Energy runs AI through the grid. The decisions that define the business never reach it.
Grid optimisation, predictive maintenance and trading are real wins, and the sector is right to chase them. The harder question is how a leadership team reasons with AI about capital that will outlive every model it runs through.
Request a Strategic Briefing →Energy and utilities have adopted AI the way they adopt any operational technology: scope it to the grid and the asset base, prove it on optimisation and predictive maintenance, run it as a project. Those wins are real. But the decisions that actually define an energy business, the capital commitments under decades-long asset lives, the transition bet, the reliability and regulatory trade-offs, mostly get made the old way, or get handed to a commodity AI that returns a confident generic answer. That is where High-Speed Waste is most dangerous, because the assets last forty years and the mistake compounds for all of them. Gildoni installs the Havruta Methodology (formerly the Think Partner Methodology) into how leadership teams reason with AI, so the judgement that defines the business gets the same discipline the grid already gets.
AI is both a tool the sector runs and a load it now carries
Energy is in a position no other industry quite shares. It is adopting AI to run the grid better, and it is being asked to power the AI everyone else is running. Both pressures land on the same balance sheet at once.
On the operational side, the use cases are genuine and moving fast. AI is forecasting demand, balancing intermittent renewables, optimising dispatch and catching asset failures before they happen. The International Energy Agency is direct about the upside: AI offers the potential to transform how the energy sector works, from grid management to plant efficiency.
On the other side, the same technology is the fastest-growing new load on the system the sector operates. The IEA's chief, Fatih Birol, frames it as one of the defining shifts of the moment, and the numbers below show why an energy leadership team cannot treat AI as someone else's problem.
Three figures that put the scale beyond an IT decision.
terawatt-hours: data-centre electricity use roughly doubles by 2030, and demand from AI-optimised data centres more than quadruples. The sector that runs AI also has to supply it.
of utility control rooms are expected to use AI by 2027, as forecasting, balancing and dispatch move onto AI by default across the operational layer.
in record capital the US electric power sector faces through 2030, capital committed under asset lives measured in decades. This is the judgement AI rarely gets near.
The grid is getting the AI. The capital judgement is not.
AI is one of the biggest stories in the energy world today, but until now, policy makers and markets lacked the tools to fully understand the wide-ranging impacts.
Why the highest-stakes decisions never reach the AI
If the sector is so good at putting AI to work on the grid, why does it stop there? Three reasons, and none of them is a lack of capability.
AI arrives as an operational-technology project, so it inherits an operational-technology scope. It is bought to optimise an asset, a substation, a trading book. The governance, the success measures and the people who own it are all built for the asset layer. The capital decision sits two altitudes up, in a different room, on a different clock, and the AI never makes the trip.
The decisions that define the business resist the project frame. A forty-year asset bet, a transition strategy, a reliability-versus-cost trade-off under a regulator: these are long-horizon, uncertain and political. They do not look like a use case, so they do not get scoped as one, and the reasoning behind them stays where it has always been.
And where AI is invited into a strategic question, it tends to confirm rather than challenge. Ask a commodity model about a transition pathway and it returns a fluent, plausible answer drawn from an industry average, not from your estate, your regulator or your balance sheet.
It is worth being honest about the rest of it. The disciplines of good capital judgement, knowing your real exposure, deciding under deep uncertainty, holding a position a regulator can test, are old and well understood in this sector. AI has not replaced them. It has offered to do them faster, and the temptation is to let it, at exactly the altitude where speed is the most expensive mistake.
Optimising the grid is the easy part
Here is the part the sector's AI story leaves out. Every vendor and analyst will tell you where AI optimises an asset, balances a grid, sharpens a trade. Almost none of them tell you how a leadership team reasons with AI about a capital commitment that will still be running long after the model that informed it has been retired three times over.
That is the real gap, and it is not a tooling gap. Put a forty-year capital question to a commodity AI and it will hand back a confident answer shaped by the average of everyone else's, never asking what your transition assumptions miss or where your reliability case is thin. At the altitude where the assets last decades, that is the Mirror Principle at its most expensive: if the reasoning going in was generic, the capital position coming out is generic, and you are committed to it for forty years. Optimising the grid is a project. Reasoning well about the bet that defines the company is a different discipline.
AI is dense where the clock is short: grid, dispatch, assets, trading. The decision that defines the business runs on a forty-year horizon and receives almost none of that reasoning. The red is the distance between where AI is concentrated and where the judgement actually lives, and how long the company is committed once the bet is made.
