What I mean when I talk about the AI-native Grid and Agentic Grid Operations
For the last months I talked and wrote a lot about AI-native Power Grid, Agentic Grid Operations, Deep Grid Data, Grid Ground Truth or Trusted State. During the weekend, I noticed something quite embarrassing: I never said what I mean with the AI-native Power Grid. I never wrote down my definition. My vision. My thoughts.
AI-native seems everywhere now. And everything seems to be AI-native, too. Quite often it simply means: we put some AI into or onto something that already existed (and worked). For me that is AI-enabled. AI-enhanced.
When I think about AI-enabled, I think about a system or workflow that works exactly the old way, but got some AI tools helping a human here and there with some tasks. Take the AI out, the system or workflow still works, but maybe less effective for the operators than with AI in it.
AI-native on the other hand, in my mind, means the system and workflow is designed around AI. AI is the execution layer. Take the AI out, and the system or workflow stops working as defined and falls back to the operators.
My Definition of the AI-Native Power Grid
The AI-Native Power grid is a power grid with AI being the execution layer of grid operations, instead of a tool bolted on top. The grid runs like a living system: within the limits, policies, and strategies set by human operators, it monitors, operates, balances, optimises and restores supply by itself. The role of the human operators changes from reacting to every event and approving every action, to defining and governing these limits. And the Human operators take over whenever the situations is outside the defined limits.
Or in short: AI does the work. Human operators set the rules. Govern it. And step in when the rules do not fit.
A new Name for the Smart Grid
No. It is not a smart grid with a new or more fancy name. Self-healing is not new. FLISR has done it for years. Rules based self-healing. Rules written by engineers. An AI-native grid also handles the situations. where no rules are written in advance.
It is not a grid with out people. It is not a dark control room. It is not fully autonomous operating in all known or unknown conditions. Human operators are reacting on events anymore. Human operators are not managing the grid anymore. Human operators are not approving every single step or action. Human operators define the policies, strategies, and limits. Monitor those. Govern those. It is not a smaller job. Not at all. It is a different job..
It is distribute, but not fragmented. Decisions can be distributed. Decisions should be distributed. To many agents. Close to where things happen. But all them must act on the same picture of the grid: the Trusted State. Or at least all agents must know exactly on which version / snapshot it can act on.
And it is not a Big Bang. The AI-native Power Grid develops over times. Step by step, task by task, as trust grows. It is a Journey of Trust.
My Definition of the Agentic Grid
When the AI-Native Power Grid describes where we want to go, then the Agentic Grid tells us how to build it. I think about the Agentic Grid being the architecture backbone for the AI-Native Power Grid:
The Agentic Grid is the architecture for the AI-Native Power Grid. Many AI agents, from different vendors, third parties, and from the utility. Each egent with a defined role, tasks, and authority. All agents act on one Trusted State. None of them acts on the grid except through the cognition/control boundary.
Instead of describing it as an architecture stack with different layers, I like to think like a system of systems. A living system.
The grid is measured. The data get snapshotted and certified to a Trusted State. All agents reason on it, within the limits the human operators set. Every action goes through the cognition/control boundary into deterministic control. Control changes the grid, And the grid is measured again. A loop. A living system. AI-native.
How to get started
The AI-Native Power Grid is a journey. We are not jumping from advice to autonomy. It is not a big bang. It is a journey of trust. Task by task. Step by step. But every step needs three things in place:
The one Trusted State of the grid that all agents share and reason on.
A clear cognition / control boundary between where AI may advise and where AI may act.
People who can check, definie the limits and strategies and understand when the AI goes wrong. We have to train the people who will check the machines.
These three are not technology problem. Utilities should start working on them. Today.
Your Points
For now, that is what I mean when I talk / write about the AI-Native Power Grid. My definition. Or vision. Or thinking. For now.
I guess, it will change as we and I all learn more. But would love to get your thoughts and inputs !
Disclaimer: The post was originally published on the AI-native Power Grid Channel on Substack, but adapted for the Energy Central Community.
Quick note on the use of AI: All articles are written by me. All ideas, thoughts, concepts, and opinions are mine. But yes, I am using AI. To challenge my thinking, to validate and fact-check some of the content. To create the banner pictures or diagrams. And occassionally, as I am a non-native speaker, to check my spelling and grammar.