Utilities need to adapt in real time. AI holds the key

This article is the fifth in a series about how energy providers can thrive in an AI-powered future. 

Many of the planning processes utilities rely on today were designed for an operating environment defined by gradual change, stable assumptions, and long feedback loops. In that environment, integrated resource plans, transmission expansion studies, distribution forecasts, capital plans, and regulatory filings worked just fine as discrete, multiyear exercises, updated periodically and governed by fixed timelines. 

That reality is a thing of the past. Disruption across the energy industry is accelerating. Load growth no longer arrives in neat, predictable increments. Distributed energy resources (DERs), demand response mechanisms, and technological advancements are scaling faster than historical adoption curves would predict. Climate-driven risks are evolving year to year rather than decade to decade. And regulatory and affordability expectations continue to expand in scope and complexity. 

The result is a growing mismatch between how fast the power system is evolving and how quickly planning decisions are made. 

Dynamic planning enabled by AI is emerging as a critical capability for closing this gap. Instead of replacing established planning constructs overnight, AI allows utilities to continuously ingest new signals, test scenarios, and adjust decisions as conditions evolve. Many utilities have already been investing in the underlying capabilities, such as advanced analytics, advanced metering infrastructure (AMI), grid sensors, and digital platforms. Generative and agentic AI can amplify those investments so that utilities can operate at the speed of change. 

To visualize this transformation, we’ve created three hypothetical scenarios that contrast a traditional utility (which we’ll call TradElectric) with an AI-native utility (called SentientElectric). Each scenario centers on how these two imaginary utilities would tackle one of three core challenges: load growth, DER adoption, and capital planning. Each scenario is accompanied by a set of concrete steps that real-world utilities can take right now to advance toward the AI-enabled capabilities each scenario describes. 

Scenario 1: Neighboring utilities compete for the same load 

TradElectric and SentientElectric serve neighboring territories under the same commission oversight. A hyperscale data center developer approaches both with a phased load plan that could reach several gigawatts—demand that would materially change load flows, transmission upgrades, and generation needs. 

At TradElectric, planners quickly run into regulatory constraints. The latest load forecast and system plans were filed 12 months ago and approved on older assumptions. Serving this new large-load customer would require new interconnection studies, accelerated procurement, and likely a plan amendment—each needing a defensible record that investments are prudent, just, and reasonable for all customers. 

In parallel, the commission’s usual questions loom: cost allocation (who pays), risk of stranded assets, and whether the utility evaluated all potential measures to optimally utilize existing grid assets, including non-wires alternatives. TradElectric can engage, but its timeline is constrained by the next filing cycle and the time it takes to respond to discovery and stakeholder challenges. The data center needs to break ground, finish construction, energize its systems, and achieve commercial operation in about 18 months. 

SentientElectric faces the same mandates but reduces regulatory uncertainty upfront. Its AI-enabled planning environment keeps a live mapping between emerging loads and commission requirements, tracking: 

• What can be done under existing approvals and tariffs 
• What the utility can anticipate and start working on now 
• What requires commission action 

Before making commitments, SentientElectric schedules a prefiling conference with commission staff to preview new load scenarios, proposed study scope, and a cost-recovery path. It comes prepared with a concise “commission packet,” including updated forecast deltas, system study results, options analysis showing least-cost/least-regrets sequencing, and a clear proposal for customer contribution or a tracker/rider with reporting conditions. When intervenors issue data requests, SentientElectric can produce an audit trail—including assumptions, model runs, and sensitivity cases—showing why the approach is reasonable. 

The data center developer chooses SentientElectric’s territory because the regulatory path can reduce its development time, de-risk the project from a technical and investment standpoint, and ensure power supply by offering an executable sequence—from study to filing to conditions to reporting—all with fewer “unknown unknowns.” 

