Case study
Responsible AI and Governance
Translating rules and principles into understandable human decisions and product behaviour.
The situation
AI use cases in a federal security company have to satisfy governance, data protection, information security and the EU AI Act, and still be usable by the people doing the work. Rules alone don't tell a team what to build or how a system should behave. One focus was helping staff decide, at the moment it matters, whether and how they may use AI for a given task or piece of information.
Audit
What I found
I analysed and structured AI use cases against user needs, organisational requirements, governance and responsible implementation, working with business units, product teams, technology and information security.
Pathway
The pathway we designed
I translated regulatory, technical and organisational requirements into concepts, decision documents and implementation measures that teams could act on. Instead of asking people to remember policy, I turned the rules into short decision aids for the moments where they apply: before entering information into an AI tool, when judging how sensitive data is, when assessing a new AI use case against regulation, and whenever someone is unsure. The aids are designed for all staff, with particular attention to non-technical and administrative roles. The wider lens I use before a use case is built is in How I evaluate AI use cases.
Prototype
What we made tangible
A set of practical decision aids: an input check before entering information into an AI tool, a data sensitivity check, a decision aid for assessing use cases against regulation, and a simple rule for uncertain moments: stop, check, ask. They are built into scenario-based learning in Articulate Storyline and Rise 360, described in AI Learning Pathways.
Test
What testing showed
Colleagues from different parts of the organisation reviewed the aids in several feedback rounds.
My role
Senior UX Specialist for AI strategy, adoption and digital trust. I analysed AI use cases, translated governance requirements into concepts and decision documents, and designed the practical decision aids as part of a cross-functional team.