BY: Paul Aaron, Co-Founder and CEO of Addition
Brands have always operated with an inherent delay.
Strategic deliberations, creative processes, and production realities create a chasm between what people are experiencing in the world and what a brand has to say.
But as technology has progressed, this gap has been closing. The internet established a foundation for realtime communication. Social media enabled two-way conversation. And now, with generative AI, brands can enter an era of true responsiveness.
R/GA company Addition’s recent work for Google at last month’s MLB All Star Game offers a glimpse at how the AI-powered shift towards realtime brands is underway.
In partnership with Google Cloud and Major League Baseball, we created an AI system that generates contextual messages that adapt to where and when they appear—understanding player lineups, in-game stats, and even fans’ geographic locations—to create unique content that engages fans in that precise moment.
Participating in Reality in Realtime
The Google Cloud team came to us with a challenge: How do we use Google’s frontier AI technologies and our partnership with Major League Baseball to intelligently engage fans in real time?
The ambition went beyond having a library of preset of messages—a series of if-then statements, MadLibs and decision trees to respond to gameplay. Instead, we wanted to show what was only possible now with Google’s latest AI technology: a system that could understand, think, and participate in realtime.
The system we came up with leverages a combination of classical AI and machine learning, data science, and Agentic AI to react to the world in a way that is not just human, but in many ways, beyond human in terms of what it can do. It’s a synthesis of the nuanced creativity of a copywriter and the analytical rigor of an MLB-obsessed data scientist.
How it Works:
Data Foundation: We leveraged MLB’s Statcast API to process years of historical data from the entire All-Star roster. This gives us a foundation of conventional stats like batting averages, home run percentages, hit direction data, exit velocity, and an understanding of how specific stadiums and weather conditions impact player performance.
Classical AI: We leverage classical AI to generate a predictive model of player performance, for example a player’s likelihood to hit a home run to a given section of the park based on a set of conditions. These pre-computed predictions get added into the data foundation to give the system real-time and predictive capabilities.
Real-Time Context: We tap back into the MLB API to get live data streams including current batting order, weather conditions, wind direction, and temperature. The system continuously updates its predictions as conditions change and the lineup evolves.
Agentic AI: Sitting on top of this data foundation of historical, pre-computed AI analysis, and realtime context is a Gemini Agent. The agent’s job is twofold – it not only analyzes real-time data to come up with engaging insights, but also comes up with creative messages to deliver those insights to fans in and around the stadium.
Humans in the Loop: To ensure brand safety and accuracy, a team supervised the AI system from a command center to review, select, and if necessary, override content as it’s being displayed.
Around the Stadium: In addition to generating messages in the stadium, we created a mobile OOH activation that generates dynamic insights and creative messages based on where it travels around All Star Village.
The result is fan engagement that feels magical because it’s predictive and personal. Before Shohei Ohtani even steps to the plate, fans in sections 152-154 might see a message like, “Heads up! You’re in Ohtani’s sweet spot based on 2 years of launch data.” Meanwhile, a mobile billboard outside the stadium might compare a player’s power to local traffic, noting, “A ball off Ronald Acuña Jr.’s bat travels 121 MPH. That’s more than 5x our average speed on Cobb Parkway.” This is the essence of realtime brand engagement: it’s not just displaying stats, but collapsing the distance between data and lived experience to create moments of anticipation and connection.
From Creative Campaigns to Creative Systems
This project wasn’t just about creating clever billboards; it’s a blueprint for a fundamental shift in how brands operate.
Instead of going through a traditional creative and production process, those efforts are instead spent creating a system that can do both on the fly.
There’s still a role for human creativity. But that energy is channeled into architecting the system itself: defining its goals, setting its creative boundaries, and curating the high-quality, human-crafted examples from which it learns its voice and style.
This new discipline—one that’s equal parts communication arts and computer science—has the potential to transform how brands show up in the world by enabling them to understand and participate in a reality that’s constantly changing.
This same approach can apply across a myriad of other contexts:
- Retail promotions based on weather, inventory, and foot traffic.
- A social presence that can participate in cultural moments as they happen.
- Localized marketing that actually reflects a real-time understanding of local context and culture.
The constraint is no longer the technology, but the organization itself. Can creative leaders shift from making stories to making systems? Can different departments—creative, data, legal—work together to unlock AI’s realtime potential? Can brands establish governance that balances speed with safety?
Building Your Real-Time Capability
For brands looking to develop real-time capabilities, our work with Google provides a practical playbook:
1. Start with contained, high-value use cases where speed is a clear competitive advantage. Think sports, trending culture, or seasonal messaging—anywhere a delay creates obvious friction between your brand and your audience.
2. Map the data streams that could inform real-time responses. What signals indicate a moment worth responding to? What information would make that response relevant?
3. Design systems with clear guardrails. Define what’s on-brand and what’s not. Build in safety checks. Create review processes that can operate in seconds, not days.
4. Test, measure, and expand based on what works. These systems get better with iteration, both in their outputs and in your understanding of how to design them.
Taken from: R/GA