IAS Agent: AI-Powered Ad Verification Assistant

Role: Lead Product Designer · Timeline: 6 months (2025 - 2026) · Platform: Web Application

THE PROBLEM

Mid-market advertisers and independent agencies were losing hours to drowning in fragmented reporting, manual campaign setup, and struggling to enforce brand safety consistently across regions - friction that slowed activation and put compliance at risk.

THE PROBLEM

Mid-market advertisers and independent agencies were losing hours to drowning in fragmented reporting, manual campaign setup, and struggling to enforce brand safety consistently across regions - friction that slowed activation and put compliance at risk.

Metric Selector: Redesigning a Legacy Workflow for Scale

Role: Lead Product Designer · Timeline: 2 months (2026) · Platform: Signal (Report Builder)

THE PROBLEM

120+ METRICS. NO STRUCTURED WAY TO FIND, COMPARE, OR TRACK WHAT YOU'VE SELECTED. A LEGACY WORKFLOW THAT HAD NEVER RECEIVED REAL DESIGN ATTENTION.

THE APPROACH

As the sole designer on this initiative, I designed a net-new component, a call-and-response interface that surfaces selections as they're made, rather than patch the existing picker. I built the selection model with a second constraint in mind from the start: Signal was concurrently developing an agentic AI layer for report creation, and the redesigned UI needed to account for AI-configured metrics appearing alongside manual ones.

KEY DECISION - TOTAL CONTROL

When the AI layer was introduced, the core design question became: how does a user stay in control of a report that's now part-AI, part-manual? My answer was total control, no exceptions - every AI-suggested metric can be edited or removed, and manual selections can be added alongside AI ones at any point.

KEY DECISION - BUILT FOR SCALE

The component's reusability became a bigger question than metric selection alone. I raised it with the broader design team - could this same pattern extend to other multi-select contexts across the platform, like media upload in the Call section? To make that possible, I worked with Engineering to ensure the component's Call section - the input mechanism itself - could be swapped or customized independent of the review-and-control model, so the same underlying pattern could adapt to different use cases - media upload among them - without a rebuild.

That exploration led to an audit of existing platform workflows to identify where similar multi-select patterns already existed. Engineering has since built the component into Storybook, making it available for any designer on the team to pull from and adapt within defined constraints.

IMPACT

With no post-launch metrics yet available, validation came from structured sessions with 5–7 external customers. Their feedback tracked precisely onto the original research.