The campaign is stuck in the design queue. How big brands scale assets with AI
The campaign idea usually comes quickly. The rest is another story. Square, vertical, stories, a banner in five sizes, a German and a Czech version. Unilever and Mondelēz show that this part can move into a process where AI handles production and people look after quality.
Where a campaign really loses time
If you work in marketing, you know the moment. The key visual is approved, everyone is happy, and then production starts. A square post and a vertical one, stories, banners in several sizes, a website graphic, a newsletter header. A multi-market campaign adds language versions, and every copy change means editing every single file.
None of this needs a new idea. All of it needs a designer’s time, so the campaign waits in a queue while the creative team spends a week nudging elements around artboards. Big brands have exactly the same problem, only at a scale where manual work stops making sense.
Unilever and one file per product
In March 2025, Unilever explained in a press release how it changed its product shoots. Instead of photographing every variant separately, the company builds 3D digital twins of its products (using, among other things, the NVIDIA Omniverse platform). According to Unilever, product imagery is now created twice as fast and at half the cost.
More interesting than the numbers is a detail that’s easy to miss. Each digital twin holds all of a product’s variants, labels, packaging and language formats in a single file. Unilever says this cut repeated work in the process from an average of five times to once. The team no longer prepares separate images for TV and for e-commerce. It prepares one source, and everything else comes from it.
“This isn't about pushing out more content – anyone can do that.”
We couldn’t agree more. Volume on its own fixes nothing if the materials start drifting apart. Unilever also claims full brand consistency across its product imagery, and for us that is the most important part of the story.
Mondelēz leaves the final call to people
The maker of Oreo and Milka is taking a similar route. As Reuters reported in October 2025, Mondelēz built a generative AI tool for marketing content together with Publicis Groupe and Accenture. The company has invested more than 40 million dollars in it and expects content production costs to fall by 30 to 50 percent. The tool already produces social media content, for example for Milka in Germany, where a short chocolate animation gets a different background depending on who it’s shown to.
The same report contains a line that matters more than the budget. A company executive says people always check what the tool produces, and Mondelēz has written rules on what its content must never do. AI took over producing variants. The team still decides what reaches customers.
What marketing leaders say
These two companies aren’t outliers. Gartner asked 402 CMOs and marketing decision-makers in North America, the UK and Europe about their 2025 budgets. As Marketing Dive summed up, budgets stayed at 7.7% of company revenue, the same as a year earlier. With a flat budget every hour counts, so it’s no surprise that 49% of respondents see time savings as the main benefit of generative AI, 40% lower costs, and 27% the ability to produce more content or take on other areas.
How to bring this to a smaller company
You don’t need digital twins or 40 million dollars. Both stories share a principle that works in a company with one designer, or with none.
It all starts with a single source. The main campaign design and the brand rules need to be written down so that tools can use them, not only people: colours, fonts, layouts, photo cropping, tone of copy. Add format templates that state clearly what can change and what stays fixed. Only on that foundation can AI sensibly produce new sizes and language versions. At the end there is always a person who approves the result and fixes whatever needs a human eye.
In our AI design workflows service we simply call this area more output from the same team. Marketing gets the full set of files sooner, and designers get back to work that needs ideas.
Where to start
Pick one campaign that comes back regularly: a monthly promotion, an event series, a seasonal sale. Count the days between key visual approval and the moment marketing has every format and language version. That’s your baseline. If automation doesn’t shorten it, something in the process is set up wrong, and it’s better to find out after one campaign than after ten.
Why it pays to measure the whole process rather than one department’s speed is the subject of our post on the bottleneck that moves after AI adoption.
AI design workflows
Let’s count how much time your adaptations take
On a free call we’ll walk through one of your campaigns and show which formats and language versions AI can prepare.
FAQ
Can AI adapt a campaign to different formats?
Yes, it’s one of the tasks AI handles best. There is one condition: AI needs clear rules, meaning the main design, format templates and a description of the brand. A designer or the brand owner approves the final files.
Will automating assets lower campaign quality?
It doesn’t have to. Mondelēz, which built its own AI content tool, says people check every result. What gets automated is repetitive production; the idea and the decisions stay with the team.
Where do I start with marketing asset automation?
With one campaign that comes back regularly. Measure the time from an approved key visual to a full set of formats and language versions, automate the most time-consuming adaptations and compare.
Sources
- Unilever, “Unilever reinvents product shoots with digital twins and AI”, press release, 18 March 2025.
- Reuters, “Oreo-maker Mondelēz To Use New Generative AI Tool To Slash Marketing Costs”, republished by ESM Magazine, 27 October 2025.
- “CMOs stretch static budgets further, with downsides for agencies: report”, Marketing Dive, 12 May 2025 (on the Gartner CMO Spend Survey 2025).
CAASUAL is not affiliated with Unilever, Mondelēz or the authors of the cited publications. Figures are quoted from the sources; savings are company claims. We quote one sentence verbatim. Company and brand names belong to their respective owners.