Databricks Deflects 73% of Support Tickets with Agentic AI Assistant R2DB, Saving $1.5M Annually
After expanding its AI assistant R2DB beyond a narrow initial scope, Databricks achieved 73% ticket deflection across IT and HR support for 8,000+ employees, cutting hiring costs by $1.5 million and doubling employee IT satisfaction scores.
Background
Databricks grew from a startup to a global AI leader in a compressed timeframe, but back-office support capacity did not scale with headcount. Employees across IT and HR faced long wait times on questions about onboarding, access provisioning, and internal policies. The company needed a way to deliver consistent, always-on support across dozens of tools and departments without proportional increases in headcount.
What Was Implemented
- Deployed R2DB, an agentic AI assistant built on the Moveworks Agentic AI platform
- Integrated R2DB with 75+ internal tools including Microsoft Teams, ServiceNow, and Microsoft 365
- Provided a single conversational entry point for all IT and HR support requests
- Used Slack to proactively push knowledge to employees within their existing workflows
- Iteratively expanded R2DB's automation scope after initial low deflection rate (10%)
- Fine-tuned processes to balance resolution speed with service quality
Results
After expansion, R2DB deflects 73% of support tickets within its scope , delivering more than 3,500 answers per week and cutting approximately $1.5 million in annual staffing costs . Employee IT NPS doubled from 30 to 70. Human agents were freed for higher-value interactions.
Lessons
- Initial low deflection (10%) was addressed by broadening the AI's scope and workflow integrations, not by replacing the system
- Meeting employees within their existing tools (Slack, Teams) dramatically increased adoption
- Agentic AI value compounds as scope expands; narrow early pilots understate eventual ROI
- Balancing speed with quality during iteration prevents deflection gains from trading off against employee satisfaction