Prompts • Skills • Connectors • Agents • Energy workflows • Tools
AI Playbook
for Energy
For engineers, operators, analysts, and managers at utilities, grid operators, power producers, renewable developers, and energy retailers. A practical guide to using AI across grid operations support, asset management, renewables and interconnection, customer programs and billing, trading, and regulatory compliance.
Who this playbook is for
This playbook is for the teams who run utilities and energy companies, not for household energy advice. It covers how utilities, grid operators, power producers, renewable developers, and energy retailers can use AI for analysis, planning, documentation, and communication. Real-time grid control stays with certified operators and approved systems.
- Investor-owned, cooperative, and municipal utilities
- Grid operators and transmission owners
- Independent power producers and renewable developers
- Energy trading and retail teams
- Regulatory, compliance, customer, and asset management leaders
- Homeowners looking for energy-saving tips
- Real-time grid control or SCADA automation
- Investment advice on energy markets
Start With Your Role
Choose the role closest to your day-to-day work. Start with one useful prompt, make it repeatable, then add the information it needs.
- Start with: event after-action summary or study results summary
- Save as a skill: summary format and open-question rules
- Connect: approved event reports and study files (non-BCSI)
- First agent: event report drafter
- Measure: report turnaround, follow-ups closed
- Start with: transformer condition summary or capital prioritization
- Save as a skill: risk ranking and scoring criteria
- Connect: EAM and test result exports
- First agent: asset risk watch
- Measure: failures avoided, overdue maintenance
- Start with: outage communication plan or collections prioritization
- Save as a skill: templates and protection rules
- Connect: aggregate CIS reports, ETR feeds
- First agent: storm communication drafter or collections protection checker
- Measure: complaints, rule exceptions, satisfaction
- Start with: interconnection queue audit or curtailment analysis
- Save as a skill: milestone and study summary formats
- Connect: queue tracker, study reports, plant data
- First agent: queue milestone tracker
- Measure: missed milestones, curtailment revenue
- Start with: day-ahead results review or risk narrative
- Save as a skill: review and report formats
- Connect: ETRM and settlement exports (restricted)
- First agent: day-ahead results reviewer
- Measure: deviation costs, report time
- Start with: CIP audit evidence organizer or rate case request tracker
- Save as a skill: evidence matrix format (titles only)
- Connect: compliance system metadata, filing trackers
- First agent: CIP evidence readiness monitor
- Measure: evidence gaps, on-time responses
100 AI Prompts for Energy Teams
Ready-to-use prompts for grid operations support, asset management, renewable integration, customer programs, trading and markets, NERC/FERC compliance, field operations, forecasting, sustainability, and billing. Copy one, add your organization's context, and review the result.
Use non-sensitive data, quote standards and tariffs exactly, and review the result before it informs any operating, trading, compliance, or customer decision.
Prompts for operations support: event reviews, study summaries, and procedures. Real-time control stays with certified operators and EMS/SCADA systems. Never paste BES Cyber System Information or CEII into unapproved tools.
Turns a dense event report into a reviewable summary quickly.
Risks: AI can misread technical data. Control: Operations engineers validate every statement.
Gives planners a fast, consistent read on where congestion costs come from.
Risks: Correlation isn't cause. Control: Planning engineers confirm with power flow studies.
Shows when forecasts miss ramps so teams can prepare.
Risks: Data quality varies. Control: Forecasting team validates.
Leadership and regulators need a clear version of complex assessments.
Risks: Figures come from approved studies. Control: Planning lead verifies.
A requirement-by-requirement checklist makes plan reviews thorough.
Risks: Restoration plans may be CEII or BCSI. Control: Use only approved tools and non-sensitive excerpts.
Helps engineers communicate results to leadership and stakeholders.
Risks: Simplifying can distort. Control: Study engineer approves the summary.
Connects customer complaints to measured data.
Risks: Settings changes are engineering decisions. Control: Distribution engineers decide.
Facility ratings documentation is a frequent audit focus.
Risks: Ratings are engineering determinations. Control: Engineers certify ratings.
Structured investigations resolve complaints faster.
Risks: Field work requires qualified personnel. Control: Engineer leads the investigation.
Saves engineers time on report writing.
Risks: Technical conclusions belong to engineers. Control: Engineer approves.
12 Claude-Ready Energy Skills
These are downloadable Claude Skill packages, not just longer prompts. Each includes a SKILL.md your team can tailor with your standards, tariffs, and information protection rules. A copy-and-paste version works in ChatGPT and other AI tools.
Connect Your Work
AI can only help with the information it can see. In energy, start with public and business data, and keep operational technology, BCSI, and CEII separate.
