Yext
2026
Turning Data Overload into AI-Driven Actions
Designed the Action Center, a single hub that turns billions of search signals into a prioritized, actionable checklist.
The Context
A powerful tool that users did not know how to use.
Scout is Yext’s AI search-visibility product. It shows multi-location businesses how they appear in search by ingesting billions of signals across locations, keywords, and competitors. As the product matured, a pattern emerged in customer feedback: users trusted the data, but didn’t know how to act on it, and some were leaving because of it.
Yext needed Scout to evolve from a reporting tool into a product that drove measurable outcomes, and to do it quickly in a competitive AI search market.
The Problem
Data wasn't the gap—direction was.
Three personas logged into Scout, and each needed a different answer to the same question:what do I do first?
Marketing Manager
Needs high-level keyword trends and competitive landscape insights to adjust overall strategy.
Local Manager
Focuses on granular location-based signals to improve specific branch visibility and engagement.
Executive
Requires summarized performance scores and ROI metrics for quarterly reviews and budgeting.
The Process
Defining what a good action looks like.
Working with engineering, we turned Scout’s core performance metrics into structured playbooks: rules telling an AI agent which signal matters, what fixes it, and how to explain why. My part was defining what a useful action looks like to a user (a clear task, the reason behind it, its expected impact, and a one-click fix) and reviewing agent output against that standard. We refined the prompts over several rounds until the actions held up across real customer accounts.
The Workflow Decision
Adapting how we designed.
Figma used for: Blueprint and alignment.
The initial 0-to-1 vision, design system logic, and early stakeholder alignment on layout. Figma remained essential for establishing the frame, but it ceased to be the primary tool for iteration.
Claude used for: Prototyping in code.
Because Scout’s AI capabilities changed weekly, static mockups couldn’t keep up. I used Claude to build working prototypes in code, connected to real data states. That let me test ideas against real edge cases and run strategy conversations with working software instead of static screens.
What this unlocked:
Prototyping in code removed the translation gap between design intent and engineering reality. Edge cases, latency states, and dynamic AI outputs that were once faked in Figma became real constraints to design around early, reducing revision cycles after handoff.
The Iterations
From two scores to one clear next step.
The first version of the home page asked users to interpret two scores and then decide what to do. Over three rounds, we moved the page from explaining performance to telling users what to do next, merging the scores, simplifying how users narrowed down their options, and making each action explain itself.
Two scores, three strategies
We placed an Optimization score beside the Visibility score and asked users to pick a strategy (Improve Visibility, Improve Optimization, or Target a competitor) before seeing recommendations as a card grid. What we learned: The two scores competed for attention, and users weren’t sure which one to improve. The grid made every action look equally important.
Optimization becomes progress
We reframed optimization as progress, showing actions completed instead of a grade. “Target a competitor” became a full action plan built against one rival. What we learned: Progress felt better than a grade, but users still had to choose a strategy before they saw anything useful.
Actions get their own home
We removed the strategy step and moved every action into one dedicated panel, with a score at the top showing how many were left to complete. What we learned: One list was faster than choosing a strategy, but a remaining count felt like a backlog. It framed the work as debt instead of opportunity.
What changed across rounds
Two scores → one score. Choosing a strategy → one list. Counting what’s left → showing the next step. Equal cards → a prioritized list with reasons.
The Solution & Impact
Introducing Agentic Actions
The final home page leads with a single combined score. Showing a prioritized list of actions, each including what to fix, why it matters, and its expected impact. Most can be completed in one click by an AI agent; we called these Agentic Actions. The upfront strategy choice was removed entirely, so users see their most important action as soon as they log in.
The Impact
The Agentic Actions turned Scout from a reporting tool into a workflow tool that tells users what to do next and does it for them. Working prototypes on real data let leadership see that shift early, and it became the direction for Scout’s roadmap. Users who used to log in, look, and leave started coming back to finish work, and customers started paying for it.
$6.8M+
ACV Secured
48%
Session Growth
