A rate case is not just a regulatory proceeding. For a utility building an AI transformation program, it is the capital allocation event that determines whether the program gets funded, at what pace, and with what accountability structure. The transformation leader who understands the rate-case machinery can fund a multi-year AI capability build inside a regulated capital plan. The one who does not will find that the best-designed program in the industry stalls for want of a defensible cost recovery path.

How Regulated Capital Actually Works for AI Investment

The rate-case machinery is built on a simple premise that becomes complicated in practice: the utility is entitled to recover its prudent and reasonable costs from ratepayers, including a return on its invested capital. AI investment fits into this framework through three channels, and understanding which channel applies to each component of the investment determines how it is presented to the commission and how quickly it is recovered.

The first channel is operating expense (O&M). Software subscriptions, data science consulting fees, program management staff, training, and governance committee overhead are all operating expenses under standard utility accounting. They are recovered dollar-for-dollar in the test-year revenue requirement of the next rate case, without any capital proceeding or separate approval. This is the fastest recovery channel and the lowest regulatory risk. For phase-one AI programs, operating expense is almost always the right starting point: you are buying access to a capability before you have proven it deserves a capital commitment.

The second channel is capital investment. Under ASC 350-40 (internal-use software accounting) and under most state commission capitalization policies, software development costs during the application development stage can be capitalized. Hardware infrastructure (servers, networking, storage) is capitalized per standard fixed-asset policy. Capitalized AI investment goes into rate base and earns a return over the asset's useful life. For a utility, the return on equity component of the weighted average cost of capital means the utility recovers more than the initial investment, which makes capital treatment advantageous over a long horizon but requires a separate regulatory process for approval. The capital proceeding is where prudency is tested most rigorously.

The third channel is avoided capital, which is not a recovery channel in the traditional sense but is the most powerful financial argument for AI investment. If AI-improved forecasting accuracy allows the utility to defer a specific transmission or distribution investment, the avoided capital cost reduces the revenue requirement relative to what it would have been without AI. This is a rate reduction story, not a cost recovery story, and it is the argument that makes AI investment most attractive to commissions focused on rate affordability.

In a regulated utility, the question is never just "does this AI system work?" The question is "can we demonstrate to a commission that this AI system's benefits justify its cost, and can we document that case precisely enough to withstand cross-examination?"

The Three-Rate-Case Progression for AI Investment

A well-designed multi-year AI investment program aligns its phases with the rate-case cycle. Rate cases typically occur every three to five years at most regulated utilities, though some utilities operate under formula rates or annual true-up mechanisms that change the calculus. Assuming a three-year rate-case cycle, the investment program has a natural three-case architecture.

First Rate Case: Operating Expense and Performance Evidence

In the first rate case that includes AI investment, the utility is presenting phase-one tools as operating expense. The key exhibit is a performance exhibit showing verified accuracy improvement: the MAPE improvement from baseline, the holdout test methodology, the production accuracy record, and the financial translation (what the accuracy improvement is worth in procurement cost savings or queue throughput improvement). The regulatory risk at this stage is the "imprudent expense" argument from intervenors: was the utility's spending on AI reasonable given what was known at the time of the investment? The defense is the business case document prepared before the investment, the competitive benchmarks that justified it, and the actual performance results that demonstrate it delivered.

The first rate case also establishes the data request response pattern: when commission staff asks for documentation of the AI program's governance, the model registry, the drift monitoring logs, and the CIP compliance assessments are all available. This demonstration of governance maturity is itself evidence of prudent management, which the prudency test requires.

Second Rate Case: Capital Investment and Avoided-Capital Benefit

The second rate case is where the capital investment ask is made. By the second rate case, the utility has two to four years of operating expense history with the AI tools, a verified performance track record, and a mature governance framework. The capital investment ask is for the platform infrastructure that takes the program from a collection of tools to a managed, enterprise-grade capability: a model serving platform, a data integration layer connecting EMS, ADMS, OMS, GIS, and market systems into a clean AI-ready feed, and potentially a model training infrastructure if the utility has decided to build rather than buy for certain use cases.

The avoided-capital argument is the centerpiece of the second-case benefit exhibit. This requires a credible utility-specific analysis: which specific transmission or distribution investments were contemplated in the capital plan approved in the prior rate case, which of those investments have been deferred due to AI-improved forecasting, and what is the net present value of that deferral to ratepayers. This calculation requires the planning team's involvement and the CFO's sign-off, because the commission will cross-examine it thoroughly. The transformation leader should be prepared to concede the portions of the benefit that are uncertain or depend on assumptions, and defend the portions that are well-supported. Overreaching on the benefit calculation is a credibility risk that is worse than a smaller, well-defended benefit claim.

