SQREEM's New CEO Bets on Behavioral AI Over Language Models

Marketing technology company SQREEM is taking a deliberate step away from the large language model (LLM) race, betting instead on what it calls a large behavioral model, or LBM, to give brands a more precise view of consumer intent. The move is being driven by new CEO Stephen Yap, who joined the company after stints at Perion and nearly two decades at Google.

Yap argues that the proliferation of AI startups has made the market feel overcrowded, with vendors pitching machine learning and language models as a universal solution to marketing challenges. SQREEM's alternative relies on state-space modeling to track how behaviors shift over time, drawing on real-time data from social platforms and open web sources rather than analyzing written or spoken language.

Founder René Raiss describes the LBM as capable of mapping behavioral chains — for example, tracing a user's path from searching for green tea to looking up antioxidants, then skin care, sunscreen, and eventually vacation planning — to build audience personas based on actions rather than demographics. Raiss acknowledges, however, that the model cannot predict pure chance or once-in-a-lifetime events.

The company points to a real-world case involving a Cyclospora outbreak — referred to by Yap as "the lettuce issue" — during which a grocery client used the LBM to identify which consumer groups were most concerned and what their specific worries were. According to the source, the model identified three main groups: elderly shoppers, parents with newborns, and people with health problems, whose primary concern was whether the parasite could spread beyond lettuce to other produce. That finding, the source reports, prompted the grocer to adjust its messaging.

SQREEM says its platform never links behavioral data to individuals, tracking instead public mood and aggregate behavior shifts. Brands can integrate the tools directly into their existing dashboards.

The broader AI sector context is relevant: Reuters has reported that major AI companies face growing pressure to demonstrate the value, safety, and differentiation of their models, citing as one example a $2 billion partnership between Anthropic and Accenture to evaluate frontier AI models over five years. Reuters has also noted that leading AI firms, including OpenAI, are incorporating behavioral signals alongside self-reported data — a spokesperson for OpenAI told Reuters that the company uses both account and behavioral signals in addition to self-reported age.

Whether SQREEM's behavioral approach can deliver on its claims at scale remains to be seen, but Yap's position is clear: knowing what consumers are doing in real time is a more actionable advantage than predicting what they might say.