AI-enabled planning actions utilities can take right now to attract and retain new load: 

1. Map regulatory constraints to planning actions in real time. Leverage AI to maintain a live view of what fits within existing approvals versus what triggers a filing or amendment. Tag requirements and link them to the data and model outputs used (assumptions, runs, exhibits) so that the rationale is traceable. 
2. Manage long-lead delivery risks. Use AI to flag procurement, transmission hardware availability, and construction bottlenecks by continuously ingesting supply chain, inventory, and project schedule updates and surfacing the highest-probability schedule drivers. 
3. Continuously refresh load changes for regulatory engagement. AI can rerun forecast sensitivities as customer plans evolve by automatically reconciling new customer-provided load profiles with forecast inputs and re-running scenario sets when key inputs change—especially if the power requirements from large-load clients will be lower today than in the future. 
4. Use a focused digital twin for large-load decisions. Working with the digital twin to model candidate substations and feeders, AI can test load phasing and upgrade sequencing and contingencies.  
5. Stress-test regulatory options inside the twin. Compare study scopes, upgrade paths, and customer contribution structures by having AI rapidly generate and evaluate comparable “option bundles” off the same validated inputs. 
6. Quantify “do-nothing” risk alongside build options. Have AI explicitly model reliability, avoided cost, and customer impacts by running counterfactual scenarios using the same input pipeline (forecast, constraints, program assumptions) and producing consistent outputs suitable for filings.  

Scenario 2: DER adoption surges 

In a fast-growing suburban corridor, rooftop solar, batteries, and EV charging trigger a sudden wave of interconnection applications clustering on the same feeders. Under commission-approved interconnection rules, both utilities must apply consistent screening, timelines, and technical requirements—and justify any departures.  

TradElectric follows an established cadence. Hosting capacity is evaluated after applications arrive, studies are run in batches, and mitigation proceeds through the queue. It’s orderly but episodic, and the queue grows faster than studies close. Customers receive conditional approvals and long timelines. Regulators see periodic updates, but by the time information is filed, conditions on the grid have already shifted—inviting complaints that the utility is slow, inconsistent, and overly conservative. 

SentientElectric uses AI to stay within the same commission rules while making outcomes faster and more defensible. AMI, inverter telemetry, queue data, weather, and real-time grid states feed a continuously updated view of feeder constraints. When a circuit becomes saturated, SentientElectric prepares a short filing or compliance report showing: 

• The technical basis for constraints (voltage, protection, thermal) 
• Evaluated mitigation options 
• Why the chosen approach is reasonable and consistent with tariffs and safety standards 

For example, if it proposes targeted controls or operating envelopes, it documents the thresholds, measurement/verification, and customer communications plan. When commission staff ask, “Are you treating applicants fairly?” SentientElectric can show transparent screening logic and a data-backed record of decisions.  

As a result, DER growth is manageable: faster interconnection where feasible, clearer timelines where upgrades are required, and fewer disputes escalating into commission complaints. 

AI-enabled planning actions utilities can take right now to manage DER growth: 

1. Move from static hosting capacity to continuously refreshed constraints. Rather than relying on periodic studies, have AI update feeder limits using AMI, inverter telemetry, and weather data by streaming these signals into a curated dataset with automated data-quality checks before updating constraint calculations. 
2. Tie AI outputs directly to approved interconnection pathways. Ensure that insights map cleanly to fast-track, supplemental review, and study screens already in tariffs by implementing AI as an assistant that applies the same tariff logic consistently while retaining the inputs/outputs used for each decision for auditability. 
3. Use an operational digital twin at the feeder level. Reflect real-time loading, inverter behavior, and protection limits while still producing filing-ready outputs by fusing real-time grid states with validated planning models in a common data model so that the twin stays aligned with both operations and filings. 
4. Identify where controls defer capital upgrades. Use AI for scenario analysis to determine where operating envelopes or targeted controls materially delay reinforcement by simulating “controls vs. wires” cases using telemetry-informed constraints and documenting the assumptions and performance thresholds. 
5. Standardize constraint explanations. Generate regulator-ready narratives explaining the constraint, mitigation selected, and basis for consistency across applicants by using AI to auto-populate explanations from the underlying data (constraint type, feeder state, options evaluated) so that writeups are consistent and reproducible. 
6. Pilot advanced methods in a defined sandbox. Have AI test new screening or control strategies on selected feeders before scaling systemwide using a parallel environment that mirrors production data flows (including telemetry transfer) so that results generalize and governance is established early. 

