Let paper and poles pick the pilot. Documents and inspections on data you already own pay first.
Run the boundary test. An output that switches grid equipment, moves a market position, changes a bill, sends a crew or enters the regulatory record clears a higher bar.
Count what already runs. Check what already runs without the AI label: EMS optimization, unit commitment, load forecasts, outage management.
Test "grid-ready AI" on your data. Run the vendor's claim on your own poles and contracts before you count on it.
Measure from day one. Set the baseline and the number to move before the pilot starts.
PG&E says it cut contract sign-off from months to days and saved $1.4 million. Look for your first pay in the same place: documents and inspections, on data you already own.
Forecasting and trading. NRECA counts peak load, operations and asset-utilization forecasting among the four popular AI categories at co-ops; the forecasting lesson takes it from there.
Asset inspection. AMP's cameras on ordinary vehicles log every pole, and its AI flags transformer health, pole lean, vegetation. The inspection lesson clears the backlog.
Outage response. PG&E reports its monitoring center has avoided 19 million outage minutes. The outage lesson owns the control room.
Field scheduling. Oxy's AI-watched cameras ended routine drive-outs to wells that might not need checking. The scheduling lesson carries it.
Wildfire and storm. PG&E says 600-plus cameras distinguish cloud, smoke and ignition and reach responders in minutes. Vegetation management, also on NRECA's list, predicts growth before it threatens reliability or wildfire. The wildfire lesson goes deep.
The customer account. APS call center agents could say which appliance drove an increase, and on what date and time. The customer lessons go deep.
The interconnection queue. Its own lesson clears the queue.
Machine and plant health. Manufacturing's when-to-act-on-a-maintenance-alert lesson owns the judgment.
Pick the first workflow by document volume and error cost. Error cost can be legal. Oxy still works from handwritten contracts from the 1910s that are legally in force. In inspections the exposure is safety and revenue. About two months after AMP's first drives, members were already finding third-party attachments they weren't aware of on their poles. Weigh both kinds of cost when you choose where to start.
Run the boundary test. Put five questions to every output. Does it switch equipment on the grid? Move a market position? Change a customer's rate or bill? Send a crew into the field? Enter a regulatory record? Any yes means a named owner, and the deep lesson on that side of the line sets the bar. James Freeman, chief data and AI officer at Trico Electric Cooperative, says the co-op labels data restricted, confidential or human-only. The co-op's AI cannot connect to operational systems unless its trust architecture approves the integration. Unclear cases default to the stricter standard.
Count what already runs. Check what is already running that nobody calls AI: EMS optimization, unit commitment, load forecasts, outage management. Patrick Bangert, chief of artificial intelligence at Oxy, says predictive maintenance built on temperature, pressure and flow forecasts has been around for decades in oil and gas. List what already runs before buying anything new. Counting it has its own lesson: c5-1, count the AI you already have.
Test "grid-ready AI" on your own data. The recurring question in NRECA's five regional workshop pre-conferences: "how do we really believe AI results." Ajay Waghray, executive vice president and chief information officer at PG&E, says trusted data comes before prediction. Once asset records meet weather and environmental feeds, a pole recorded on the left or the right side of the road changes the failure forecast. Field personnel validate records for traceability, accuracy and verifiability before the data feeds decisions. Run any "grid-ready AI" claim on your own poles and contracts before you count on it. The c7-5 lesson, vendor real AI or relabel, goes further.
Measure from day one. A KPMG survey found 42% of organizations have only partial visibility into their AI spending. Set the baseline and the number to move before the pilot starts. The c1-8 lesson, baseline and bar before the pilot, carries the method. Branndon Kelley, senior vice president of strategy and innovation and chief strategy officer at American Municipal Power, tells peers to pick programs whose return on investment is immediate or within arm's reach.
Exercise (20 minutes): fill in the first-workflow card
Take the workflow you would most like to point AI at first, the one already on someone's list. Fill in the card for it; you won't know every line yourself. Every line you can't fill gets an owner and a date before you close the card. The card download includes a filled-in example of a first workflow that isn't ready yet, and why.
