Enter the password to view this case study.
Microsoft Security · AI + Human oversight · 2024
Designing the human oversight layer for AI-driven security workflows, ensuring analysts stay in control when automated systems flag critical decisions.
The context: before agents were everywhere, this was a design-led exploration into incorporating autonomous agents into Purview workflows. Imagining a security admin delegating risk monitoring to an agent, and what it takes for that to be trustworthy rather than terrifying.
The inquiry centred on three questions:
How do we win customer trust with the right level of transparency around the agent's reasoning process?
How do we design human–AI interactions that keep the human in control of outcomes?
How can the human provide feedback to the agent for course correction?
The thesis: a human-in-the-loop approach enhances decision-making by reducing cognitive load and improving efficiency. Modernizing job-management workflows without removing human judgement from them.
My role: designed the end-to-end concept and narrative with my design manager, and presented it to Microsoft Security leadership.
Outcome: leadership aligned on the human-in-the-loop approach as the direction for agentic experiences. The transparency, control and feedback patterns explored here informed how agent experiences were framed in later work. The full walkthrough is below.