NXP Semiconductors · Power BI and Tableau site admin · SQL, AI assistant, Power BI Copilot
New data platform, enabled with AI
I helped 200+ colleagues move to a new data platform with built-in AI, using AI myself to learn it first and fast, and to bridge the SQL differences.
Impact
The challenge
The company was moving its analytics to a new data platform with built-in AI. Most analysts had never used it, and their SQL did not carry over as written.
What I built
- Step 1
AI helps me learn the platform and map the change
AI assistant
I used AI to understand the new platform quickly and to analyze what the change would affect.
- Step 2, human in the loop
Plan the transition as a team
Our admin team worked out together how users could move their work over with as little disruption as possible.
- Step 3
Translate the first SQL samples with AI
The platform’s built-in AI assistant
The biggest technical gap was SQL syntax. I translated the first sample queries with the platform’s built-in AI assistant, which showed everyone the patterns. The platform team later built a translation tool.
- Step 4, human in the loop
AI tool evaluation: the platform’s AI vs Power BI Copilot
I ran both on real stakeholder questions, compared the outputs side by side, and wrote up quality, failure modes and fit, so the teams knew how to approach the new AI tools with confidence. Microsoft Copilot judged each round, so the evaluation itself ran with AI.
- Step 5, human in the loop
Trial runs and recordings
I tried things myself and recorded what worked.
- Step 6
AI drafts the guides; I review them
AI writing assistant
AI turned my recordings and notes into clear guides and playbooks quickly, so users could learn faster. I reviewed every one before it went out.
- Step 7, human in the loop
Office hours, live demos and coaching
AI can draft a guide, but people trust a person who shows it working on their real work. Weekly office hours, live demos at the Power BI and Tableau user groups, 1:1 coaching and a newsletter to every Power BI and Tableau user kept it practical.
What it looked like
The evaluation template: both tools on every question, judged by AI each round
| Sample stakeholder question | Criterion | Platform AI | Power BI Copilot |
|---|---|---|---|
| "Which region grew fastest last quarter?" | Answer correct? | Yes / No | Yes / No |
| Chart useful? | Yes / No | Yes / No | |
| Failure mode, if any | What went wrong | What went wrong | |
| Fit for which team | Team | Team | |
| AI judge's verdict (Microsoft Copilot) | Which answer was better, and why | ||
Where AI helped
AI was in the work from the start: it helped me learn the platform and map the change, the platform’s assistant bridged the SQL differences, in the comparison with Power BI Copilot, Microsoft Copilot judged each round, and AI drafted the guides and playbooks that I reviewed.
Human in the loop: what I kept
Collaboration Our team planned the transition together; I brought my AI impact analysis and my own trial runs into those decisions.
Judgment I decided which AI tool to trust for which job, from real stakeholder questions.
Initiative I tried things first myself, then turned the trial runs into recordings, guides and live demos for everyone.
Cross-functional impact The key BI teams piloted the change with our team, and champions now onboard their own teams.
What happened after
A newsletter went to every Power BI and Tableau user.
Champions now onboard their own teams from the playbooks, guides and knowledge base articles, without me.