First AI workflow card (PDF)
A blank card and a filled-in example, one page each.
Show the three prompts: fill the card, test it, turn it into a plan
Use AI to fill in and test your card
These three prompts go into your company's approved AI assistant. Use the first to fill in the card one row at a time; it won't guess a number for you. Use the second when the card is done, to hear the case against this workflow going first. The third asks about your industry and your project and turns the lesson into a 30-day plan, with a checklist, rubric or brief for your boss if you want one. Use placeholders for company, system and people names; the help doesn't depend on them.
Prompt 1: Fill in my card with me
Paste this, then add your candidate workflow in one or two sentences and the card below.
Prompt
Before you start: use your company's approved AI assistant, not a personal account. Replace company, system, supplier and people names with placeholders (Company A, System X) and round any figures. The help you get does not depend on them.
You are helping me fill in a first-AI-workflow card for one candidate workflow. The card is pasted at the end.
Rules:
- Ask me about one row at a time, in order, and wait for my answer before moving on.
- Write down only what I tell you. Never suggest, estimate or invent a number, a name, a yes/no answer or a decision for me, even if I ask you to guess. If I ask, say that this row needs a real answer and ask who could give it.
- If I don't know an answer, write "Not known" and ask who should find out and by when. Put that in the Owner / by when column.
- After the last row, show the whole card as a table with three columns: Question, Answer, Owner / by when. Then list every row marked "Not known" with its owner, and remind me that any unanswered row means the workflow is not ready to go first.
My candidate workflow in one or two sentences: [DESCRIBE THE WORKFLOW YOU WANT TO START WITH]
The card:
[PASTE THE CARD]
Prompt from The AI Decision Course by Dr. Tony Hoang, host of The Artificial Intelligence Podcast (Spotify, Apple Podcasts): tonyphoang.com/learn
Prompt 2: Stress-test my card
Paste this, then your filled card.
Prompt
Before you start: use your company's approved AI assistant, not a personal account. Replace company, system, supplier and people names with placeholders (Company A, System X) and round any figures. The review does not depend on them.
Below is a first-AI-workflow card I filled in for one candidate workflow. Act as a skeptical reviewer at the decision meeting and argue the case for NOT FIRST.
For each row:
- Say whether the answer is specific enough to act on (a number, a named role, a date) or too vague.
- Point out where answers contradict each other, for example an error-cost answer of low against a boundary yes, or a vendor-test yes against a blank baseline row.
- Say what evidence you would ask for before accepting it.
Do not rewrite my answers and do not invent facts about my workflow. If a row is blank or says "Not known", treat it as not ready.
Finish with: your recommendation (START HERE, NOT FIRST or RETEST ON a date), the three rows that most need work, and one question I should ask my team on Monday.
My card:
[PASTE YOUR FILLED CARD]
Prompt from The AI Decision Course by Dr. Tony Hoang, host of The Artificial Intelligence Podcast (Spotify, Apple Podcasts): tonyphoang.com/learn
To paste the card into prompt 1, copy its rows from the PDF.
Prompt 3: Make it yours
Paste this and answer its four questions.
Prompt
Before you start: use your company's approved AI assistant, not a personal account. Replace company, system, supplier and people names with placeholders (Company A, System X, Supplier B).
You are helping a leader apply one lesson to their own situation. The lesson's five decisions are below. Use only them and what I tell you. Never invent facts, numbers, vendors, laws or results about my organization or my industry. Where a number is needed and I have not given one, say what to measure and who should measure it, instead of guessing.
First, ask me these four questions, one at a time, and wait for each answer:
1. My industry, and roughly how many people work in my organization.
2. The AI workflow I want to start with, in one or two sentences.
3. Where it stands today: an idea, a vendor demo seen, or a pilot being scoped.
4. What worries me most about it.
Then give me the following. Keep it short and practical:
A. My plan. For each of the five decisions: one action written for my workflow, the role that should own it (a role, not a name), and what should be done within 30 days. If my industry makes a decision harder, for example regulated data, grid-facing outputs or commission scrutiny, say which decision and why in one sentence.
B. Three questions to ask my team on Monday, specific to my workflow.
C. Ask me which one I want: a checklist for my workflow, a rubric to score it, or a one-page brief for my boss. Then write it, and finish with a prompt I can paste later to update it.
Mark anything that rests on an assumption with "(check)".
The five decisions:
1. Let paper and poles pick the pilot: documents and inspections on data you already own pay first.
2. Run the boundary test: an output that switches grid equipment, moves a market position, changes a bill, sends a crew or enters the regulatory record clears a higher bar.
3. Count what already runs: check what already runs without the AI label: EMS optimization, unit commitment, load forecasts, outage management.
4. Test "grid-ready AI" on your data: run the vendor's claim on your own poles and contracts before you count on it.
5. Measure from day one: set the baseline and the number to move before the pilot starts.
Prompt from The AI Decision Course by Dr. Tony Hoang, host of The Artificial Intelligence Podcast (Spotify, Apple Podcasts): tonyphoang.com/learn
Your company's assistant may answer differently, so read what it writes before you use it.
Check yourself
A vendor demo impresses on their data. What must happen before you sign?
Show the answer
A run on your own records; what it did on their data doesn't count.
One budget line: a filing summarizer or a switching-order optimizer. Which goes first?
Show the answer
The summarizer; the optimizer crosses the boundary.
Nobody can say how much AI the company already runs. Where do you look?
Show the answer
List what already runs: load forecasts, unit commitment, EMS optimization, outage management. Count it before buying new.
Sources
Venkat Nimala, Director of Digital Transformation and Enterprise Architecture, Arizona Public Service, on Open Circuit (2026): "How APS dove head-first into AI [partner content]". https://exa.ai/library/podcast/5q07rnrfm6s/episode/274145qj1c8 Produced by Bidgely.
Ajay Waghray, Executive Vice President and Chief Information Officer, PG&E, on Technovation with Peter High (CIO, CTO, CDO, CXO Interviews) (2026): "PG&E CIO Ajay Waghray on Building Trusted Data for Predictive AI". https://exa.ai/library/podcast/4s4ym6nt1dh/episode/ctcgwmbqtx6
Branndon Kelley, Senior Vice President of Strategy and Innovation and Chief Strategy Officer, American Municipal Power, on Public Power Now (2026): "AMP’s Branndon Kelley Details How AMP Members Will Benefit from New AI Program". https://exa.ai/library/podcast/6s94dc35079/episode/j702lc59ly9
James Freeman, Chief Data and AI Officer, Trico Electric Cooperative, and Venkat Banunarayanan, Leads integrated grid and AI initiatives, Business and Technology Strategies team, NRECA, on Along Those Lines (2026): "How Trico Is Taking Its AI Program to the Next Level". https://exa.ai/library/podcast/d2n52kk0nkr/episode/297pp3vvkfx
Patrick Bangert, Chief of Artificial Intelligence, Occidental Petroleum, on Oil and Gas Upstream (2026): "AI in Oil & Gas: From Mainframes to Large Language Models with Patrick Bangert, OXY | Ep 357". https://exa.ai/library/podcast/r347tmk2jzs/episode/vpmdxc3zlkd Sponsored by IFS.
Two pilots are proposed. One reviews the contracts and standards your teams already hold. One forecasts peak load using sensors you have yet to install. Which starts first?
A proposal's output would automatically schedule field crews to cut vegetation near lines. What does that mean under the boundary test?
A pilot proposal promises to reduce outage minutes but names no starting measure. What do you require before it starts?
A vendor says its AI can forecast which poles will fail from your asset records. Those records have never been validated. What comes first?
Where AI pays first in energy and utilities, and the bar it has to clear
Ask about this lesson, or how it applies to your team. Every answer cites the lessons it draws on.
Try asking
Free preview: 3 questions a day, answered from this lesson. Follow-ups remember your earlier questions on this visit. Don't paste confidential details.