Market research is a $120 billion industry, but the tools that predict how people will react to new ideas are still slow, expensive and often out of reach for startups or creators. London-based Artificial Societies believes it has a solution: AI simulations that mimic how groups of people interact with each other online.
It has just raised $5.35 million to bring the technology out of academia and into the hands of marketers, creatives, and businesses everywhere. The funding includes a $3.35 million seed round led by Point72 Ventures, alongside Y Combinator, and a $2 million pre-seed round with backing from angel investors tied to DeepMind, Strava, and Sequoia Scout.
James He, CEO and Co-Founder of Artificial Societies, shares, “Until now, access to experiment with human behaviour at scale has been available only to those with corporate-level market research tools. This investment accelerates our mission of bringing advanced behavioural insights to our users, enabling them to understand societies, customers and public opinion better, and make more informed decisions.”
Bridging the gap between research and real-world insight
Artificial Societies is part startup, part social science experiment. He, a Cambridge-trained behavioural scientist, once led an experiment that pitted 33,000 AI chatbots against each other to study their collective behaviour, one of the largest projects of its kind in Europe. That research eventually became a hit at computational social science conferences and even landed in the British Journal of Psychology.
But what started as academic curiosity became the seed for a company.
“It started when I was a research student at Cambridge, and with a series of questions I had: ‘How do we create a computer system that can understand how a group of people would react to a piece of content? … What if we created a group of AI personas and let them play a group of humans?” This would be a social sandbox that can tell us what narratives come up before you take the bet in reality,” He shares with TFN.
Predict, don’t guess
The company’s flagship simulator, nicknamed AS, claims to predict the performance of social media content and ad campaigns with more than 80 per cent accuracy. Instead of waiting weeks for survey responses or field-testing ads, users can run instant experiments with AI “societies” that interact and respond to content.
“Artificial Societies is essentially a whole collection of AI personas that lets you test how a group of people will react to your content, and how they will interact with each other in response to your content, before you actually put it out in the real world,” He adds.
Unlike other analytics or research platforms, Artificial Societies is free to try, with no enterprise onboarding. The founders point out that, for now, no other social simulation product is this accessible.
What’s next?
The fresh cash will be invested in building out partnerships in market research, audience insights, and content testing. In the following year, the company plans to expand simulations across more verticals.
“Over the next 6-12 months, we are focusing on developing more partnerships and applying simulation technology to a wide range of verticals, while continuing to conduct foundational research into the science of predicting large-scale human behaviour,” concludes He.