Easy start
Start with a limited, low-risk connection your team can test quickly.
- Access: EIA data, ISO/RTO public market and load data, and NOAA weather data; community-built MCP servers for EIA and grid data exist but aren't government- or ISO-supported
- Useful for: market context, forecast reviews, and planning summaries
- Setup & limit: prefer official downloads or APIs; verify any community connector before use
- Access: Outlook, SharePoint, Excel, and Teams connections, or Microsoft Copilot in your tenant
- Useful for: studies, procedures, reports, and regulatory work
- Setup & limit: keep BCSI and CEII sites out of AI scope unless approved
- Access: Gmail, Drive, Sheets, and Calendar connections in Claude and ChatGPT
- Useful for: business documents and analysis
- Setup & limit: follow information classification rules
- Access: scheduled, non-sensitive exports from EAM, CIS (aggregate), ETRM, and planning tools into an approved shared location
- Useful for: asset risk lists, program analysis, market reviews
- Setup & limit: classify each export; name an owner
Needs an administrator
Useful, but someone needs to set access and permissions first.
- Access: Databricks, Snowflake, BigQuery, and Azure offer governed AI and agent features over curated data
- Useful for: forecasting analysis, asset analytics, and customer program analysis
- Setup & limit: governed views only; classification and access controls set by IT
- Access: EAM, CIS, ADMS, and historian vendors (e.g., IBM Maximo, SAP, Oracle Utilities, AVEVA PI) are adding AI features
- Useful for: working inside systems of record
- Setup & limit: assess vendor AI under CIP-013 where applicable; inventory AI service accounts
- Access: Esri ArcGIS and its AI features
- Useful for: asset and vegetation mapping, hosting capacity visuals
- Setup & limit: detailed network maps may be CEII; restrict access
Integration project
Plan the use case, source data, permissions, and owner before connecting.
- Access: PI and other historians hold operational data
- Useful for: summarized, one-way data copies to a business analytics zone for asset and forecast analysis
- Setup & limit: never connect general AI tools into OT networks; use one-way, approved data flows
- Access: CIS and MDM hold customer usage and payment data
- Useful for: aggregate program and billing analysis
- Setup & limit: individual customer data only in approved systems; privacy review required
- Access: ETRM and settlement systems hold positions and confidential deal data
- Useful for: results and risk narratives
- Setup & limit: restrict to the trading and risk teams; follow market conduct rules
Good to know
Limits and honest gaps.
- Access: EMS, SCADA, ADMS, DERMS, and protection systems
- Useful for: nothing in this playbook; general AI tools never connect to OT or issue control actions
- Setup & limit: OT AI belongs in validated vendor systems under your CIP program
- Access: BES Cyber System Information and critical infrastructure information
- Useful for: only tools approved under your CIP-011 and CEII procedures
- Setup & limit: default: keep out of AI tools
Tools
Choose the AI tools and energy software that fit your organization. The goal is to make your existing systems easier to use, not replace everything at once.
LLMs & Foundation Models
8Grid Optimization & DERMS
7Predictive Maintenance Platforms
7Demand & Renewable Forecasting
6Smart Meter & AMI Analytics
7Energy Trading & Risk Management
6Asset Management & GIS
7SCADA & Operational Technology
7Customer Engagement & CRM
7Sustainability & Carbon Tracking
7Data Platforms & Analytics
612 AI Agents for Energy
An agent is an assistant for one recurring job. It gathers the right information, follows your process, and prepares the work for someone to review and decide.
A good first agent handles one task your team already does, reads non-sensitive business data, never touches OT systems, and hands its work to a named person.
# Event Report Drafter ## Job to be done Get after-action summaries to engineers within a day of an event. ## When it runs When an approved event report is filed. ## Approved context Approved event reports (non-BCSI), procedures. ## Operating loop Build timeline; summarize what happened; list data gaps and follow-ups. ## What it delivers A draft after-action summary. ## Human owner Operations engineering manager. ## What it must never do Never recommend real-time actions or reach compliance conclusions. ## How to measure it Report turnaround, follow-ups closed. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# Asset Risk Watch ## Job to be done Flag worsening asset condition and overdue maintenance early. ## When it runs Weekly; when new test results are loaded. ## Approved context Asset registry, condition data, maintenance schedules. ## Operating loop Re-rank assets by risk; flag worsening trends and overdue work; draft planner list. ## What it delivers A weekly asset risk list. ## Human owner Asset management lead. ## What it must never do Never create work orders or change maintenance intervals. ## How to measure it Failures predicted vs. actual, overdue maintenance. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# Queue Milestone Tracker ## Job to be done Never miss an interconnection milestone or deposit. ## When it runs Weekly; 60 and 30 days before each milestone. ## Approved context Queue tracker, tariff milestones, project status. ## Operating loop List upcoming milestones and gaps; flag at-risk projects. ## What it delivers A queue status report. ## Human owner Development director. ## What it must never do Never submit filings or deposits. ## How to measure it Missed milestones, withdrawals avoided. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# DR & VPP Event Performance Reviewer ## Job to be done Measure every demand response or VPP event. ## When it runs Two days after each event. ## Approved context Aggregate event performance, enrollment, incentives. ## Operating loop Compare delivered vs. nominated; analyze opt-outs; update program metrics. ## What it delivers An event performance report. ## Human owner DER program manager. ## What it must never do Never call events or change enrollments. ## How to measure it Delivered kW, opt-out rate, cost per kW. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# Storm Communication Drafter ## Job to be done Keep customers informed consistently during major outages. ## When it runs When a storm is forecast and at each ETR update. ## Approved context Templates, operations-confirmed ETRs and affected areas. ## Operating loop Draft updates by channel; check facts against operations data; prepare critical customer notices. ## What it delivers Channel-ready drafts. ## Human owner Customer communications manager. ## What it must never do Never publish messages or create ETRs. ## How to measure it Update timeliness, customer satisfaction after storms. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# Day-Ahead Results Reviewer ## Job to be done Give the desk a consistent morning review of market results. ## When it runs Each morning after results post. ## Approved context Bids and awards, prices, forecasts. ## Operating loop Compare awards to expectations; quantify deviations; list questions. ## What it delivers A morning results review. ## Human owner Trading desk lead. ## What it must never do Never submit bids or trades. ## How to measure it Deviation costs, review usage. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# Risk Report Narrator ## Job to be done Speed up daily risk reporting. ## When it runs Daily after risk runs complete. ## Approved context Positions, VaR, limits, credit exposure. ## Operating loop Draft narrative; flag limit breaches and near breaches for escalation. ## What it delivers A daily risk narrative. ## Human owner Middle-office risk manager. ## What it must never do Never change positions or limits. ## How to measure it Report timeliness, breaches escalated. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# PPA Performance Tracker ## Job to be done Track contract delivery and obligations every month. ## When it runs Monthly after settlement. ## Approved context Delivery and settlement data, contract terms. ## Operating loop Compare to contract; flag shortfalls and obligations at risk. ## What it delivers A PPA performance report. ## Human owner Asset or contracts manager. ## What it must never do Never contact counterparties. ## How to measure it Shortfall penalties avoided. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# CIP Evidence Readiness Monitor ## Job to be done Keep audit evidence current and gaps visible. ## When it runs Monthly; weekly in the 90 days before an audit. ## Approved context Evidence inventory (titles and locations only), requirement list. ## Operating loop Update requirement-to-evidence matrix; flag stale or missing evidence. ## What it delivers An evidence readiness report. ## Human owner NERC compliance manager. ## What it must never do Never access BCSI content or make compliance determinations. ## How to measure it Evidence gaps closed, audit findings. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# Rate Case Data Request Tracker ## Job to be done Keep every data request on time. ## When it runs Daily during a rate case. ## Approved context Request tracker, assigned owners, due dates. ## Operating loop List due and overdue requests; draft response outlines from supplied data. ## What it delivers A daily request status report. ## Human owner Regulatory affairs manager. ## What it must never do Never file responses. ## How to measure it On-time responses, follow-ups. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# Forecast Error Monitor ## Job to be done Spot forecast degradation before it costs money. ## When it runs Daily for short-term; monthly for medium-term. ## Approved context Forecast vs. actual, weather data. ## Operating loop Calculate error and bias; flag degradation; list likely causes. ## What it delivers A forecast performance alert. ## Human owner Forecasting lead. ## What it must never do Never change models or operating plans. ## How to measure it MAPE trend, bias. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
# Collections Protection Checker ## Job to be done Make sure collections activity follows disconnection protections. ## When it runs Before each collections cycle. ## Approved context Collections worklist, protected statuses, rules. ## Operating loop Remove protected accounts; flag assistance-eligible customers; check rule compliance. ## What it delivers A checked worklist. ## Human owner Credit and collections manager. ## What it must never do Never disconnect service or contact customers. ## How to measure it Rule violations, assistance enrollments. ## Operating rules 1. Use only approved sources and follow the linked energy skill. 2. Cite source records and separate facts from assumptions. 3. Route safety, reliability, and compliance concerns to the human owner immediately. 4. Keep an audit record of trigger, inputs, output, reviewer, decision, and override. 5. Stop and ask for human review when standards, tariffs, or approval authority are unclear. 6. Never access BCSI, CEII, or individual customer data outside approved systems, and never connect to operational technology networks.
