From production houses to prompt rooms: AI rewrites the agency workflow

Production was the fixed middle stage of every creative agency's pipeline for several decades. A brief left the creative floor, went into a production house's hands for casting, shoot, edit, and VFX, and came back weeks later as a finished final film. That sequence is now being compressed into a single desk. Creative teams are using generative tools to storyboard, previsualise, generate multiple cuts of a script, and adapt a hero film across languages and formats, often before a production house is even briefed.

The scale of the shift shows up in the numbers: 83% of ad executives globally now deploy AI somewhere in the creative process, up from 60% in 2024, according to Luma's 2026 review of advertising campaign data, and 86% of buyers are already using or planning to use generative tools to build video ad creative. The same research links AI-assisted creative development to a roughly 30% cut in production timelines, a figure that lines up closely with what Indian creative leaders are describing on the ground.


The efficiency question

Much of the industry conversation around AI has been framed as a trade-off: agencies can either be lean and efficient, or they can be bold and original, but supposedly not both. Neville Shah, CCO of FCB Kinnect and FCB/SIX India, rejects that framing outright.

“You can't not be an efficient agency if you want to be a great agency. You have to be efficient all the way through the chain,” he said, arguing that the real differentiator is agility and integration rather than a choice between craft and speed.

For Shah, the pressure AI has introduced is less about output volume and more about ambition: how well an agency reads its audience, and how quickly it turns that reading into execution. “It's like you don't get to decide whether your ad is good. Eventually, the audience does,” he said, a reminder that faster production tooling changes very little about how work is ultimately judged.

That judgement now happens against a backdrop where brands themselves are recalibrating what they expect to pay for. According to the World Federation of Advertisers, 58% of brands expect to pay agencies less once AI is deployed in their workflows, even as 80% of multinational marketers say they remain concerned about how their agency partners are actually using the technology on their behalf. It is this tension, cheaper and faster on one side, accountable and original on the other, that is forcing agencies to rethink what a production department is even for.


Inside the modular studio

Nowhere is that rethink more visible than in how much production has simply moved in-house. A joint study by the WFA and Mediasense, covering more than 30 global companies through 2025, found that all of the in-house creative teams surveyed reported using generative AI somewhere in creative production, with 71% already implementing it in specific workflow areas.

Archit Agarwal, Creative Producer at White Rivers Media, a Mumbai headquartered digital and social first agency, sees this playing out directly in how his own team is structured.

“AI is clearing executional bottlenecks, so more production now happens inside the agency itself,” he said. What does not get automated away, in his reading, is judgement: “What stays valuable is creative judgement, quality control, workflow management, and rights and compliance.”

Teams, he added, are becoming more technology-enabled and sitting closer to strategy, which for clients translates into faster turnaround and less rework, with brand integrity still owned by someone accountable. That does not mean external partners disappear from the equation, only that they get called in differently. Complex shoots, VFX and high-stakes brand work still route to specialists, but the timing of that engagement has shifted from being a default step in every job to a deliberate choice for specific craft.

“The change is in when partners come in. They will be engaged for specific craft, scale or quality needs,” Agarwal said, describing an ecosystem that has become modular, with coordination value sitting squarely inside the agency rather than the production house.


Craft, sameness, and the limits of the prompt

Not everyone in the production ecosystem reads the shift the same way. Sarah Thomas, Director and Founder of STOM Productions, a Mumbai-based film and content production house, is candid about where she thinks AI's usefulness currently ends.

“I see AI more as a powerful pre-production assistant than a replacement for production,” she said, pointing to visualisation, referencing and process simplification as its real strengths, while complex execution, technical challenges, realism and nuanced storytelling still require people.
Thomas flags a subtler risk too, one that goes beyond headcount: originality itself.

“When creativity relies too heavily on AI, a certain sameness begins to emerge because it is ultimately learning from what already exists,” she said, warning that an industry optimising purely for efficiency risks getting very good at producing the same ideas on loop.

Thomas also makes the point that the agency, production house and post-production boundary was already dissolving well before generative AI entered the picture, with agencies building in-house production capability even as brands increasingly bypass agencies to go straight to production houses for creative thinking, and production houses in turn absorbing post capabilities of their own. What AI does, in her view, is accelerate all three shifts at once, particularly at the low and mid budget end of the market, even as it introduces its own ceiling.

“No matter how sophisticated prompting becomes, you are ultimately working within what the machine is capable of producing, rather than having complete freedom to execute what is in a creative's head,” she said. That, she argues, is precisely where strong production partners become more valuable rather than less, because the job shifts from simply executing a storyboard to asking what the most exciting way to make something actually is.


Where the idea still lives

Strip away the disagreement over how much production shrinks, and the two camps converge on the same conviction: tooling is not the point, the idea is. Shah frames it as a question of relevance rather than scale.

“I think it's about relevance in terms of ideas. It's never about largeness,” he said, arguing that Indian creatives need to stop treating the brand film as the default unit of output and instead focus on getting work into conversation, wherever an audience happens to be watching, liking and listening.

Thomas arrives at a near identical position from the production side of the table. “At STOM, AI is already used across ideation, previsualisation and pre-production,” she said, “but we approach it as another filmmaking tool rather than allowing the capabilities of the tool to dictate the idea. The idea must always come first.”

Whether the production department, as Indian agencies have traditionally staffed it, becomes smaller, more specialised or simply restructured around new choke points, the direction of travel looks settled.

What generative tooling has removed is the assumption that every job needs the full pipeline. What it has not removed, on the evidence of creative leaders across the agency and production divide, is the need for someone with taste to decide what is worth making in the first place.