Mondelez invests $40m in generative AI to cut marketing costs by up to 50%
Mondelez is rolling out a generative AI tool to overhaul how it produces marketing assets, targeting a 30%-50% reduction in production costs. The move comes as the company faces rising tariffs and shifting consumer spend.
The tool has been developed with Publicis Groupe and Accenture. Early outputs include short TV spots built in a fraction of the usual time and cost, with first runs planned for the holiday season and potential use in Super Bowl 2027 placements.
What's new
Mondelez is concentrating on product-centric animations for brands like Chips Ahoy, Milka, and Cadbury, and is currently avoiding human likenesses in AI content. The company has already shipped social assets made with the tool-like an eight-second Milka video with dynamic chocolate visuals aimed at specific audience segments-work that would typically cost hundreds of thousands to produce.
Beyond ads, the tool will refresh retail product pages. In the U.S., Oreo PDPs on Amazon and Walmart will begin using AI-generated assets starting in November. Rollouts for Lacta in Brazil and Cadbury in the UK are next.
Why this matters for marketers
- Cost compression: 30%-50% savings on production unlocks more creative testing without ballooning budgets.
- Speed to market: Short-form video and retail assets can be produced in days, not weeks.
- Variant scale: Dozens of creative iterations for audience segments become feasible.
- Operating model shift: More in-house production and leaner agency scopes.
Peers are moving too. Kraft Heinz and Coca-Cola are experimenting with AI-generated ads. Coca-Cola's 2024 holiday work drew mixed reactions, a reminder that quality control and brand alignment still make or break performance.
Governance and guardrails
Every AI asset will be reviewed by humans before it goes live. Mondelez has guidelines in place to avoid promoting unhealthy eating habits and other sensitive topics. For now, the brand is skipping human likenesses to reduce legal and reputational risk while the tech and policies mature.
How to apply this playbook in your team
- Start where the volume is high and the format is standardized: six- to fifteen-second cutdowns, snackable social videos, banners, and PDP images/videos.
- Build a review workflow: legal, brand, and nutrition or claims checks before publishing.
- Create prompt and style libraries: tone, visual rules, disclaimers, safe topics, and "do-not-use" lists.
- Instrument measurement: time-to-market, cost per asset, creative variance produced per sprint, and performance lift vs. control.
- Pilot on retail media: prep multiple image/video variants for Amazon and Walmart; test copy, visuals, and sequencing.
- Rescope with agencies: negotiate AI-inclusive SOWs, asset ownership, and model/data rights.
- Upskill your team: train producers, brand managers, and media leads on prompt craft, QA, and brand safety.
If you're building capability in-house, consider structured upskilling that maps AI skills to marketing roles. See our AI courses by job or the AI Certification for Marketing Specialists to speed up adoption.
Rollout snapshot
- Holiday season: first AI-generated TV and social spots go live.
- November: Oreo retail pages on Amazon and Walmart get AI-built assets.
- Next: Expansion to Lacta (Brazil) and Cadbury (UK); Super Bowl 2027 under consideration.
Bottom line
AI-assisted production is shifting from experiments to core workflow. The brands that set clear guardrails, measure aggressively, and ship fast will win on cost, speed, and creative range-without sacrificing trust.
Your membership also unlocks:
