In 2016, I left Digitas to join MDC Media Partners↗ where I would start BORN AI, a creative agency built to take advantage of artificial intelligence—in every facet of our work. At the time, LLMs were not common knowledge, nor practical for our purposes, and conversational interfaces were still fairly robotic. Nevertheless, emotional computing frameworks had evolved considerably and we believed that, with the right creative concept, even the limited options available to us in 2016 could be orchestrated to produce a magical experience. Specifically, using Assembly's media tech and chatbot technology widely available at the time, it was possible to (a) understand what a human being wanted and (b) to infer the best possible solution, in real-time. So that's what we set out to do. I called this the AI bow-tie.

Our initial focus was on crafting experiences driven by AI and designed to establish 1:1 relationships between brands or IPs and individuals. Through projects featuring various IPs held by 20th Century Fox (e.g., Alien, Deadpool), and others, we proved that the zero-party data that these interactions produced could be incredibly rich, if the experience was designed correctly. For us, that meant using characters and psychology to hide some of the limitations of the existing tech. We used experiments like Xiaoice in China as inspiration and proof, in our earliest presentations, that our approach would work. In our view, she worked so well for so long because she was a teenage girl—changing the subject or being vague/aloof wasn't out-of-character, it added to the verisimilitude. At BORN, the Deadpool project benefited from a similar effect. The character is meta at his core—Deadpool can know he is a chatbot and make fun of the fact that he is a chatbot and it's not a bug, it's a feature.
We also developed techniques for detecting real-time signals about an audience's need state, as well as triggered responses. For instance, Cuppy (Dunkin's official mascot) has the right to Tweet you about coffee in the afternoon if you're actively sending off signals about how tired you are.

Concepts such as those would later become part of the offering at ATTENTION, where I would serve as CCO toward the end of my time with MDC (which had, by then, become Stagwell). There, we created segments in which multiple clips could be recombined in different permutations, with different audio tracks, based on the need state of the audience (which can be detected in various ways, depending on the signal we're seeking).
