Making AI-Powered Search Usable for Non-Technical Users

Overview

This was an early-career project completed between 2015 and 2019. The focus was designing a Boolean search interface within a three-week timeline. This case study highlights my process and design decisions; detailed research data is no longer available.

Veritone's AI-powered media platform relied heavily on a search interface that had grown unwieldy. The existing search bar didn't support Boolean logic across all cognitive categories, offered no way to edit individual query terms, and cluttered limited space with icons, forcing users to reconstruct their entire search string whenever a change was needed.

The goal was to redesign the search experience to make complex queries fast, intuitive, and learnable, while maintaining the power media professionals relied on.

The Problem

Users needed to build multi-criteria Boolean queries combining AND, OR, and NOT logic across Veritone's cognitive categories such as transcription, face recognition, and object detection. The existing search interface made this nearly impossible:

• Users could not edit individual search terms, forcing them to retype the entire query for any change
• Boolean logic was only partially supported and limited to transcription, not across all engines
• Cognitive category icons cluttered the search bar and reduced usable input space
• The interface was overwhelming for average users and offered no guidance for constructing complex queries
• There was no way to save or reload search profiles for frequently repeated queries

Understand and Define

To begin the design process, I conducted interviews with stakeholders as well as internal and external users to better understand the problems the search UX/UI needed to solve. Questions focused on user goals, how they used the search function, pain points, needs, and workflow patterns.

This research helped align goals, prioritize features, and clarify workflow requirements. Based on these insights, a primary persona was created to guide design decisions.

Primary Persona: Victor Abbott

Victor represents the Customer Success persona, an expert-level Veritone user who constructs sophisticated, multi-variant search queries to manage client accounts and demonstrate platform value. His goals and frustrations directly informed the redesign priorities.

User Journey

Based on the research findings, a User Journey and Emotion Map was created to trace the search experience from arriving ready to search, to constructing a query, reviewing results, and attempting to refine it. The map clearly showed how quickly user sentiment dropped at the search query step and pinpointed where the interface caused the most frustration.

Wireframes and User Testing

Low-fidelity wireframes were created to map out the new search structure, interactions, and layouts. This helped communicate the UX/UI vision to developers, assess technical feasibility, and gather early feedback from stakeholders. Wireframes went through several rounds of iteration to refine the optimal user workflow.

Internal and external user testing validated the proposed search improvements. The three key usability enhancements below were tested, refined, and confirmed before moving into high-fidelity design.

Usability Improvement 1: Search Bar Redesign

The search bar interface was redesigned to create more usable space. Cognitive category icons were moved inside their respective input containers instead of the main search bar, freeing horizontal space and reducing visual clutter.

Usability Improvement 2: Visual Boolean Logic

Users can now build advanced queries using AND and OR operators across all of Veritone's cognitive categories, not just transcription. This enabled more complex search strings and reduced the time spent manually filtering results.

Usability Improvement 3: Search Pill Grouping

Search terms are displayed as individually editable pills that can be grouped to create complex, prioritized queries. Grouping pills allows users to set query priority and edit individual terms without reconstructing the entire search.

Solution

Three key changes solved the core problems. Cognitive category icons were moved out of the search bar and into their own input containers, freeing up query space. AND and OR Boolean operators were extended across all cognitive categories, allowing users to build searches the platform was already capable of. Search terms became individually editable pills that could be grouped to set query priority, letting users edit a single term without starting over. The result was a search experience that aligned with users' understanding of Boolean logic.

Sketch was used to design the new search UI, and InVision was used for pixel-perfect handoff to developers. Additional UX/UI interaction documentation highlighted specific behaviors, while Google's Material Design guidelines ensured a modern, consistent UI in line with Veritone's product direction.