AI for Grid Operations & Optimization
Use AI to summarize events and studies, review forecasts, and prepare planning inputs. Real-time control stays with certified operators and approved systems.
- Prompts: frequency event summary, congestion review, ramp event review
- Skills: grid event & study summaries; forecast performance review
- Connect: historian and market data exports (non-BCSI), study reports, public EIA and ISO data
- Agent: event report drafter; forecast error monitor
- What AI does: Improves load forecasts that feed your EMS and dispatch tools
- Integrates: Weather, calendar, behind-the-meter solar, and history
- Control: Operators and EMS make every dispatch decision
- What AI does: Improves solar and wind forecasts and flags likely ramp events
- Handles: Cloud cover swings, wind ramps, and forecast uncertainty bands
- Control: Accuracy varies by site and horizon; measure it yourself
- What AI does: Flags equipment and weather risk patterns ahead of events
- Supports: Crew staging, storm planning, and after-action reviews
- Control: Operators decide all switching and restoration steps
- What AI does: Summarizes dispatch and market results for review
- Factors: Fuel costs, congestion, renewable availability, constraints
- Control: Dispatch optimization runs in approved, validated systems
- What AI does: Analyzes grid topology, identifies congestion, simulates upgrade impact
- Uses: Historical flow data, weather patterns, growth projections to plan infrastructure
- Result: Avoid over-building. Right-size upgrades based on predicted demand.
- What AI does: Coordinates solar, batteries, EVs, and loads as one virtual power plant
- Orchestrates: Who charges when, who exports energy, how to respond to grid signals
- Enables: Islanding during outages, peak shaving, demand response at scale
Grid Operations Implementation Checklist
WorkflowPre-Implementation
- Map current grid topology and SCADA systems
- Baseline grid stability metrics and outage frequency
- Identify data sources: SCADA, PMU, AMI, weather, market
- Define success KPIs: frequency stability, outage duration, cost savings
- Assess cyber and operational risk tolerance for AI integration
Post-Implementation
- Monitor AI-assisted dispatch decisions weekly
- Track forecast accuracy vs. baseline (hourly, day-ahead)
- Measure grid stability metrics: SAIDI, SAIFI, frequency excursions
- Audit operator feedback and AI recommendation acceptance rates
- Review cost savings from optimized fuel use and avoided outages
- Operators decide: AI never issues dispatch, switching, or control actions; certified operators and approved EMS/SCADA/ADMS systems do.
- Protect BCSI and CEII: Keep them out of tools not approved under your CIP-011 and information protection programs.
- No OT connections: General AI tools never connect to operational technology networks.
- Explain assumptions: Any AI-assisted analysis used in planning lists its data sources and assumptions.
- Vendor AI in OT products: Assess under CIP-013 supply chain risk management and inventory AI service accounts.
- Audit trail: Log AI-assisted analyses used in operating or planning decisions.
AI for Asset Management & Maintenance
Use AI to rank asset risk, organize inspection results, and plan maintenance and capital. Engineers and qualified technicians decide the work.
- Prompts: transformer condition summary, pole inspection summary, spare parts review
- Skills: asset health & maintenance planning; field operations support
- Connect: EAM and inspection exports (Maximo, SAP), test result files
- Agent: asset risk watch
- What AI does: Assigns health score 0-100 to each asset based on sensor data, age, history
- Predicts: Remaining useful life (RUL) and failure probability in next 30/60/90 days
- Impact: Move from calendar-based to condition-based maintenance
- What AI does: Ingests vibration, temperature, pressure from thousands of devices
- Detects: Anomalies humans miss: bearing wear patterns, insulation degradation, corrosion
- Outcome: Catch issues at Stage 1, not Stage 5 when catastrophic failure imminent
- What AI does: Flags possible defects in drone and patrol imagery for review
- Scales: Reviewers focus on flagged images instead of every frame
- Control: Qualified inspectors confirm every finding
- What AI does: Routes maintenance crews based on failure risk, geography, skill requirements
- Schedules: Preventive visits to risky assets before they fail, clustering nearby locations
- Reduces: Emergency callouts. Maximizes crew productivity and asset uptime.
- What AI does: Models which assets reach end-of-life when, cost of replacement vs. repair
- Optimizes: Annual capex spending. Avoids clustered failures from aging cohorts.
