This live workshop drew more than 50 attendees for an hour-plus of demos, discussion, and post-event networking in breakout rooms.
Amit Patel walked through real-world use cases showing how utilities can safely bring AI into day-to-day operations, from compliance and document automation to forecasting, dashboards, and live data analysis
The theme was simple.
AI only works when it’s reliable, secure, and grounded in real data.
When applied correctly, it becomes a force multiplier—helping teams cut reporting hours, remove manual steps, and operationalize repeatable code rather than one-off experiments.
Watch the full recording, go through the slide deck or read below for key takeaways on building trustworthy, practical AI inside the plant.
First, meet our expert:
Amit Patel, Managing Partner @ Integ Consulting
Amit leads utility AI and compliance solutions across North America. With over 25 years in power generation analytics, he helps plants operationalize AI securely—transforming reporting, forecasting, and compliance.
Before you get into the summary…
Share GridInnovationHub.com with your peers, colleagues, and friends.
The data problem utilities face
You have more data than you can use, but less access than you need. Integration silos and long IT cycles slow you down. You need interfaces that speak plain English, not SQL.
Prioritize a governed data layer you can query safely
Map the top 20 data sources to use cases, not to systems
Give teams natural language access with role-based controls
Track data lineage so answers can be trusted and cited
Why utilities are slow to adopt AI
Your cybersecurity bar is high. Generic enterprise AI does not meet plant, grid, or customer ops needs. Group-specific AI beats enterprise-wide AI when accuracy and context matter.
Start with OT-adjacent use cases where data is controlled
Build small, group-specific models or prompt libraries
Involve security on day one, define logging and access rules
Prove value in weeks, then scale to neighbor groups
Chat with documents and data
Put your document repositories behind governance, then let teams query in plain English. With the right setup, you can reach 90 to 95 percent answer accuracy with references.
Index SOPs, directives, compliance reports, emails where needed
Require citations to the source paragraph for every answer
Enforce retention, permissions, and version control
Measure accuracy weekly and tune prompts and sources
Document writing automation
AI drafts compliance documents, SOPs, and directives. You review, edit, and approve. Teams see up to 80 percent time savings and higher consistency.
Use approved templates and styles as the base
Auto-fill boilerplate, tables, and cross-references
Keep human review on by default, with tracked changes
Expect 2,000 to 3,000 hours saved per year at fleet scale
Dashboard and visualization creation
Ask for charts in plain English, get Power BI-style outputs. Great for GADS analysis and recurring KPIs. Once validated, lock the template.
Define metric names, units, and filters up front
Compare auto-charts to ground truth before sharing
Export to BI for refresh and permissions
Promote only the tested dashboards to production
Alerts, monitoring, and automation
Automate routine checks and responses, like solar PI tag validations. Make the code self-documenting and run it in a safe container.
Specify triggers, thresholds, and escalation paths
Force the AI to output code plus a readable runbook
Execute in isolated, signed containers with logs
Store every decision and change for audit
Data analysis and forecasting
Chat with live data sources to build repeatable analyses. Add group knowledge so the AI behaves like your team.
Create templates for EAF forecasting, load and cost code analysis
Attach definitions, formulas, and exceptions as policies
Version prompts and keep sample outputs for QA
Schedule runs and compare forecasts to actuals
The two-mode framework
Review Mode: AI creates, you verify and test. Citations on, logs on, no automated actions.
Operationalized Mode: After testing, it becomes reliable code. No free-form AI or hallucinations, just deterministic runs.
Bottom line: Start in review mode to explore quickly and safely. When a result is correct, operationalize it so it runs as secure, deterministic software.
That’s how utilities get reliable AI in production—faster documents, faster dashboards, fewer manual steps, and better decisions.
Explore the full deck
![[Workshop] Applying AI to GADS & Power Plant Performance.pdf](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/static_assets/file_attachment.png)
Thanks for reading, here are some more grid technology resources for you to check out:
👉 Watch the whole 60-minute panel event here.
👉 Find more grid technology panels, meet-ups, and events here.
👉 Register for our networking coffee chats here.
And if you like our content, check out some of these newsletters we love:
