How to use GenAI to create SQL, Python, DAX and M Source to Target documentation in Excel?#
Me and my Colleague have run internal EPAM training on this topic. Goal was simple — enable colleagues to use this approach on projects right away.
Business Value#
- 5 days → 0.5-1 day for S2T Excel documentation
- Business Summary output: distills ~10,000 lines of code into ~10-step description — perfect for senior stakeholders to grasp what the team actually delivers
Key Enabler: Power BI Project Format (.pbip)#
This is the door-opener. Saving Power BI as .pbip exposes tables, visuals, DAX, and M code as plain text files — making them accessible to GenAI.
Bonus benefits:
- Git integration (parallel development, structured CI/CD)
- Version control that actually works (bye-bye SharePoint chaos)
Security note: Data stays local (cache.abf). Only metadata (catalog addresses, schema/table names) goes to repo.
Our Approach vs Alternatives#
What we use: MS Copilot
- Semi-automated, fully controlled workflow
- Low learning curve — minimal ramp-up needed
- Direct Excel/Word output, M365 integrated
Alternative: GitHub Copilot
- More advanced capabilities, broader model selection
- Steeper learning curve, higher entry barrier
- Possibly less control over outputs — still exploring this path
Critical Considerations
- Human review is mandatory — GenAI hallucinations happen, validation is non-negotiable
- Huge time savings — ROI speaks for itself
- Minimal learning required — grab prompts, follow the pattern, deliver
Resources#
Ready to try it yourself? The Sandbox is avalible below. Enjoy your prompt enginering!
🔗 Demo Repository: github.com/datameisterpl/genai-s2t-energy-demo-public