The Skill That Outgrows the Tool

I want to start with a blunt, kind truth: learning new automation tools as your defining career skill in 2026 is a risky bet. I say that as the architect at SCE who is accountable for Software, Artificial Intelligence, and Automation. I have a clear vantage point on how these technologies actually work today — and where they’re headed.

Here’s the uncomfortable part: the same technology that’s helped us cut operational costs the last few years is on the verge of becoming worth-less in the sense that the skills that made those tools valuable will be eclipsed. Not because the tools stop producing value, but because something more capable sits above them now.

AI is advancing fast. Tools and the technical know-how that surround them get stale faster than most of us admit. The most valuable capability in 2026 won’t be memorizing UI flows or APIs. It will be the ability to translate real business problems into AI solutions, and to see — systemically — how value actually flows through an organization.

The Lie Everyone Believes

There’s a comfortable story we’ve been sold for decades: “Master the tools and you’re set for life.” Learn Windows. Learn SAP. Learn Pega. Learn the APIs. Stitch the checklist together and voilà — job security.

That story broke long ago. AI is advancing so rapidly that the technical skills you’re trying to master now will likely be automated away before you can fully master them — and before the market pays you for that mastery. This isn’t a new rhythm in history; it’s the same pattern repeating.

Skills at the Margins Invalidate Fast

Let me tell you a short genealogy of a craft.

Agnes Stitchwell, circa 1795, carried forty-seven hand-stitching techniques in her head — French seams, blind hems, embroidery — each one a small miracle of labor and value. Then the Industrial Revolution arrived. The rarity evaporated. Value moved up.

A generation later, her granddaughter Brenna Shuttlecock didn’t master all the hand stitches; she learned to operate a loom, keep it oiled, and feed it cloth. The labor moved from fingers to machine. Later, computers abstracted another layer away. The detailed finger-work at the bottom of the stack kept getting pushed down until something else sat above it.

Fast-forward: Agnes’s distant descendant Cleo Cadwell learned CAD to design patterns to feed automated manufacturing. Cleo shipped designs to customers she reached on TikTok. Then came her daughter, Ada Cadwell: she no longer needs CAD. She prompts an AI, “Create a summer dress with floral patterns suitable for professional settings that mirrors the client’s corporate colors,” and the work is done — or close enough for iteration.

Every major technical wave invalidates the marginal skills of the previous era. As the edges of the stack get automated, the value migrates up the stack.

I lived through this in my own career. I was born the same year the computer chip was invented. In grade school I learned the nitty-gritty: how values lived in memory to change a screen’s colors. Later came programming languages, markup, databases, and systems. At one point I built tools so non-coders could create websites. Each time, the rear-end work fell away and the value moved forward.

Today, AI lets us create applications by describing them in plain English. It’s been around for a couple of years as experiments; now it’s moving into production. You describe the outcome and the system scaffolds code, features, and exception paths as you iterate. That means the thing you get paid for shifts: from finger-work and tool mastery to the living interface between business problems and AI.

So stop memorizing tool features and API docs. It doesn’t matter as much anymore. We have a Sonic Screwdriver — instead of fussing over which buttons to press, worry about the people stuck in repetitive, dangerous, or soul-draining processes. Find them. Care for them. Ask: how can we remove the drag on their day?

First: have empathy for yourself to let go of the past. Second: have empathy for your neighbor and the pain they are in. Third: Start learning business processes and repeatable patterns of value creation so you can eliminate that pain. This used to be the domain of executives; with computers it became the domain of knowledge workers. Now it must be everyone’s concern.

We must Stop cobbling together old modules and databases as an end in themselves.

We must Start identifying problems that remove more than $100,000+ net a year in waste.

The future superstars won’t be those fluent in “the best AI/automation tool.” They’ll be the people who know their business, its broken places, and how to solve those problems.

Communicating With Models Is the New High‑Leverage Skill

AI will soon instantiate entire workflows from natural language. Not small automations — whole business systems. The people who can describe outcomes clearly, structure the logic, and iterate with discipline will be the ones who scale impact.

I believe that in 12–18 months, natural language will create 50%+ of routine workflows by describing outcomes and having models or agents implement them. Deep technical skills will still matter, but primarily to troubleshoot AI-generated edges where data is messy or the process isn’t well understood.

In 24–36 months, expect AI to build full systems from business requirements: CRMs, inventory systems, data pipelines — spun up from the right prompt and the right feedback loops.

Yes, that’s scary. Especially if you’ve built identity and security around deep technical mastery. But it’s also an enormous opening for people who know the business problems and aren’t afraid to speak plainly to an AI. When you move up an abstraction level, you move up orders of magnitude. If today you can drive X, one level up you can drive 10X.

AI’s greatest strength is its flexibility; its greatest liability is that flexibility, too. To channel it into consistent business value you need a framework. That’s the skill to learn.

The POWER Framework

Here’s a compact prompting framework with enterprise legs. It’s simple; I actually used it years ago when I first introduced Copilot; it underpins a lot of high-quality prompts in the wild. Use it.

P O W E R

Acronym: POWER — because everything in utilities comes back to reliable, resilient, measurable power.