Third Rate Case: Standard Operating Model Infrastructure

The third rate case is the destination: AI as standard operating model infrastructure. By this case, the utility can present a three-case track record (operating expense history, capital investment performance, avoided-capital realized benefit), a workforce model that is genuinely built around AI-augmented roles, and a governance framework that has survived at least one audit cycle. The argument is that AI investment is now as fundamental to the utility's operating model as the EMS or the OMS: it is not a technology experiment but a required component of reliable, affordable service in the current grid environment.

The third-case ask is typically for ongoing capital maintenance (refreshing the model training infrastructure as technology evolves), ongoing operating expense recovery for governance and model management, and possibly a new capital investment for extending the AI capability to a new operational domain (real-time topology optimization, for example, which was not yet supported by the evidence base in the first two cases). By the third case, the commission and its staff are familiar with the program, have reviewed its performance for years, and have a working understanding of how AI advisory systems operate. The adversarial dynamic of the first two cases is replaced by a more collaborative one, because the utility has built the regulatory relationship over time.

Managing the Test-Year Challenge

The test year is the period on which a rate case is based, and it creates a structural timing challenge for technology investments. The rate case typically reflects costs and revenues from a twelve-month period that ended six to twelve months before the case was filed. If the AI program's most significant costs and benefits fall outside the test year, they may not be fully captured in the case. The solution is a combination of known-and-measurable adjustments (where the commission allows projections of costs that are committed but not yet in the test year) and post-test-year evidence (where the commission considers the full production period in evaluating prudency, even if the cost recovery is limited to the test year).

The transformation leader should work with the regulatory affairs team to identify the test year for the upcoming rate case early, and to ensure that the AI program's cost and benefit recognition aligns with the test year's accounting. For capital investments, the timing of when assets are placed in service determines when they enter rate base; the transformation leader should be aware of the rate case test year when planning phase-two capital deployment milestones. A capital investment placed in service three months before the test year closes generates more rate-base value in the upcoming case than one placed in service after the test year closes.

Defending Prudency Under Cross-Examination

The prudency standard in rate regulation asks whether the utility's decision, at the time it was made, was consistent with the decision a reasonable, well-managed utility would have made under the same circumstances. For AI investment, the prudency defense rests on four elements: the decision-making record (what analysis was done before the investment was approved?), the governance framework (was the investment managed with appropriate oversight?), the performance result (did the investment deliver its intended benefits?), and the alternatives analysis (did the utility consider and reject less expensive alternatives?)

The decision-making record is built before the investment, not after. The transformation leader should ensure that every significant AI investment is documented with a business case that includes: the problem statement, the alternatives considered, the basis for the AI solution's selection, the financial projections with stated assumptions, the risk factors and mitigation, and the authorization level required for the investment. This document is the primary defense in a prudency challenge.

The performance result is the most powerful prudency defense when it is positive, but it is not a complete defense on its own. A commission that finds the decision-making process was inadequate may still disallow costs even if the investment performed well, on the theory that the utility got lucky rather than made a good decision. The converse is also true: a commission that finds a well-documented, well-governed investment that underperformed due to circumstances outside the utility's control may allow full recovery, because prudency is a process standard, not an outcome standard.

Formula Rates and Alternative Rate Structures

Not all utilities operate under traditional rate cases. Transmission utilities subject to FERC jurisdiction often operate under formula rates, where the revenue requirement is calculated annually based on a defined formula and the actual costs and rate base from the prior year. For these utilities, AI investment that qualifies as transmission plant goes into rate base automatically upon completion and in-service, with the annual formula rate true-up reflecting the new asset. This is a significantly faster recovery mechanism than the three-to-five-year rate-case cycle at many distribution utilities.

Distribution utilities in states with alternative rate mechanisms (performance-based rates, multi-year rate plans, revenue decoupling) have different investment recovery structures. A multi-year rate plan with a technology investment fund or capital investment rider can allow AI platform investment to be recovered more quickly than under a traditional rate case, with less regulatory risk, because the investment fund mechanism is designed for exactly this kind of technology-driven capital deployment. The transformation leader should understand which rate mechanisms are available in each jurisdiction and work with the regulatory affairs team to identify the optimal structure for the AI investment program.

Some states have developed specific regulatory mechanisms for utility technology investment: grid modernization plans, advanced grid programs, or technology investment mechanisms. These are worth examining carefully. Where they exist, they typically provide a defined approval process, a defined cost recovery path, and sometimes a defined performance metric that the utility must meet to retain the investment recovery. The transformation leader should be deeply familiar with any such mechanisms in the utility's operating jurisdictions.