Scenario 3: Capital plans confront a changing resilience landscape 

TradElectric builds a multiyear capital plan for grid hardening and wildfire mitigation that’s filed and approved through a formal commission process emphasizing justification, affordability, and planned outcomes. Once approved, the plan provides certainty—but conditions change faster than the regulatory cycle. Successive summers of extreme weather shift wildfire risk materially, with some circuits seeing repeated stress events and others performing better than expected. Field teams know that priorities should change, but changes trigger a familiar burden—new analyses, revised testimony, stakeholder review, and commission approval—with risk often shifting before they’re completed. 

SentientElectric treats regulatory commitments as binding guardrails while using AI to manage change inside them. It operationalizes the commission’s objectives (risk reduction, reliability, affordability) as measurable performance metrics tied to approved programs. When risk shifts, the utility doesn’t “rewrite the plan” informally; it prepares a commission-ready justification showing that resequencing projects still meet approved goals and remain prudent. It can file a targeted update or provide required periodic reporting that: 

• Quantifies the change in risk drivers 
• Shows alternative project sequences and expected outcomes 
• Explains why the revised sequence is the least-cost option while maintaining safety and reliability obligations 

If the commission requires it, SentientElectric proposes guardrails—such as quarterly reporting, a limited amendment docket, or thresholds that trigger notification—so that the commission retains oversight without forcing a full re-litigation of the entire capital plan. 

AI-enabled planning actions utilities can take right now to develop a more agile capital planning process: 

1. Translate commission objectives into measurable metrics. Encode approved reliability, safety, and affordability goals into quantitative indicators by defining standardized metric definitions and storing them alongside the data sources and calculation logic used for reporting. 
2. Use AI to continuously rescore projects as risk shifts. Update wildfire, storm, and asset-health inputs to reassess project sequencing—not just scope—by ingesting updated risk signals into a governed data pipeline and re-running scoring when material inputs change.  
3. Define resequencing thresholds upfront. Establish clear triggers for notification, reporting, and limited filings, and have AI monitor those thresholds against incoming risk and condition updates to flag when governance steps are required. 
4. Anchor capital decisions in a risk-aware digital twin. Use a planning-grade twin that links asset condition, climate exposure, and outage performance to re-rank projects within approved programs, with inputs standardized across IT/OT systems so that the twin is fed by consistent, validated data rather than one-off extracts. 
5. Use the twin to justify resequencing—not scope creep. Show regulators that changes in project order preserve approved outcomes on safety, reliability, and affordability by using AI to generate an auditable comparison of “original vs. revised” sequences using the same metrics, data lineage, and assumptions. 

Across all three scenarios, AI only performs as well as a utility’s ability to collect, maintain, validate, and securely move planning and operational data to keep models current. This entails robust data management and low-latency networking, which most utilities have or are now putting in place. 

Dynamic planning can start now 

AI-enabled dynamic planning offers a way to move beyond static assumptions and periodic updates toward continuously informed decision-making. Utilities don’t need to predict the future perfectly to plan effectively, but they do need planning processes that can adapt quickly as the future unfolds.  

Moving from static deterministic planning to AI-enabled dynamic planning doesn’t require dismantling existing processes. In practice, the most effective starting point is to layer dynamic capabilities into formal planning and study cycles, using those capabilities to improve speed, insight, and defensibility without disrupting approved timelines. 

By starting with real system studies, narrowly scoped use cases, and regulator-credible governance, utilities can begin moving toward dynamic planning today—without waiting for regulatory frameworks to change. 

Explore the series