that prompted images make it easier to communicate aesthetics, characters, and overall look and feel with more precision than the old habit of pulling stock photos. For teams passing briefs between strategists, creatives, and production partners, that specificity narrows the gap between what someone describes and what the next person actually builds. "You can really easily show the aesthetics you want in your film, the characters, the look and feel. Before, you often had to use old pictures from a random website, but now you can get really specific. The better you can translate your vision on paper, the better it is," Leijten adds.
The next phase of AI in production, in Leijten's view, moves from assisted ideation into delegated execution. He expects agents from companies like Adobe to take on more of the multi-step work in editing, design, and post-production. For ad-tech buyers building out video pipelines, that prospect changes how a creative team's roles and headcount get drawn up. "Taste, storytelling, and original thinking still need to be fed into the AI. No doubt it's going to be a mix of humans and agents working together," he says.
How a team frames the relationship shapes the output. Agents working without clear instructions tend to produce average results, while agents working against defined intent retain the creative signal of whoever set them in motion. Teams that invest in tighter briefs get more from their tools. That dynamic grows more pronounced as production costs continue to fall. When a polished spot can be generated in a fraction of the time and budget it once required, what separates one piece from another lives in the direction given to the tool.
Developing the kind of creative judgment AI demands calls for a notably low-tech habit. Leijten encourages collaborators in design, video, and marketing to read deeply about scriptwriting. Narrative principles that govern a feature-length film translate cleanly to a 30-second spot, a connected-TV ad, or a corporate broadcast. "Scriptwriting principles apply across marketing and content, whether the storyline runs 30 seconds or an hour and a half. The components are the same. I always tell people to read widely. It's hard to get people to read in this digital age, but it's the one thing that can really make you stand out," he says.
His advice runs against how most creative skills get taught today. Short-form tutorials and platform trend cycles dominate the learning curve, with both formats rewarding replication over invention. AI in production rewards similar habits, leaving teams that lean on familiar formulas to compete in crowded territory.
"You can always get clones, and people love to reproduce the success of others. But the real hits come from people coming up with their own ideas, their own taste, their own uniqueness," Leijten says. For ad teams thinking about how to train the next generation of creatives, AI raises the value of inputs it cannot supply on its own. A well-read team brings sharper thinking to its briefs, and that depth carries through in the work that follows.
The other piece sits in the systems running underneath production. As AI lets teams generate more work faster, the odds of duplicating effort rise alongside it. At a company the size of SAP, a creative team in one country can spend days building an asset that another team has already produced somewhere else, and neither side ever finds out. "If you had a big second brain, a shared search database, you'd already know the asset is there and could work together. Having that system in place is key going forward," Leijten says.
A similar dynamic plays out at ad teams running parallel video work across regions and brand portfolios. Without shared metadata, searchable libraries, and clear handoffs between departments, fast-moving teams end up duplicating each other's work. The teams getting real leverage from AI are treating content infrastructure as a creative resource, letting work set in motion by one group build on what another has already produced.
Knowing what to make remains the part AI has not absorbed. Briefs, references, story sense, and the institutional memory of what a team has already tried are the inputs that turn cheap execution into something worth watching. The AI investments paying off are the ones where teams put as much care into shaping inputs as they do into chasing speed. "The real hits are people coming up with their own ideas, their own taste, their own uniqueness. I foresee everyone doubling down on that in the future," Leijten says.