What the Havruta Methodology installs at leadership level
The Havruta Methodology is that discipline. It changes the default behaviour of the machine from agreeing with you to reasoning with you, which is exactly what a forty-year capital decision needs.
The Flip
The Flip puts the machine on the other side of the question. Instead of confirming the transition pathway or the capital case, it argues against it: where is this demand forecast optimistic, what would have to be true for this asset to strand, what is the regulator going to ask that the board is not. The leadership team gets challenged before the next forty years do the challenging.
Ground Truth
Ground Truth keeps the reasoning anchored in the organisation's real position, its actual estate, load profile, regulatory obligations and balance sheet, rather than the industry-average view a commodity AI returns by default. A capital decision built on a plausible average is the most expensive output AI can produce in this sector.
Decision Velocity
And Decision Velocity lets the team reason at the speed the transition and the load are moving, compressing the path from question to defensible position, without surrendering the forty-year judgement to a model that will be obsolete long before the asset is.
The fuller account of how all of this works is on the methodology page.
What this is not
This is not grid software and it is not an asset-management platform. It is not an energy-trading tool, not a SCADA or forecasting system, not a digital-twin product. It is not AI training or general AI literacy. The optimisation tooling is a separate market, and a good one. This is the reasoning underneath the capital decisions that tooling never touches.
It changes how the leadership team reasons about the decisions that define the business: the capital commitment, the transition bet, the reliability-versus-cost call under a regulator, the long-horizon judgement no optimisation tool ever reaches.
Where a leadership team starts
The methodology is installed along a ladder, and a leadership team enters at the rung that fits.
Most begin with the Eye-Opener Workshop, a half-day in which the team sees the shift on its own real work.
A leadership group embeds the practice through the Havruta programme, taking the discipline across the team.
A single high-stakes question, a transition pathway, a major capital commitment, a reliability case before a regulator, can be worked through Advisory Havruta.
How a COO and operations leadership reason with AI
For the operations leaders specifically, the role page takes the same discipline to the people who run the asset base and the grid day to day. A Strategic Briefing is how to decide where to begin.
Go to the COO pageLeadership questions about AI in energy and utilities
Is AI in energy a grid-technology project or a leadership decision?
Both, and the trouble is that most organisations only treat it as the first. Grid balancing, dispatch and predictive maintenance are genuine wins worth running as projects. But the decisions that actually define an energy business, the capital commitments under decades-long asset lives, the transition bet, the reliability trade-offs under a regulator, are leadership decisions. Scoping AI to the grid alone leaves that judgement made the old way, or made with a confident generic answer the AI was never the right place to get.
Why does the sector apply AI to the grid but not to capital decisions?
Because AI arrives as an operational-technology project and inherits an operational-technology scope. It is bought to optimise an asset, so its governance, its measures and its owners all sit at the asset layer. The capital decision lives two altitudes up, on a longer clock, in a different room. The reasoning never makes the trip. The highest-stakes, longest-horizon decisions are precisely the ones the project frame is built to exclude.
What does AI's own energy demand mean for an energy leadership team?
It means AI is not just a tool you run, it is a load you have to plan, supply and price. The IEA projects data-centre electricity use roughly doubling by 2030, with AI-optimised demand growing far faster. That reshapes demand forecasts, interconnection queues and capital plans at once. So the same technology your control room is adopting is also a structural new driver of the forty-year decisions your leadership team owns. The two cannot be reasoned about separately.
Is this grid software, a digital twin, or a trading tool?
No. It is not grid or asset-management software, not a forecasting or SCADA system, not a digital twin, not an energy-trading tool, and it does not touch your operational stack. Those address the asset layer, and they do it well. This addresses the thinking above it: how a leadership team reasons through a capital or transition decision so the answer is genuinely theirs, anchored in their real estate and obligations, and stress-tested before forty years of asset life do the stress-testing for them.
How should we reason about AI in a long-horizon capital or transition decision?
Start by refusing to scope it as an optimisation task. Then install the reasoning discipline: make the AI argue against the capital case rather than confirm it, where the demand forecast is optimistic, where an asset could strand, what the regulator will test. Anchor it in your real load, estate and balance sheet rather than an industry average. And decide at the speed the transition is moving without surrendering a forty-year judgement to a model that will be obsolete long before the asset is.
Where should we start?
With a Strategic Briefing, or with the Eye-Opener Workshop, where a leadership team sees the difference between instructing AI and reasoning with it on its own real work. From there the path depends on whether you are setting how the whole leadership group reasons, embedding the practice across the team, or working a single high-stakes question such as a transition pathway or a major capital commitment.