- Enables: Data-driven budget requests with 5-year replacement roadmap
- What AI does: Highlights spares exposure for long lead time equipment
- Supports: Reorder points and shared spares discussions
- Control: Materials and asset managers decide purchases
Asset Management Implementation Checklist
WorkflowPre-Implementation
- Catalog all critical assets: transformers, breakers, poles, cable routes, substations
- Install or connect sensors to highest-risk assets first
- Establish baseline maintenance costs and downtime rates
- Define asset health metrics and failure modes by equipment type
- Select drone platform and AI vision vendor for pilot inspection route
Post-Implementation
- Compare actual failures to AI-predicted failures weekly; refine models
- Track maintenance cost savings from preventive vs. reactive work
- Measure inspection efficiency: assets checked per day, cost per asset
- Monitor crew satisfaction with work order routing (reduced emergency calls)
- Review spare parts inventory turns and stockout incidents monthly
- Model accuracy: Validate AI health scores against actual failure outcomes monthly
- Safety first: Only qualified technicians perform work flagged as safety-critical
- Data quality: Audit sensor data for gaps, outliers, and calibration drift quarterly
- Privacy: Drone images not stored/shared without encryption and access controls
- Escalation: Any AI-flagged 'imminent failure' triggered manual inspection within 48 hours
- Transparency: Maintenance teams can see exactly why asset was scheduled for work
AI for Customer Engagement & Billing
Use AI to draft outage and billing communications, analyze programs, and prepare collections worklists that respect protections. Customer data stays in approved systems.
- Prompts: outage communication plan, high bill templates, collections prioritization
- Skills: outage & customer communication; DER & customer program analysis; sustainability & billing analytics
- Connect: CIS and CRM reports (aggregate), outage management ETR feeds, program data
- Agent: storm communication drafter; collections protection checker
- What AI does: Predicts grid failures and sends notifications before customers lose power
- Provides: Realistic outage duration, restoration ETA, affected area, reason for outage
- Builds: Trust. Reduces angry calls to support. Lets customers prepare.
- What AI does: Turns aggregate usage patterns into program and messaging ideas
- Suggests: Targeted efficiency offers by segment
- Control: Savings claims come from EM&V, not AI estimates
- What AI does: Answers common billing and program questions from approved content
- Handles: Routine questions with a clear path to a person
- Control: Measure resolution and handoff rates; disconnection and hardship questions go to staff
- What AI does: Flags meter data anomalies and repeated estimates for review
- Supports: Faster investigation of disputes
- Control: Billing staff approve every correction
- What AI does: Offers customers incentives to shift usage to off-peak hours
- Optimizes: Who participates when based on grid need, customer schedule, and preferences
- Outcome: Flattens demand curve. Reduces peak generation costs. Increases renewable absorption.
- What AI does: Segments customers by behavior and sends targeted messages (SMS, app, email)
- Tailors: Message tone and content. Bilingual. Proactive for at-risk customers.
- Improves: Customer satisfaction scores, payment rates, engagement with programs
Customer Engagement Implementation Checklist
WorkflowPre-Implementation
- Baseline current customer satisfaction: NPS, CSAT, support wait times
- Identify top support questions and outage notification gaps
- Define customer segments and engagement goals per segment
- Select AI chatbot platform and integration with billing system
- Prepare meter data for analysis (privacy-compliant, aggregated).
Post-Implementation
- Monitor chatbot satisfaction, handoff rates, and resolution accuracy
- Track engagement with proactive outage alerts (open rates, customer actions taken)
- Measure adoption of energy efficiency recommendations and usage impact
- Review demand response participation and revenue from peak reduction
- Compare customer NPS before/after AI initiatives quarterly
- Privacy first: Never share individual meter data without explicit customer consent
- Transparency: Explain how AI recommendations are made; show comparable peer usage
- Easy escalation: Chatbot must offer human agent option within 2 interactions
- No manipulation: Demand response offers must be fair. No predatory pricing for vulnerable groups.
- Bias check: Audit recommendations by income, zipcode quarterly for fairness
- Data security: All customer data encrypted in transit and at rest; access logs audited monthly
AI for Renewable Integration & Interconnection
Use AI to analyze curtailment, track interconnection milestones, summarize studies, and evaluate storage and DER programs. Engineers and development leads decide.
- Prompts: curtailment analysis, interconnection queue audit, hosting capacity summary
- Skills: interconnection & development tracking; DER & customer program analysis
- Connect: queue trackers, study reports, plant monitoring exports, public ISO and EIA data
- Agent: queue milestone tracker; DR & VPP event performance reviewer
- What AI does: Reviews forecast accuracy by site and horizon
- Uses: Forecast vs. actual, weather, and plant data
- Control: Measure accuracy yourself; vendor claims vary
- What AI does: Reviews storage dispatch results against warranty and market outcomes
- Supports: Strategy discussions and warranty compliance
- Control: Dispatch runs in approved optimization systems
- What AI does: Analyzes curtailment by site and reason
- Supports: Storage, contract, and hedge decisions
- Control: Curtailment instructions come from the ISO and operators
- What AI does: Coordinates rooftop solar, home batteries, EVs as one distributed power plant
- Manages: Local trading, islanding during grid failures, demand response
- Outcome: Higher renewable utilization. Better resilience for communities.
- What AI does: Summarizes market results and curtailment revenue impact
- Covers: Day-ahead, real-time, ancillary services, RECs
- Control: Traders decide bids and hedges
- What AI does: Analyzes wind/solar resource maps, grid capacity, land availability
- Simulates: Generation, transmission costs, and ROI for candidate sites
- Enables: Data-driven siting decisions and environmental impact assessment
Renewable Integration Implementation Checklist
WorkflowPre-Implementation
- Assess current renewable penetration % and grid stability limits
- Install anemometers/irradiance sensors at key locations or use satellite data
- Establish baseline renewable curtailment and wasted generation
- Define grid services markets and revenue streams available
- Select AI forecasting and dispatch platform
Post-Implementation
- Monitor forecast accuracy daily; refine models based on misses
- Track renewable penetration % and curtailment rates weekly
- Measure battery dispatch optimization: revenue vs. naive peak-shaving baseline
- Audit microgrid islanding tests and DER response times monthly
- Calculate revenue impact from better forecasting and trading decisions
- Stability limits: AI analysis never overrides ramp limits, reserve requirements, or ISO instructions.
- Reserves: Reserve requirements come from your reliability standards and BA obligations, not AI tools.
- Curtailment: Follow ISO/RTO and contract rules for curtailment order and compensation.
- Storage warranties: Review dispatch strategies against warranty cycle and state-of-charge limits.
- Interconnection: Tariff milestones and deposits are confirmed by interconnection counsel.
- Transparency: Document assumptions behind any AI-assisted forecast or revenue analysis.
AI for Energy Trading & Markets
Use AI to review market results, summarize rule changes, compare contracts, and draft risk reports. Traders and risk managers decide every position.
- Prompts: day-ahead results review, rule change summary, term sheet review
- Skills: market results & risk reporting; contract & PPA review
- Connect: ETRM and market settlement exports, ISO public data, contract library
- Agent: day-ahead results reviewer; risk report narrator; PPA performance tracker
- What AI does: Awards vs. expectations and deviations
- Helps: Better bidding over time
- Control: Desk decides
- What AI does: What changed and who's affected
- Helps: Adapt faster
- Control: Regulatory confirms
- What AI does: Deviations and risk allocation
- Helps: Safer deals
- Control: Legal decides
- What AI does: Daily positions, VaR, and limits
- Helps: Clear oversight
- Control: Escalate per policy
- What AI does: Structured options for committees
- Helps: Better risk decisions
- Control: Risk committee decides
- What AI does: Delivery vs. contract obligations
- Helps: Avoid penalties
- Control: Contracts verifies
Trading & Risk Control Checklist
ControlsBefore You Start
- Confirm market data license terms
- Document risk policy and limits
- Export results and positions
- Baseline: deviation costs, report time
- Define escalation rules
After You Start
- Review results daily
- Escalate limit issues per policy
- Track contract performance monthly
- Review rule changes as filed
- Audit AI-assisted reports
- Decisions: Traders and risk managers make every bid, trade, and hedge.
- Market conduct: Follow FERC market behavior rules and your compliance program; never use AI to coordinate bids with others.
- Data: Market data use follows license terms; confidential positions stay in approved tools.
- Contracts: Legal approves all terms.
- Escalation: Limit breaches follow the risk policy.
AI for NERC/FERC & Regulatory Compliance
Use AI to organize evidence, track deadlines, and prepare regulatory work, without exposing BCSI or CEII. Compliance staff and counsel decide.
- Prompts: CIP audit evidence organizer, reliability event checklist, rate case request organizer
- Skills: NERC/FERC compliance evidence; rate case & regulatory support
- Connect: compliance management system (evidence titles and locations only), regulatory filing trackers
- Agent: CIP evidence readiness monitor; rate case data request tracker
- What AI does: Requirements mapped to evidence titles
- Helps: Organized audits
- Control: No BCSI in the tool
- What AI does: Reports, filings, and data requests
- Helps: Nothing late
- Control: Compliance confirms dates
- What AI does: Exercise and event reports
- Helps: Improvement and evidence
- Control: Keep CEII protected
- What AI does: Orders and tariffs to programs
- Helps: See gaps
- Control: Counsel interprets
- What AI does: Data request tracking and outlines
- Helps: On-time responses
- Control: Witnesses approve
- What AI does: Inventory and vendor AI review
- Helps: CIP-aligned AI use
- Control: CIP-013 and CIP-011 apply
Compliance Implementation Checklist
ControlsBefore You Start
- Classify information (BCSI, CEII, confidential)
- Approve tools per classification
- Inventory AI tools and vendor AI features
- Baseline: evidence gaps, late items
- Name compliance owners
After You Start
- Review evidence monthly
- Track deadlines weekly
- Review vendor AI under CIP-013
- Treat AI tools as service accounts under CIP-007 where applicable
- Report AI use to compliance leadership
- Standards: CIP standards neither permit nor prohibit AI; existing requirements apply to AI tools' access (per FERC staff's FY2026 CIP audit lessons learned).
- BCSI: Limit AI tools' access to BCSI under CIP-011.
- Supply chain: Assess vendor AI risk under CIP-013, using frameworks such as the NIST AI RMF.
- Determinations: Compliance staff and counsel decide; AI organizes.
- Filings: Counsel and regulatory staff file.
Set Up & Get Running
If you already use prompts, here is how to turn one into a repeatable skill and then a useful first agent. Start with a task that already works.
From prompt to first agent
- Choose a recurring business or engineering-support task
- Use non-sensitive data
- Good first choices: forecast error review, asset condition summary, or outage communication templates
- Write down inputs, output, and reviewer
- Test on three real cycles
- Improve when results are unclear
- Start with public data and business exports
- Classify every source
- Keep OT, BCSI, and CEII out
- Start with a monitoring task
- Inventory it as a service account where CIP applies
- Track accuracy and time saved
It's useful, measurable, and uses no sensitive data. It compares forecasts to actuals, tracks bias, and flags degradation for the forecasting lead. It doesn't change models or operating plans.
Copy these builders
Use the first after a prompt works. Use the second only after the skill has been tested on real work.
Days 1-30: Foundation
- Name an executive sponsor and involve CIP compliance and IT security
- Write AI rules aligned with CIP-011, CEII, and privacy
- Set up enterprise accounts
- Inventory existing AI tools and vendor AI features
- Pick 1-2 pilots with non-sensitive data
- Record baselines
Days 31-60: Expand
- Save pilots as skills
- Add a 2nd workflow (asset risk or compliance evidence titles)
- Assess vendor AI in CIP-013 plans
- Train engineering and business teams
- Measure time saved and accuracy
Days 61-90: Standardize
- Adopt the AI policy
- Pilot one review-only agent
- Write SOPs
- Report results to leadership and compliance
- Maintain the AI tool inventory
- Plan the next wave
Implementation Success Measures
Measurement30-Day Targets
- No sensitive data in unapproved tools
- Pilots running
- Baselines recorded
- AI inventory started
60-Day Targets
- Time and accuracy measured
- Second workflow live
- Vendor AI assessed
- Teams trained
90-Day Targets
- Policy adopted
- First agent piloted
- SOPs written
- Inventory current
- Decision on next wave
How the Pieces Fit Together
Start with a prompt. Save the ones that work. Add the systems and files they need. Then use an agent for recurring work, with someone checking the result.
- Ask AI to do one clear task
- Use it for a quick draft, analysis, or explanation
- Check the result before you use it
- Save the step-by-step instructions for a task your team repeats
- Add your policies, examples, and review steps
- Update it when your process or regulations change
- Bring the right systems, files, and reports into the task
- Stop copying and pasting information by hand
- Start read-only, with only the sources the task needs
- Run a skill on a schedule or when something happens
- Gather the information, prepare the work, and flag issues
- Leave grid control, trading, compliance, and customer actions with people
Where AI Can Help Energy Teams
Use AI to get analysis, documentation, and communication into a clear first draft. Keep operations, trading, compliance, and safety decisions with qualified people.
- Event and study summaries
- Forecast performance reviews
- Asset risk rankings
- Maintenance and capital inputs
- Outage and high-bill communications
- DR, VPP, EV, and efficiency analysis
- Collections worklists that respect protections
- Rate impact summaries
- Market result and risk narratives
- Interconnection milestone tracking
- Compliance evidence matrices
- Rate case and regulatory drafting
- Real-time operations and switching
- Bids, trades, and hedges
- Compliance determinations and filings
- Field safety and maintenance decisions
AI Capabilities Explained
No jargon. What AI actually does in energy operations, in plain English.
