Mondelez's new AI tool is cutting marketing content costs by 30-50%. Here's how they're doing it-and what you can copy
Mondelez is building its own generative AI tool with Publicis Groupe and Accenture to cut the price and time it takes to make ads. The goal: 30-50% lower production costs now, with more savings as the tool scales to complex video.
The company has put more than $40 million behind it. Leadership expects the system to produce short, ready-to-air TV spots by next holiday season-and potentially content fit for the 2027 Super Bowl.
What's live right now
- Social content for Chips Ahoy in the U.S. and Milka in Germany.
- An eight-second Milka clip with chocolate waves over a wafer, with backgrounds that change by audience segment.
Traditionally, animations like this run in the hundreds of thousands. This setup is "orders of magnitude" cheaper, according to the company.
Next moves
- Oreo will use the tool for Amazon and Walmart product pages in November.
- Rollouts planned for Lacta and Oreo in Brazil, and Cadbury in the UK.
Mondelez isn't using AI-generated human likenesses yet. That's a deliberate choice after seeing how AI-made people in other brand campaigns drew criticism for feeling flat or emotionless.
Why this matters for marketers
Tariffs and tighter shopper budgets are putting pressure on margins. AI is the lever to reduce agency fees, build more content variants, and shorten the gap between brief and launch.
Peers like Kraft Heinz and Coca-Cola are experimenting too. And with platforms racing to automate ad creative (Meta is targeting heavy automation by 2026), the bar for speed and efficiency is moving.
Guardrails and governance
Every output gets human review. The company's rules prohibit promoting unhealthy eating habits, vaping, overconsumption, emotionally manipulative language, and offensive stereotypes.
This isn't "set and forget." It's AI-assisted production with brand safety baked in.
Playbook: how to apply this in your org
- Start with the cheapest wins: short social cuts, cinemagraphs, and modular animations that can be versioned by audience or channel.
- Move into retail media next. Product page videos and image sets are high volume, measurable, and perfect for testing.
- Build a creative spec library: product angles, lighting, color palettes, and brand voice prompts that the tool can reuse.
- Stand up a "human-in-the-loop" review step with clear reject reasons tied to brand rules and legal guidance.
- Track a simple scorecard: cost per asset, turnaround time, approval rate, and lift from personalization versus generic creative.
- Avoid synthetic people until your team proves consistency on objects, textures, and motion. Real humans still carry the emotional load.
- Negotiate new agency models focused on concepts, brand platforms, and QA-let automation handle the asset factory work.
- Plan for TV only after you've nailed social and retail assets at scale. Treat broadcast as a downstream formatting step, not the pilot.
Benchmarks to aim for
- 30-50% reduction in production costs within the first quarter of use.
- Turnaround times measured in days, not weeks.
- Personalized variants that actually outperform control-measured by CTR, ROAS, or basket size, not vanity metrics.
Bottom line
The edge here isn't the tool-it's the system. Clear guardrails, modular creative, and ruthless measurement. Do that, and you get cheaper assets, faster cycles, and personalization that earns its keep.
If you're skilling up a marketing team for this shift, see our AI certification for marketing specialists and our rundown of AI tools for generative video.
Your membership also unlocks:
