AI Was Supposed to Handle It. The Deadline Arrived. Now What?

There's a moment every art director or agency producer has already lived — or will soon. The brief arrived, the client approved the concept, the deadline is locked, and the question nobody wants to ask out loud is: will AI actually handle this or not?

For most of 2024 and 2025, the default agency answer was: yes, it will. Client pressure for cost reduction and agility pushed AI to the center of the creative process — often before any real testing of what it could actually deliver in each job's specific context. The results started showing up in numbers and cases nobody volunteers to share.

Cost-reduction pressure pushed AI to the center of agency creative processes — often before any real testing of what it could deliver in each specific context.

What the numbers are saying about agencies and AI

The pressure isn't perception — it's structural. WPP, the world's largest communications group, saw its shares fall around 60% over 2025, driven by client losses and fears that AI was eroding margins . Omnicom, Publicis and IPG all felt the impact. The market is signaling that the traditional agency model is under real pressure.

On the client side, the movement is clear: brands are increasingly building internal AI teams, bypassing agencies for routine tasks . And the expectation that's been created is direct — if AI cuts production cost and time, the service price should drop proportionally.

The problem is that this logic ignores something fundamental: only 13% of consumers trust ads created entirely by AI, compared to 48% who trust pieces co-created by humans with AI support. The efficiency AI delivers in process doesn't automatically offset the credibility loss that a visibly machine-generated piece can cause to a brand.

The case that became a symbol — and what it really teaches

Coca-Cola is the most documented example of this tension. The 2025 holiday campaign was generated entirely by AI for the second consecutive year — this time featuring polar bears, pandas and sloths admiring the brand's trucks, an attempt to avoid the human face problems that marked the previous version. Technically more refined, it still drew criticism for its artificial quality .

The campaign was produced with 20 professionals instead of the 50 traditionally used — prioritizing cost savings despite backlash over visuals described as "uncanny" and lacking soul . Executives defended the decision as forward-thinking. Consumers called it "slop" — a term that has become shorthand for low-quality AI-generated content.

What this case teaches isn't that AI is bad for advertising. It's that efficiency without technical and creative judgment has a brand cost — and that cost rarely appears in the spreadsheet that justified the AI decision.

Efficiency without technical judgment has a brand cost. The problem isn't using AI — it's letting it make creative decisions that only a specialized human eye can properly evaluate.

Two real cases: when the problem arrives at the studio

Global cases are revealing — but what happens day to day in agencies is equally representative. Two recent jobs that arrived at Kado show precisely where the process breaks.

Case 1 — The text AI didn't know how to write

February 2026. An agency developed advertising material that involved technically precise text — not creative headlines, but content that needed to be reproduced exactly in the final layout. The art director built the concept, the client loved it. The logic adopted was the same as previous jobs: if everything else had been AI-generated up to that point, why would this one be different?

Because this was one of those cases where "close enough" doesn't work. The artwork contained text that needed to appear legibly, faithfully and technically correctly in the final piece. AI was useful for building the layout — but the text it generated was invented, deformed or simply wrong. Any text wouldn't do: it needed to be that specific text.

The result was complete rework. The deadline that had seemed comfortable stopped being comfortable. The budget that had been calibrated for an agile process needed to be renegotiated. And before any of that, there was a difficult conversation with the client — explaining why the concept-approved material couldn't be delivered as planned. In the end, the work was beautiful and the client loved the result. But the path to get there cost more than it should — in time, money and relationship strain.

Case 2 — The image that couldn't be fixed

January 2026. A studio developed a complete job with generative AI: models, clothing, scene, light, interaction — everything integrated. The on-screen result was impressive. Client approved immediately. No one needed to travel, no risk of bad weather on shoot day, no model or photographer bookings. AI had handled everything.