Guardrails: Reliability, Security & Customers
Simple rules for using AI in energy. The grid, critical infrastructure information, and customers' data all demand care.
- AI never issues dispatch, switching, or control actions
- No general AI tool connects to OT networks
- Operating decisions stay with certified operators
- Blackstart and restoration never depend on AI tools
- BCSI only in tools approved under CIP-011
- CEII handled per your procedures
- Detailed network models restricted
- Inventory AI tools as service accounts where CIP applies
- Individual usage data stays in approved systems
- Disconnection rules applied by staff
- Fair program design and bias checks
- Clear path to a person in chatbots
- Assess vendor AI under CIP-013
- Use frameworks like the NIST AI RMF
- Know what new AI features can access
- Monitor AI tool communications
- Measure forecast and model accuracy yourself
- Cite standards and tariffs exactly
- Label hypotheses
- Verify sustainability claims
- Grid operations and switching
- Bids, trades, and hedges
- Compliance determinations and filings
- Field safety and maintenance decisions
Energy AI Readiness Checklist
ControlsLeadership & Policy
- Executive sponsor named
- AI policy aligned with CIP, CEII, and privacy programs
- Information classification applied to AI use
- AI tool inventory maintained
- Vendor AI assessed in supply chain risk plans
Practice & Review
- No AI connections into OT networks
- Human approval on filings, bids, and customer actions
- Aggregate customer data by default
- Accuracy monitored for AI-assisted forecasts
- Quarterly review of AI use and incidents
Scope: Business and engineering support functions: planning, asset management, customer operations, billing, trading support, regulatory, compliance, sustainability, and field support.
Approved tools: [enterprise AI accounts approved by IT security; vendor AI features assessed under CIP-013]
Approved uses: event and study summaries, asset risk rankings, program analysis, communication drafts, market result reviews, evidence organization, regulatory drafting.
Prohibited: connecting AI tools to OT networks; issuing control actions; entering BCSI, CEII, or individual customer data into tools not approved for them.
Human decisions: operations and switching, maintenance and safety, bids and trades, compliance determinations, filings, customer disconnections and credits.
Oversight: AI tools inventoried; access authorized and revoked under CIP-004 where applicable; communications monitored.
Approval: [CIO / CISO / Chief Compliance Officer], [date]
AI Maturity Model for Energy & Utilities
Assess your readiness. Define target state. Plan progression.
Manual + Ad Hoc
- Operators using ChatGPT occasionally for analysis
- No integrated AI tools in SCADA or EMS
- Zero governance or measurement
- Demand forecast still done manually or with spreadsheets
- Maintenance reactive, based on failures not predictions
Tool-Based Adoption
- 1-2 AI tools deployed (forecasting or asset scoring)
- Basic data integration with SCADA
- Simple metrics tracking (accuracy, cost savings)
- Staff trained on approved tools
- Some cyber security controls in place
Embedded AI Workflows
- 3-5 AI tools integrated into daily grid operations
- Formal AI governance policy and decision audit trails
- Full team trained; SOPs documented
- Monthly KPI reviews and continuous improvement
- AI-assisted decisions override-able by humans in seconds
AI-Native Grid Operations
- 6-10+ AI tools spanning generation, transmission, distribution
- Predictive maintenance and condition-based work on 80%+ of assets
- Agentic AI: semi-autonomous systems optimizing grid in real time
- Cross-functional AI (ops, trading, customer, engineering)
- Board visibility; AI as competitive advantage for reliability/cost
Maturity Self-Assessment
AssessmentOrganization
- AI champion assigned (CIO, CTO, or chief engineer)
- AI governance board formed (IT, OT, ops, legal, security)
- Executive sponsorship and budget allocated
- Training program established for staff
Technology & Process
- 3+ AI tools deployed and integrated with existing systems
- Demand and renewable forecasting automation live
- Predictive health scoring on critical assets
- SCADA, EMS, and billing system integrations working
Controls & Compliance
- Manual override and human control always available
- Audit trail logging for all AI recommendations and actions
- Cyber security assessment completed; isolated networks for critical systems
- Regulatory alignment verified (NERC, FERC, state)
Measurement
- Baseline KPIs established (outage duration, cost, forecast accuracy, RUL)
- Monthly ROI tracking (fuel savings, maintenance cost reduction, reliability)
- AI recommendation acceptance rates by operators (should be >80%)
- Continuous improvement loop (feedback, model retraining, SOP updates)