Pauline Kabitsis

Spending on AI products is enormous while a large majority of AI initiatives are failing to deliver ROI, and the common pattern is that users bounce after a first frustrating encounter. The bottleneck isn’t model capability anymore — it’s design and adoption. What makes the talk unique is that it’s not opinion or vibes: it’s a structured taxonomy backed by published research and live product teardowns the audience uses every day, delivered by someone who runs these diagnoses on real products.

This talk is built on real product teardowns mapped to each of the seven sins: ChatGPT’s guidance-free blank canvas (Neglect), Amazon’s near-invisible AI icon (Obscurity), Google AI Studio’s option overload (Gluttony), Zoom’s AI Companion failing a mid-meeting recap (Mediocrity), Samsung’s “Bespoke AI” ad that sells the label instead of the benefit (Vanity), usage-limit lockouts that strip away user control (Tyranny), and the sea-of-sameness chat interfaces (Envy).