Case Studies
Product design for complex, technical software — AI features shipped into enterprise products, and the strategy work that decides what gets built.
- 01 · Product design · AI behaviour
Multilingual Detection & Filtering
I shipped language filtering for an AI search product, then found the model classifying inconsistently across document types. I mapped the four patterns behind it, defined behaviour rules with engineering, and made confidence visible in the UI.
- Edge case mapping
- Confidence in the UI
- Shipped
- 02 · AI transparency · Retrieval UX
Sources: trusted context
The panel showed every document the system had scanned, so the ones that actually shaped the answer were buried. I split sources into three roles, turning verification from opening every document into reading it off the panel.
- Source hierarchy
- Verification UX
- In development
- 03 · Permissions & governance · Designer + PM
Document Metadata Enrichment
Flat permissions let anyone create a label, so duplicate metadata multiplied on its own. Nobody owned the vocabulary. I split metadata into three governed layers and removed the permission that caused it, rather than redesigning the picker.
- Three governed layers
- Default field set
- Designer + PM
- 04 · AI workflows · In development
Workspace Skills
Ready-made actions — summarise, translate, compare — plus skills a client configures once, like filling in a questionnaire from their own documents. They run while the conversation keeps going.
- Built-in & client skills
- Never blocks the chat
- In development
