
Sign up to save your podcasts
Or


What happens when AI stops being a tool you type into and starts becoming something you talk to? In this episode, Galen Low sits down with Oliver Shoulson, Agent Design and Engineering Lead at PolyAI, to unpack the surprisingly human problem at the heart of conversational AI: most AI conversations still feel weird.
From clunky chatbot scripts to overly polite “LLM voice” personas that sound like they were trained by a committee of HR robots, Oliver explains why good conversational design is less about mimicking humans perfectly and more about removing friction. The conversation explores the psychology of trust, the mechanics of social presence, and why the future of AI interfaces may depend less on visual design and more on understanding how people naturally speak, interrupt, hesitate, clarify, and collaborate.
Resources from this episode:
By Galen Low4.7
2020 ratings
What happens when AI stops being a tool you type into and starts becoming something you talk to? In this episode, Galen Low sits down with Oliver Shoulson, Agent Design and Engineering Lead at PolyAI, to unpack the surprisingly human problem at the heart of conversational AI: most AI conversations still feel weird.
From clunky chatbot scripts to overly polite “LLM voice” personas that sound like they were trained by a committee of HR robots, Oliver explains why good conversational design is less about mimicking humans perfectly and more about removing friction. The conversation explores the psychology of trust, the mechanics of social presence, and why the future of AI interfaces may depend less on visual design and more on understanding how people naturally speak, interrupt, hesitate, clarify, and collaborate.
Resources from this episode:

87,538 Listeners

112,161 Listeners

56,561 Listeners

284 Listeners

8,052 Listeners

233 Listeners

5,539 Listeners

19,692 Listeners

172 Listeners

12,343 Listeners