Making Analytics Accessible Through AI Summaries
Hireroad I 2024-2025
OVERVIEW
People Insight by HireRoad delivers people analytics through Tableau dashboards embedded directly within the product. While the dashboards provided powerful data, users often had to navigate multiple tabs within a workbook to understand performance, trends, and outcomes.
As usage grew, it became clear that users needed help interpreting the data—not accessing more of it. We introduced an AI insight overlay to summarize dashboard data and surface key takeaways, helping users understand what mattered at a glance.
Learn more about what People Insight does.
PROBLEM
-
Users were overwhelmed by dense, multi-dashboard workbooks, making it difficult to quickly identify insights.
-
Embedded Tableau dashboards limited opportunities for guided interpretation or contextual explanations.
-
Non-analytical users struggled to translate charts into clear conclusions.
-
Summaries and reports were created manually, slowing down decision-making and increasing effort.
-
Important trends and anomalies were easy to miss without deep exploration.
RESEARCH
-
Users were overwhelmed by multiple, data-heavy dashboards, especially non-power users who struggled to quickly scan and interpret insights.
-
Even small changes or clarifications required CX team involvement, creating friction and slowing iteration.
-
Report generation was manual and CX-dependent, limiting self-serve access to insights.
-
Dashboards followed a one-size-fits-all model, despite different user roles needing different levels of detail.
-
Embedded Tableau constraints limited customization, guidance, and modern interaction patterns.
-
Tableau’s analytical depth came at the cost of usability and approachability.
-
Metrics changed frequently, but users relied on static views and exports, increasing the risk of outdated insights.
APPROACH
Users relied on Tableau dashboards embedded in an iframe but struggled to interpret the data—most ended up submitting requests to the CX team for help understanding what they were seeing. Since we couldn't modify the Tableau interface itself, I designed an AI summary side panel that translates complex visualizations into plain language insights.
Users relied on Tableau dashboards embedded in an iframe but struggled to interpret the data—most ended up submitting requests to the CX team for help understanding what they were seeing. Since we couldn't modify the Tableau interface itself, I designed an AI summary side panel that translates complex visualizations into plain language insights.
Users relied on Tableau dashboards embedded in an iframe but struggled to interpret the data—most ended up submitting requests to the CX team for help understanding what they were seeing. Since we couldn't modify the Tableau interface itself, I designed an AI summary side panel that translates complex visualizations into plain language insights.
The side panel sits alongside the dashboard so users can read summaries while viewing the data. When users apply filters in Tableau, a refresh indicator lights up to let them know the summary needs updating—keeping the AI insights synchronized with what they're actually looking at.
Given the data-heavy nature of the reports, we categorized summaries into three types: descriptive (what's happening), diagnostic (what's wrong), and predictive (what will happen next).
SOLUTION
Organizing insights by intent
I structured the panel with three tabs—Descriptive, Diagnostic, and Predictive—so users could jump directly to the type of insight they needed. Descriptive answers "what's happening," Diagnostic explains "why," and Predictive shows "what's likely next." This categorization matched how users naturally think about their data questions.
Keeping summaries synchronized
When users apply filters in Tableau, a refresh indicator appears on the panel, signaling that the summary reflects outdated data. Users click refresh to regenerate insights based on their current view. This keeps the AI layer accurate without automatically refreshing on every change, which would be disruptive.
Preserving the source of truth
The panel generates summaries from Tableau's data but doesn't modify or replace the dashboard itself. Users can reference the exact charts and numbers while reading AI-generated insights, making it easy to verify claims and dig deeper when needed.
Takeaways
The AI summary panel helped users interpret complex dashboards without relying on the CX team for basic explanations. Non-power users who previously avoided Tableau could now understand their data by reading plain-language summaries alongside the visualizations.
Diagnostic and predictive insights were particularly well-received—users valued understanding not just what was happening, but why trends occurred and what might happen next. This validated our phased approach and confirmed demand for more interpretive AI features beyond simple description.
The design proved that AI summaries could enhance existing tools without requiring changes to dashboards or underlying infrastructure, establishing a pattern for future AI integrations across the platform.