The problem appeared when the material arrived at Kado for finishing. The images came out at AI's standard output — 2,000 pixels. For professional campaign-quality use, the minimum required is 8,000 pixels. Upscaling resolves part of the problem: the scene, the background, ambient elements can be enlarged at acceptable quality. But the product — the clothing, the campaign's central subject — is the hero. And the exclusive texture of that product, what makes it identifiable and faithful to what the consumer will purchase, AI cannot invent. It's not possible to enlarge something that never existed with sufficient fidelity for commercial use.

Kado didn't take the job. We analyzed the material, identified that there was no viable technical solution within the quality requirements that professional work demands, returned it to the studio and explained exactly why. There's no responsible way to deliver a result that will disappoint the end client — regardless of how much deadline or budget pressure is involved.

Not every problem has a viable technical solution. Identifying that before accepting the job — and being transparent about why — is what separates a serious studio from one that accepts everything and delivers anything.

Artistic quality and technical quality — and why confusing the two is expensive

There's a distinction that's rarely explained clearly in image production processes — and that sits at the center of most problems that arrive at studios. Every image has two types of quality, and they don't measure the same thing.

Artistic quality is what the eye evaluates on screen: the beauty of the composition, the energy of the scene, the message the image conveys, the creativity of the concept, the proportion between elements, the expression of the moment. It's what makes the client say "I love it" in the approval meeting. It's subjective, immediate and real — an image can have genuine artistic quality and still be unusable for its intended purpose.

Technical quality is what parameters measure: resolution in pixels, file size, real versus apparent sharpness, color profile calibrated for the application medium — RGB for screen, CMYK for print — information density per square centimeter. It's what determines whether the image holds up when enlarged to billboard size, whether colors will reproduce faithfully at the printer, whether the product will appear with the texture the consumer expects to find when they open the box.

Generative AI, in most cases, delivers impressive artistic quality — and insufficient technical quality for professional use without additional processing. The problem happens when approval is made based on the first and delivery is judged based on the second. That gap between what was approved on screen and what can actually be delivered at final size and quality is precisely where jobs break down — and where rework begins.

The pattern appearing in jobs that arrive with problems

The two cases have different structures but share the same root: the decision to use AI was made before evaluating whether AI was the right tool for that specific job. It's not an execution error — it's a diagnostic error.

"AI-generated content can lead to unrealistic expectations," says a creative director with two decades of experience on campaigns for Apple, Samsung and Coach. "There's a lot of education happening between clients and agencies. Doing it in real life isn't the same as virtual."

That education process has a cost — and it falls on whoever is in the middle: the agency that made the promise, the studio that receives the problem, and the client who approved something that can't be delivered as planned.

What agencies can do differently before the deadline arrives

The answer isn't to avoid AI. It's to include a technical diagnostic step before any delivery commitment.

Before locking the budget and timeline for any job involving AI in image production, three concrete questions are worth answering: will the material need readable typography or text? Does the product need to be reproduced with color and texture fidelity? Does the final application size require resolution above what AI natively outputs?

If the answer is yes to any of the three, the plan needs to include real photography, specialized technical post-production, or both. Not as plan B — as part of the workflow from the start. This doesn't necessarily increase the job's total cost. In many cases it reduces it, because it eliminates the rework that costs more than doing it right from the beginning.

The brands that faced the biggest problems with AI in 2025 shared a common mistake: they let AI lead their creative decisions instead of supporting them. The same applies to production technical decisions.

The role of the specialized studio in the agency process

Kado isn't where the job goes when AI didn't work. It's where the conversation should happen before any decision about how the job will be produced.

Twenty years of post-production for fashion, beauty and advertising means being able to identify in minutes whether a material has technical viability for professional use — and what the most efficient path is to get to the result the client needs. Sometimes that's photography. Sometimes it's AI. In most cases it's the combination of both, with each part handling what it does best.

That conversation before the job starts costs nothing and can prevent the difficult conversation after the deadline has already arrived.

If you're at an agency with an image production job on the horizon — especially one involving AI — talk to Kado before closing the budget. It's a consultation, not a commitment. And it can make all the difference in the final result.