Which Projects Are Not Right for AI?
AI doesn't solve every job, and few suppliers will say so. Human performance, material fidelity, documentary value and regulated categories — in four areas a real shoot is still the only right answer. Where not to use AI, and why.
- Work carried by human performance — emotion, micro-expression, spontaneity — shouldn't be generated.
- Where material and texture fidelity decide the sale (luxury, textile, food), a real shoot is required.
- Documentary content — news, real places, real events, customer stories — cannot be generated.
- In regulated categories (health, food claims, finance) the risk cost outweighs the saving.
Most writing about AI production covers what's possible. The more useful information for a brand runs the other way: what's possible but wrong. This piece sets out the four areas where we don't propose AI, and why.
First: work carried by human performance. If the emotion of a commercial rides on an actor's micro-expressions, that job is shot. AI can generate a human face; what it can't carry is the consistency a viewer reads in a real performance. Wherever emotion is central, that gap shows on screen.
Second: material and texture fidelity. In luxury, textile and food, the purchase decision is made on the realness of texture. The fall of a fabric, the grain of leather, the moisture on food — generated, these come out 'almost right', and almost right is wrong in a luxury category.
Third: documentary content. News, real locations, real events, customer stories. Here the image's value comes from being true; generating it zeroes that value out. This is an ethical boundary rather than a technical one, and it's the most expensive one to cross in brand-trust terms.
Fourth: regulated categories. Health, food claims, financial products. In these areas an image that doesn't reflect reality isn't just a reputation risk, it's an enforcement risk. The time and cost saved is irrelevant next to that exposure.
Frequently Asked Questions
Which projects are not right for AI production?
Four areas: work carried by human performance (emotion, micro-expression, spontaneity), work where material and texture fidelity is critical (luxury, textile, food), documentary content (news, real places and events) and regulated categories (health, food claims, finance).
Can AI replace an actor in a commercial?
Technically a face and motion can be generated, but where emotion is central the result falls short — viewers read the micro-consistency of a real performance. AI is strong in product-led films with simple narratives; it is not where performance carries the film.
Why is AI risky for luxury products?
In luxury, the purchase decision is driven by material perception: leather grain, metal reflection, the fall of fabric. AI produces these approximately, and a luxury buyer notices the approximation. Production in this category shouldn't happen without a reference shoot.
What do you do on jobs where AI isn't appropriate?
Shoot for real and use AI only as a support layer: format multiplication, background variation, social adaptations. The question isn't 'AI or shoot' — it's 'where does AI enter'.
Why does it matter that a studio says this out loud?
Because a supplier who proposes AI for everything is transferring risk to you. A team that can name what isn't suitable at the start of a project is removing problems before they appear. It's one of the most revealing questions to ask during selection.
At Pam AI Studio we make this call at the start of a project. We don't propose AI for every job — which frames sit on set and which variations run through the pipeline is separated at brief stage.
Let's work out which part of your project is right for AI.