Minnie AI Superdecoder

Minnie is a first-of-her-kind AI led machine learning tool that helps advertisers categorise, benchmark and decode every aspect of their advertising material. Named after Minerva, Roman goddess of wisdom, Minnie is a super-decoder that analyses everything from copy, vocabulary, tone to imagery, structure and size of models.

Australian advertising hasn’t kept up with reality and the diverse nature of society. And indeed is perpetuating stereotypes of what women ‘should’ look like, what they see and read. Working with clients, and their customers, we set out to quantify and contextualise diversity issues through data.

Quickly realising the complexity and manual nature of the task, and the inherent human biases of reviewers, we built a team from various departments and backgrounds, who undertook the task of building an AI bot that would remove subjectivity and biases. 

Whilst a human being can review one ad or a set of ads and form an opinion, Minnie can review a whole body of work quickly and objectively — eliminating the unconscious bias that impacts how we perceive content. This gives brands a unique way to evaluate their body of work, over time and across channels. Revealing insights into how the language and imagery they use impacts the appeal of their brand, and reflects or stereotypes its target audience.

Minnie has been built using best-in-class and custom-made machine learning tools to provide a complete overview of the components of advertising. These components involve object identification, optical character recognition (image to text), sentiment analysis, gender/age/emotion classification, colour coding and more. All of this is stored in a database with the ability for AI to extract and visualise trends and enable analysis on different industries and eras. 

Work By