Session

Unified SRE Observability: How Workday Democratizes Multi-Platform Insights with Databricks Genie

Overview

Experience	In Person
Track	Analytics & BI
Industry	Enterprise Technology
Technologies	Genie
Skill Level	Intermediate
DOWNLOAD SESSION SLIDES

As Workday’s infrastructure scaled across multi-cloud environments, the Site Reliability Engineering (SRE) teams faced a common hurdle: fragmented data silos. Monitoring reliability across Databricks, Snowflake, ServiceNow and FiveTran required pivoting between tools, leading to delayed insights and manual reporting toil. To solve this, Workday built a unified cross-platform KPI dashboard on Databricks, centralizing telemetry into a single pane of glass for platform adoption, job performance and ITSM health. This session explores the architecture behind this “one-stop shop” — leveraging third-party integrations for seamless ingestion and the power of Databricks Genie. By embedding conversational AI into the SRE workflow, Workday enables stakeholders to “talk to their data” — getting instant, natural-language answers without SQL. Learn how this shift from static charts to AI-driven insights has accelerated incident response, improved governance and freed SREs to focus on engineering.

Session Speakers

Aniruddha Vishnupurikar

/Director, AI Platforms and SRE
Workday

Manan Bhavsar

/DataSRE Manager
Workday

Full Summary

How Workday unified SRE observability across 30+ platforms with Genie and AI/BI Dashboards

Workday's SRE leaders describe how they replaced fragmented, tool-by-tool reporting with a governed, AI-enabled observability layer.

The team consolidated data and metrics from more than 30 data and AI platforms into a single environment on the Databricks Platform, then exposed shared KPI definitions through AI/BI Dashboards and Genie for conversational analytics.

The approach cut manual back-and-forth, aligned leaders on one source of truth, and created a blueprint other groups at Workday now want to adopt.

The operational pain: fragmented data and ad hoc reporting

SREs supported critical databases, ETL tools, reporting systems, and code platforms, while handling roughly 7,000 to 8,000 incidents per year across dozens of services. Leadership's routine questions on uptime, incident trends, or MTTR triggered repetitive data pulls, bespoke joins, and inconsistent answers.

The bottleneck was not data scarcity but semantic drift and silos. Each platform exposed different metadata at different grains, so even basic questions like quarterly P2 MTTR required custom work.

A governed semantic layer on Databricks

The team built a classic medallion architecture, ingesting metadata from about 99 connectors into Databricks-managed Delta Lake tables across raw, base, and certified layers. Predictive optimization improved query speed, and managed Delta Lake tables reduced maintenance overhead.

Unity Catalog became the single control plane for data, dashboards, and AI assets. Metric views, authored as version-controlled YAML with sources, joins, dimensions, measures, and filters, encoded agreed KPI logic for MTTR, incidents, and change requests. Storing definitions in Git enabled review and approvals, preventing ad hoc changes. Using the same metric views for dashboards and Genie aligned numbers across visual and conversational experiences.

Conversational analytics and assisted authoring

AI/BI Dashboards now surface incident health, cloud adoption, and job-level DBU usage with flexible filters, replacing one-off consolidations. Genie handles ad hoc questions in natural language and, guided by concise routing instructions, selects the right metric view. The team attached six metric views to one Genie space to cover incidents, change requests, query performance, database performance, cloud adoption, and job execution. A simple helpfulness rating lets Genie reuse validated queries for similar prompts, improving consistency.

Genie Code speeds dashboard editing. Analysts can restyle or reformat visuals by prompt, shrinking changes that once took a week to minutes. Genie One automates scheduled reporting, emailing daily status digests that include AI-driven insights such as potential recurrence patterns and callouts of teams with elevated backlogs or slow resolution times.

Outcomes and roadmap

The unified foundation yielded tangible gains, including about 46 percent faster query performance, shorter pipeline execution times, and lower cost through optimization. Leaders now ask questions against a single semantic layer, reducing interruptions and variability in answers.

Next steps include Slack and mobile alerting through Genie, deeper Agentic AI capabilities, and exploration of Lakebase. Interest from other Workday teams indicates the approach generalizes well once new domains ingest metadata and define KPIs.

Why shared definitions turn AI into a reliable teammate

Agreement on KPI semantics unlocked everything else. Once MTTR and related measures lived in governed metric views, dashboards, Genie answers, and scheduled insights all aligned, accuracy improved, and changes became safer and faster to ship.

FAQ

They wanted governance, semantic consistency, visualization, and AI in a single place without adding new BI tools. Unity Catalog centralizes access control, Delta Lake tables boost performance with predictive optimization, AI/BI Dashboards provides dashboards, and Genie adds conversational access.

Metric views encode vetted sources, joins, dimensions, measures, and column-level context, while routing instructions steer questions to the correct view. Grounding queries in these definitions led the team to report roughly 99 percent accuracy in practice.

SRE KPIs span several domains, including incidents, change requests, query and database performance, cloud adoption, and job execution. Attaching six metric views, combined with routing instructions, lets one Genie space cover the breadth of questions without blurring each domain's semantics.

Dashboards and Genie cover MTTR, incident counts and trends by priority and state, change requests, cloud adoption activity, and job-level DBU usage. Genie One schedules daily emails that include AI-driven insights and predicted recurrence patterns, plus highlights of teams with higher backlogs or slower resolution.

They cite about a 46 percent improvement in query speed, shorter pipeline execution times, and lower cost from better optimization. Leaders also get faster, consistent answers from a single source of truth, reducing manual data pulls and rework.