• P — Precision: Define the operational or business challenge in measurable, utility-specific terms. Not “optimize the grid,” but “Reduce transformer overload events by 20% in feeder circuits serving >5,000 customers.”
• O — Operations Logic: Give the AI structured system flows it can reason through: work orders, outage sequences, SCADA inputs, switching procedures, maintenance cycles.
• W — Worksite Examples: Provide concrete scenarios from the field: storm response playbooks, wildfire shutoff edge cases, EV load surges, vegetation management cycles. Ground the AI in lived utility operations.
• E — Evolution: Iterate with AI feedback and real-world telemetry. Plans aren’t one-and-done — they evolve with load forecasts, climate models, regulatory updates, and crew feedback.
• R — Reliability Results: Validate outcomes against regulatory, safety, and reliability requirements. Can it pass CAISO/NERC standards? Did it lower SAIDI/SAIFI? Can you prove ROI in outage minutes saved, fines avoided, or megawatts restored?

A bad prompt: “Build me a Wildfire Risk Program.”

A strong prompt: “Create an end-to-end Wildfire Ignition Risk & PSPS Decision-Support system for a Southern California investor-owned utility operating in HFTD Tiers 2–3. Objective: reduce equipment-caused ignitions by 25% year over year while cutting PSPS customer-minutes by 15%, without compromising safety or regulatory compliance. The system covers overhead distribution and sub-transmission segments and their buffers; during Red Flag or Santa Ana conditions it updates every 15 minutes, otherwise hourly. It must support day-ahead (24–72h), near-real-time (0–6h), and post-event analyses. It ingests and time-aligns multiple domains; each line segment receives a 0–100 risk score; automated actions are tied to thresholds; executive/operator briefings auto-generate; integrates with Esri; reports and PM dashboards exist; designed to comply with CPUC; acceptance criteria include five-year backtests showing the claimed reductions.

“It ingests and time‑aligns multiple input domains…

“Each line segment receives a 0–100 risk score…

“Automated actions are tied to score thresholds…

“For executive and operator decisions, the system auto‑generates…

“The platform integrates with Esri…

“Reporting and performance management…

“The system is designed to comply with CPUC requirements…

“Acceptance criteria are explicit. Backtests over five+ years must show…

“The scoring, actions, and integrations can be tailored to any technology stack…”

Send the first prompt and you get a template. Send the second and you start building the autobahn. Flexibility remains — but it’s constrained into the path you need.

Enterprise Thinking Transcends Specific Skills

Why could Serena Williams have dominated basketball if she’d wanted to? Not because she was secretly a hidden basketball prodigy, but because elite athletes think one level up. They understand patterns of movement, training systems, mindset, strategy, recovery—the shape of performance. Techniques live inside that shape.

Business works the same way.

Work Item: Trees and vegetation growing too close to power lines.

1. Inspect the field (patrol crews or drones collect data).
2. Assess risk (is it within clearance distance? species growth rate?).
3. Create a work order.
4. Dispatch a crew.
5. Perform the job (trim, remove, or treat vegetation).
6. Close the work order with before/after evidence.

Final Deliverable: A tree branch safely trimmed or removed.

Work Item: A failing or overloaded distribution transformer in a neighborhood.

1. Inspect the field (crew identifies overheating, oil leak, or overload condition).
2. Assess risk (failure likelihood, customer outage potential).
3. Create a work order.
4. Dispatch a line crew with the right equipment.
5. Perform the job (de-energize, lift out the old transformer, install the new one).
6. Close the work order with completion notes and compliance checks.

Final Deliverable: A new transformer bolted onto a pole or pad, powering homes safely.

Work Item: A required CPUC/DOE compliance report (e.g., wildfire mitigation, PSPS post-event report).

1. Inspect the field (collect data from sensors, systems, and crews).
2. Assess risk (does the data show compliance issues, gaps, or penalties?).
3. Create a report order.
4. Dispatch analysts and writers.
5. Perform the job (compile evidence, draft narratives, validate against rules).
6. Close the work order with submission to the regulator.

Final Deliverable: A formal document submitted to regulators.

One ends with a trimmed branch; another with a new transformer bolted to a pole; another with a PDF submitted to a regulator. Different deliverables, same enterprise workflow. But the skeleton process is the same: inspection → risk validation → work order → execution → closure. Learn the shape of the business, and you will see AI implement it over and over again.

More generally, this shape is: Preexisting conditions necessary for value → Construction → Delivery → Sales → Service. Under each of those are nested steps. These are the containers you should learn to see and fill.

The Takeaways

• The automation skills many of us learned over the last 60 years are being invalidated.
• The new, higher-leverage skill is communicating precise business needs — sometimes to coworkers, sometimes to models.
• The even higher skill is enterprise thinking: seeing the flow of value, recognizing the container, and learning how to fill it.

That skill has a name: Enterprise Architecture (EA) — is a playbook for how a company works, showing how people, processes, technology, and information fit together so leaders can make clear, coordinated decisions that produce value for its owners.

Automation skills always have an expiration date. That sounds bleak; it need not be. History shows us repeatedly that technology raises the floor on human life and work. The hard part is often that dawn looks like a high wall from the bottom of the well. But the sunrise is there — and we can climb.