Mark Splisbury from Tribus Digital has watched clients turn up to calls having already built a working prototype in Claude. In this episode, he explains why that's rarely as far along as it looks, and what agencies need to rethink about pricing, ownership, and training as a result.
In this episode, we cover:
- Why a clickable AI prototype can look finished while most of the real work is still hidden, using a theatre-backdrop analogy
- How a report page went from half a day of design work to ten minutes once Claude was fed a screenshot
- The friction between designers embracing AI-built prototypes and developers feeling their role encroached on
- Why Mark asks his team to accept code that's “78% as good” if AI delivers it in a fraction of the time
- The accountability problem when nobody who prompted the code feels responsible for it
- Why junior developers risk losing the ability to debug once AI stops being helpful
Guest: Mark Splisbury, Tribus Digital
Interviewer: David Wain-Heapy, Co-founder, Prodigi
Aperture: Agency AI is produced by Prodigi.