A Worked Example: Constructing the Five-Year Capital Plan

Consider a large investor-owned utility planning its AI investment program over a five-year horizon spanning two rate cases. The transformation leader works with the planning team, the finance team, and the regulatory affairs team to construct the capital plan in three tiers.

Tier one is operating expense throughout all five years: the AI tool subscriptions (forecasting platform, queue study platform), the governance staff (AI systems manager, data governance lead), and the governance infrastructure (model registry platform, drift monitoring tooling). These are estimated at approximately $3 million annually and are recovered as O&M in each rate case test year. This tier requires no separate capital proceeding.

Tier two is the capital investment ask in the second rate case: the data integration layer ($8 million, capitalized over a ten-year useful life), the model serving platform ($5 million, five-year useful life), and the governance infrastructure enhancements ($2 million). Total capital ask: $15 million, generating approximately $1.5 million annually in rate base return at a 10 percent weighted average cost of capital. The second-case benefit exhibit shows verified avoided-capital of approximately $12 million in deferred distribution investment from the improved forecast, generating a positive rate impact for customers within the rate case period.

Tier three is the capital ask in the third rate case for the real-time topology optimization advisory system and the predictive maintenance infrastructure: approximately $18 million combined. By the third case, the utility has a two-case track record, a reliability performance record that includes the topology advisory system's shadow-mode validation, and avoided-capital calculations that are now supported by three years of actual investment deferral data rather than projections. The third-case capital ask is the most defensible because it is the most evidence-supported.

The five-year capital plan total is approximately $33 million in capital and $15 million in operating expense for AI-specific investments. The benefit calculation shows approximately $45 million in avoided distribution and transmission capital over the same period, a net ratepayer benefit of approximately $12 million, and a reliability improvement measured in reduced MAPE, improved queue throughput, and reduced outage duration. This is the summary a commission needs to understand the program. The exhibits behind it are what make it defensible.

The Great Crew Change and Investment Sizing

One of the most frequently overlooked variables in the AI investment sizing calculation is the workforce cost avoided by AI augmentation. With more than 25 percent of utility workers retirement-eligible within a few years, the analytical and planning capacity that those workers represent must be either replaced by new hires or augmented by AI. EPRI's projection of more than 30 percent growth in digital and analytical roles through 2030 is not a coincidence; it is the reflection of utilities that are deciding to hire more analytical staff rather than, or in addition to, deploying AI tools.

The transformation leader who can quantify the workforce investment avoidance creates a more complete financial case for AI than the one who focuses only on avoided capital. If the AI-assisted queue management program allows the utility to process three times the queue applications with the same engineering staff, the alternative cost of achieving the same throughput by hiring additional engineers is a valid component of the benefit calculation. Many commissions that are skeptical of technology investment are more sympathetic to an argument that frames AI as a workforce productivity tool that avoids rate increases from hiring costs than to one that frames it as a technology modernization investment.

The workforce investment avoidance calculation requires care. The comparison is not "what would we pay the new hires?" but "what would the utility's O&M cost trajectory be without AI augmentation, and how much lower is that trajectory with AI?" The difference, discounted at the appropriate rate, is the workforce cost avoided. This calculation must account for the training and transition costs associated with AI augmentation, which are not trivial but are significantly less than the ongoing cost of additional headcount at fully loaded rates.

Key Takeaways

• AI investment at a regulated utility flows through three channels: operating expense (fastest recovery, lowest risk, no capital proceeding required), capital investment (rate-base treatment with return, requires prudency documentation and approval), and avoided capital (the most persuasive financial argument for commissions focused on rate affordability).
• The three-rate-case architecture aligns the transformation phases with the rate-case cycle: operating expense and performance evidence in the first case, capital investment and avoided-capital benefit in the second, standard operating model infrastructure in the third.
• The prudency defense is a process standard, not an outcome standard: the decision-making record prepared before the investment is the primary defense, and it must document the problem statement, alternatives considered, financial projections with assumptions, and risk factors.
• Test-year timing affects cost and benefit recognition: the transformation leader should coordinate capital deployment milestones with the regulatory affairs team's test-year calendar to maximize rate-base value in the upcoming case.
• Formula rates (FERC transmission utilities) and alternative rate mechanisms (grid modernization plans, capital investment riders) can provide faster AI investment recovery than the traditional rate-case cycle, and should be evaluated for each operating jurisdiction.
• The avoided-capital calculation must be utility-specific, must identify the specific investments deferred, and must survive cross-examination on its assumptions; the transformation leader should concede uncertain portions rather than overreach and lose credibility on the well-supported portions.
• A well-constructed five-year capital plan integrates operating expense, capital investment, and avoided-capital across two rate cases, with the third case presenting AI as standard operating model infrastructure supported by a multi-year reliability and financial track record.