How Artificial Societies steered a client through a major crisis

A major consumer goods company was faced with a headline-grabbing crisis, threatening both its customer credibility and its broader licence to operate. The firm’s Board cared deeply about repairing its reputation and finding clarity about the optimal route forward to retain market share. In addition, there was a wide range of perspectives about the source, culpability and proposed consequences of the crisis, confounding the Board’s ability to decide how to communicate and move forward. It was impossible to conduct adequate research using live participants:

We supplied five specialised Artificial Societies matching the five audiences whose opinions mattered most in the situation — Policymakers and Regulators, Media, Influencers, Consumers, and Commercial Customers. To construct these societies, we built 1,174 AI personas grounded in real, anonymised profiles and conversations, sourcing data from social listening and deep research. We also mapped out the audiences’ social influence networks, so that the societies accurately model how opinions organically form in real-life settings.

From the crisis message test, Artificial Societies revealed that the best path forward for the business lay in evidence-backed messaging that directly addressed its steps to mitigate and resolve the issues in question. Our simulation also revealed nuance about how and when to deliver messages to the firm’s key audiences, uncovering a likely backlash against certain activations in the broader context of the issue, while alighting on delivery mechanisms that achieved the greatest possible reputational uplift. As a result of this message testing project, the client was able to decide on its crisis strategy and communications backed by robust evidence and recommendations. In addition, they were able to conduct follow-on simulations to test how to deliver these stress-tested messages, and where certain activations might backfire across the audience segments — all based on nuanced research-grade simulations of audiences that matter, at a scale, speed, and security that would’ve been impossible with human participants.