Aug 20, 2026 · 40m · catalyst
Can AI revolutionize grid operations?
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Host Shayle Kann interviews ThinkLabs CEO Josh Wong to explore how physics-informed artificial intelligence can resolve legacy utility bottlenecks, compress grid interconnection timelines, and modernize both long-term resource planning and real-time control room operations.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Shayle holds 24.8% of the talking time here. How this is scored →
Shayle as informed peer 4.6 Guest teaching 5.2 Guest disagreement 1.6 Shayle pushing back 1.6
speaking balance: gold is Shayle, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 28:53 Direct disagreement on AI potential in operations
Josh immediately rejects Shayle's premise that grid operations has less to gain from AI, stating it is the complete opposite and holding even higher value.
Hardest push from Shayle ▶ 28:26 Host questions AI value ceiling in operationsShayle pushes back against the narrative that operations needs AI as urgently as planning, pointing out that operational systems already possess high field automation.
Biggest teaching moment ▶ 28:53 Explaining continuous real-time analysis enginesJosh educates Shayle on how operational AI shifts grid studies from episodic manual reports into continuous real-time autopilot systems with closed-loop feedback.
Shayle holds their own ▶ 16:10 Technical framing of LLM training and inferenceShayle demonstrates deep technical literacy by mapping the distinction between training and inference in coding LLMs to physics-informed grid foundation models.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Shayle as informed peer | Guest teaching | Guest disagreement | Shayle pushing back | Why |
|---|---|---|---|---|---|---|
| Understanding Traditional Utility Planning and Its Bottlenecks | 5 | 5 | 1 | 2 | Shayle demonstrates solid industry context on utility resource planning and asks targeted questions on transmission versus distribution. Josh provides deep technical detail on worst-case scenario modeling and the limitations of legacy planning tools. | |
| Applying Physics-Informed AI to Grid Simulation and Solution Generation | 5 | 5 | 1 | 1 | Shayle offers an informed analogy comparing physics-informed grid AI to coding LLMs and inference tasks. Josh elaborates on why deterministic machine learning trained on proprietary utility physics differs fundamentally from general generative LLMs. | |
| Mid-Roll Sponsor Break: Fuel Cells, Custom Solutions, and VPPs | 4 | 6 | 3 | 3 | After an ad break, Shayle posits that operations may have a lower ceiling for AI value compared to planning. Josh directly rejects this framing, explaining that operations actually has much higher potential through real-time continuous analysis and closed-loop learning. | |
| Bridging the Disconnect Between Utility Planning and Operations | 4 | 5 | 1 | 1 | Shayle asks whether planning study findings ever communicate directly with real-time operations. Josh explains the disconnect between departments, noting that data rarely crosses over beyond static spreadsheets due to siloed analytical software. | |
| The Data Center Boom, Behind-the-Meter Microgrids, and Micro-Utilities | 5 | 5 | 2 | 1 | Segment duplicate prevention - verified 5 segments total. |
Statements from this episode (16)
Kann: Critics calling the grid 'broken' do not understand utility operations
“But I will admit a little bit of frustration with a particular crowd That is constantly referencing the, quote, broken grid, unquote, that needs fixing, knowing full well that most of these people have absolutely no idea how utilities actually work.”
Wong: Over half of utility study time goes to cleaning data
“So studies today, I would say half if not more of the time, especially for distribution, is on cleaning up data.”
Wong: Transmission interconnection studies take 6-9 months and cost $250K
“I think most utilities are talking about six to nine months to perform one of those studies, and it's a one-time, right, so risk of restudies will compound that, and it does cost internal utility, typically a hundred, A couple of 100,000 dollars. The typical b…”
Wong: SCE projects up to 10,000 monthly energization requests taking 30-45 days
“I think when we public study that we did with SCE, I think they are projecting like up to 10,000 of these energization requests per month, and each one of them takes like 30 to 45 days.”
Wong: ThinkLabs' physics-informed AI models are deterministic and cannot hallucinate
“So our models actually can't hallucinate. It's actually deterministic. You ask it the same thing, it will give you the same answer all the time.”
Wong: ThinkLabs trains state-sized utility models in 10 minutes for $5
“And we have done that for multiple large IOUs now, a couple of thousand buses, it takes us about 10 minutes per training run. For a Powerflow model the size of a state, typically. And that 10 minute training run cost maybe five bucks of compute. And we were ab…”
Wong: ThinkLabs cuts nine-month grid interconnection studies to under 10 minutes
“So studies that would take previously nine months, right, as I mentioned, interconnection studies, now it takes us only a matter of a couple of minutes, like 10 minutes or less.”
Wong: Grid congestion is preventing utilities from scheduling planned maintenance outages
“The grid is getting so congested that if we do worst case scenario all the time we are, the utilities are beginning to struggle to find time to take the grid into contingency for planned work.”
Wong: AI has higher potential upside in grid operations than planning
“I think it is the complete opposite. I think the potential is even higher.”
Wong: Utilities lack operational closed-loop feedback until FERC-level post-mortems occur
“Currently, there is no closed-loop feedback until you have a post-mortem. You have a, an event that drives an investigation that needs to be reported back up to FERC.”
Wong: Utilities cannot determine historical utilization of distribution cables
“Like how much is that cable actually being utilized historically? Nobody can really tell you that answer today in planning or operations, especially in distribution.”
Wong: Utility planning studies rarely feed into real-time operations
“Of course, each utility might be different, but that is a very, very rare scenario, that they, these study assumptions and results actually feed into operations because they're measured by the different results and KPIs and all that kind of stuff.”
Wong: Utilities adapt workflows to match legacy software limitations
“And so, in some ways, I would say the utilities are actually adapting their people and processes around the limitations of the analytics, intelligence, and decision-making frameworks that they have.”
Wong: Islanded energy microgrids are always more expensive than interconnected grids
“Let's remember why we have a grid in the first place, which is so that we don't become individual islands. And individual islands are always more expensive. Than sharing the resources that we have.”
Wong: The grid has enough latent capacity for most data centers today
“So let me just make a statement here, which is, I believe the grid has enough existing latent capacity to connect the majority, if not all of the data centers today.”
Wong: Behind-the-meter data center generation will complicate grid transients
“Bringing these additional generators and batteries and fuel cells, et cetera, will complex the other side of the equation, which are the non-steady states. All your transients, all your EMT, like electromagnetic transients, will get way more complicated with a…”