A structured proof-of-concept to validate self-service analytics for HR leadership — grounded in evidence, built on trust.
Six-phase path to an evidence-based go/no-go decision by 15 July.
Power BI dashboards confirm the data is correct. The real question: can non-technical HR users get trusted answers without waiting for analysts?
Power BI works, but every new question requires analyst time, SQL skills, and manual extracts. Days, not seconds.
Natural language queries on live HR data, aligned to business terminology. Same data, same governance — radically faster answers.
Trust, but verify. Every Genie answer is validated against SQL source queries and Power BI screenshots — producing structured evidence for leadership review.
Genie output → SQL source query → Power BI screenshot match → pass/fail score. Fully traceable, fully auditable.
Each response has a verified answer key and context notes for tuning the Genie space. Leadership gets a transparent compliance trail.
From feasibility to go/no-go decision with built-in review and user testing checkpoints.
Clear operational metrics and a structured decision framework for leadership.
Overall accuracy: 67.5%. Privacy compliance: 100%. Strong foundation, iterating on training.
Zero privacy violations in Exam 1. All sensitive queries correctly refused per policy.
Prioritize Easy (85%) and Medium (80%). Hard questions require analyst interpretation.
Exam 2 scheduled for next week. Expecting meaningful performance increase with refined training.
Strategic focus: Easy and Medium questions drive self-service value. Hard questions require analyst interpretation and domain expertise, so we target lower accuracy there by design.
Three clear asks — low commitment, defined timeline, collaborative ownership.
Core team roles and the immediate path to the first formal accuracy report.
Finalize the question bank · Complete phase 2 Genie space training · Produce first formal accuracy report with SQL + Power BI evidence.
Phase 5 includes a formal external validation session with the Databricks team.