The Next AI Race Is Predicting What People Will Do

AI companies are trying to predict what people will do next, and investors are putting hundreds of millions of dollars behind the idea.

Last week, TechCrunch profiled Mirror Particle, a two-year-old San Francisco startup building what it calls a world model of human behavior. It joins a growing field. Over the past year, Simile raised $200 million at a $2 billion valuation, Aaru raised $88 million at a $1 billion headline valuation and Humans& launched Persimmon, its behavior-modeling product, after a $480 million seed round.

These companies are attracting major investment to tackle a problem marketers have been dealing with for years. Audience intelligence companies like RAD Intel already look beyond demographic profiles to understand consumer behavior. Now, startups like Mirror Particle are trying to model why that behavior changes and predict what people might do next.

Why Mirror Particle Is Moving Beyond Language Models

Many attempts to predict consumer behavior still rely on large language models. Describe an audience, have the model role-play it, and use its answers to predict how real people might respond. Mirror Particle CEO Abhivyakti Ahuja argues that this approach misses much of what drives human behavior. People respond to social pressure, culture, visual cues, and changing circumstances that rarely show up cleanly in text.

Mirror Particle is taking a different approach to behavioral AI by focusing on revealed behavior, meaning what people actually do rather than what they say in surveys. The company combines customer information with sources including current events, pop culture, and social media to better understand what shapes people's decisions.

Audiences Change Before Their Profiles Do

Ahuja told TechCrunch that Mirror Particle wants to model the “changing person.” Someone can stay the same age, live in the same place, and earn the same income while becoming a very different customer. A new job, a cultural trend, an economic worry, or a new community can reshape what they pay attention to without touching a single demographic field.

Changes in what people engage with, the communities they participate in, and the creators they follow can offer clues about shifting interests. Those patterns give marketers another way to understand audiences beyond the demographic information in a customer profile.

Audience intelligence can help marketers make sense of these changes. RAD Intel uses AI-driven technology to understand audiences through their behaviors, interests, and communities, helping marketers make more informed decisions about creator selection, creative strategy, and campaigns.

What Behavioral AI Could Mean for Marketing

Mirror Particle’s most telling example is a pet food pilot. The brand wanted to know whether chicken, beef, or vegetables on its packaging would lift sales. According to Mirror Particle, its analysis suggested the imagery wasn't the main issue. Consumers already perceived the brand as inexpensive and mass market, and changing the picture on the package was unlikely to change that perception.

Ahuja offered a hypothetical: a beauty brand trying to sell eyeshadow palettes to Gen Z when the audience might actually prefer blush.

For marketers, behavioral intelligence can help identify which decisions actually influence consumer behavior, so less time goes into optimizing options that may have little impact. The same applies to creator marketing. Understanding what an audience cares about can help brands choose creators and develop creative that resonates with them.

Prediction Still Needs Proof

Predicting people is a far bigger claim than predicting language. Mirror Particle is still early, and its approach has to prove itself across more brands and categories. The same is true for the whole field. To prove these models work, companies will need to compare their predictions with what consumers actually do.

Still, the direction is worth watching. Investors are betting that the next useful AI model needs to understand how people change, not just reproduce how they talk. For marketers, audience intelligence could become an even more useful way to understand how consumer behavior changes and what might influence people's decisions next.