Sony Innovation Fund Backs Primitive Labs, a Startup That Builds Simulated Crowds to Stress-Test Products Before Launch

AI has made building software dramatically cheaper, and Primitive Labs exists because of the gap that leaves behind: understanding whether anyone will actually use it. The San Francisco startup announced on October 5, via PR Newswire, an investment from Sony Innovation Fund to continue building Instinct, an enterprise platform that models customers and simulates how they interact with digital products before launch. The investment amount was not disclosed.

The founders' argument starts with a distinction most analytics tools blur. Rohit Talluri, the CEO who previously worked in Amazon's AGI Autonomy Lab on computer-use agent initiatives, says AI systems today can predict the next action but cannot reason about human behavior — how people form goals, change their beliefs, and decide to persist or quit under uncertainty and time pressure. Instinct takes a company's user research, product analytics and behavioral data, and constructs simulated customer populations that navigate real product funnels, so teams can see likely outcomes before committing to a launch.

The company says its research has generated hundreds of thousands of simulated behavioral trajectories across hundreds of customer populations and thousands of digital environments — figures that are company-reported and not independently audited. The stated use cases concentrate on high-value journeys: onboarding, account creation, subscription and checkout, including how those experiences need to change for different customer segments or new markets. Retail, e-commerce, media, advertising and gaming are the current engagement areas, with healthcare, financial services, travel and marketplaces flagged as longer-term.

The founding team is entirely ex-AWS: Talluri (CEO), Jean Farmer (CTO), who worked in Amazon AGI and supported the Nova model launches, and Gabriel Fong (COO). The broader team draws from Amazon, Google, Harvard, MIT and the University of Washington. Before the Sony round, the company was part of a16z speedrun's SR007 cohort — the accelerator's partner Josh Lu framed the bet as building "a new class of behavioral models for enterprises" — alongside Olive Tree Capital and Pack Ventures.

Sony Ventures' US managing director Atsushi Kitagawa tied the investment to the agent era: as enterprises become more agentic, understanding the customer journey is one of the core areas his team evaluates. The logic is straightforward — if autonomous agents will soon be doing the building, the differentiating asset is a trainable model of the people on the other side of the screen.

The pitch lands in an industry where launch decisions are still mostly gut plus dashboards. If simulated populations can reliably predict where real users abandon a funnel, pre-launch testing stops being a survey and becomes a simulation discipline — with the usual caveat that a model of customer behavior is only as good as the data and assumptions underneath it. Founded this year, Primitive Labs is among the first wave of startups betting that "understanding people" becomes a capability you train rather than a team you hire.