Unilever
Unilever builds a Beauty AI Studio to make marketing assets up to 30% faster
Unilever's Beauty & Wellbeing division runs an in-house generative AI system that produces marketing content for brands such as Dove, Vaseline, POND'S and CLEAR, from first idea to personalised versions. Unilever reports assets up to 30% faster and more than doubled video completion and click-through rates.
Key figures
Copied exactly from the sources. Copy any figure with its citation.
up to 30% faster than before
Speed of making marketing assets with the Beauty AI Studio compared with before1
more than doubled
Change in Video Completion Rate and Click-Through Rate, key performance metrics for content made with the system1
What it took
The problem
Unilever's beauty and wellbeing brands need a steady flow of assets for paid social media, programmatic display advertising and e-commerce. A Unilever executive told Digiday the old way was to send briefs off and get content back, at about 20 assets per campaign. The company wanted a faster, more iterative way of working.
What they built
The Beauty AI Studio is an in-house system inside the Beauty & Wellbeing division. Unilever says it covers the whole content cycle, from first idea to personalised versions, for priority markets, for brands including Dove, Vaseline, POND'S and CLEAR. Digiday reports it runs on Pencil Pro, a generative AI tool from marketing services firm The Brandtech Group, which uses several large language models (software that generates text and images from instructions). Each brand in each market has a brand tool called Brand DNAi. It holds approved brand voices, values and rules so the AI draws only on approved material.
How they did it
Marketers write instructions and add what they know about the target audience. They generate images and video from 3D renders of the real products. Digiday says Unilever started working with Brandtech in 2023, tested Pencil in the first quarter of 2024 and spent much of the year before July 2025 designing and running the studio with an in-house specialist unit. It also reports 25,000 staff trained in AI fluency. Unilever says the studio lets teams improve assets in real time, guided by the best performing creative versions.
Results
Unilever reports that the studio produces assets up to 30% faster than before. It also says the system more than doubled video completion rate (how many viewers watch a video to the end) and click-through rate (how many viewers click). Digiday adds that a Unilever executive estimated about 400 assets per product on average, compared with about 20 per campaign before. The Unilever executive gave Digiday no estimate of cash savings.
What to copy
Treat the tool as a small part of the change. Unilever paired it with brand rules, training for thousands of staff and a performance feedback loop, and it measured business metrics rather than output alone.
Lessons for your project
Budget for people as well as software. Digiday reports Unilever trained 25,000 staff in AI fluency and spent much of a year designing and running the studio, not only buying a tool.
Put brand rules inside the system. Unilever gives each brand and market an AI tool that holds its guidelines and regulations, so output stays on brand.
Measure results by channel, not only by speed. Teams can double down on the best-performing creatives and improve assets in real time.
Use the extra volume deliberately. Digiday reports one Unilever manager estimating hundreds of assets per product against about 20 per campaign before, so decide how many variants your teams can actually test.
Caveats
- The figures are Unilever's own. Digiday repeats them from a Unilever executive, and no independent measurement is published.
- The baseline for 'up to 30% faster', the sample of assets and the period measured are not given. Digiday says the Unilever executive gave no estimate of cash savings.
- Unilever's own article calls the system 'AI-powered' and does not name the technology. Digiday describes it as generative AI built on Brandtech's Pencil Pro, which draws on several large language models it does not name, so the model is shown as undisclosed.
- Neither source gives a go-live date. Digiday (July 2025) says the studio is in place across 18 markets; the deployed date shows only the year.
- The same Unilever article covers other AI projects (Sketch Pro, Closeup SuperShoots, Recipe Intelligence). Only the Beauty AI Studio figures are used here.
- Digiday says the system is not used to create images of people and is built around product renders.
Sources
- How AI is helping drive Desire at Scale across Unilever
Unilever · 2025-09-30 · Company blog · unilever.com
- Unilever is building a gen AI assembly line for its digital creative
Digiday · 2025-07-30 · News · digiday.com
Summarized in our own words from the sources above. Spotted an error? See corrections.