Significantly, where the brief allows for it. Modern production compresses traditional stages while holding the creative output, which moves a brief to finalised assets in a fraction of a fully traditional process. For brands operating at the speed of culture, that responsiveness has become a competitive requirement.

Speed serves the work. The time saved upstream goes back into direction, iteration and craft.

In most cases it makes production meaningfully cheaper. A set that once required a full art department, a build and a strike can be generated, or built physically only for the shots that require it. The same logic applies to environments, crowd extension and product variants.

The savings also change what becomes possible. Multiple environments, large-scale visual worlds and dozens of campaign variants move inside reach of a brief that once afforded a single execution. Cheaper is part of the answer. The bigger part is what the budget can now do.

‍

Human performance is the clearest case. The unscripted glance, the live chemistry between performers, the involuntary detail nobody planned: a real shoot delivers those. Dynamic product-in-motion work sits in the same territory, from liquid pouring to fabric moving with a body to food prepared in real time.

When the brief lives in those spaces, we shoot.

When part of the brief belongs in a camera and part belongs in generation. Performance gets shot. Environments, crowd extension, product variants and seasonal versions get generated around it.

One shoot day then carries a campaign across markets, formats and seasons, with the budget concentrated on the material that earns a camera.

AI is most effective where its constructed nature forms part of the appeal: abstraction, transformation, speculative worlds, impossible environments, symbolic and narrative world-building. Generative work amplifies imagination in that territory. Hyper-real simulation of documentary or lived experience is the fragile end.

On training data: We do not prompt with direct references to other creatives' work. References inspire, the way they always have, but we never feed another artist's images to AI as something to copy. Every project carries a creative signature: our own art direction, or the vision of the artist we are working with. We actively seek out AI artists who have a visual perspective of their own, not the ones who are skilled at imitating an existing style. That choice does real work to sidestep one of the core ethical problems with generative tools.On the people who appear in the work: When we generate human faces, we run them through similarity checks (tools like FaceCheck.ID, PimEyes, and platform-level safeguards) before anything reaches a client, to flag any drift toward a real person's likeness. That said, we cannot guarantee a generated face is 100% unique, which is why our actual preference is to work with real models under proper consent agreements that explicitly cover AI-generated extensions of their likeness: additional coverage, lip-sync localisation, scene extensions. The generated layer sits on top of a consented performance, not in place of one.On bias: Cultural bias in AI output is not a model problem we hand off to better software; it is a creative decision made at the art direction stage, every time. It is addressed in how we brief, how we prompt, which references we choose, and (just as importantly) what we actively reject. Experienced creatives carry that responsibility into the work. The model does not.On the people working in the creative industry, whose roles are changing: We will not pretend the disruption is not real. What we can say is that the AI artists we work with came almost entirely from traditional production: retouching, photography, motion graphics, CGI. The roles are changing shape, not disappearing. The craft transfers; the workflow adapts. That is the more honest picture than either "AI takes the jobs" or "nothing has changed.AI without ethics is a shortcut. Modern Production is not.