Utility Rate Case Preparation Agents: Building the Filing

The Evidence Problem at the Heart of Every Rate Case

Rate case proceedings sit at the intersection of two disciplines that rarely cooperate well: engineering operations and administrative law. The utility must translate decades of capital investment, depreciation schedules, load forecasting data, and cost allocation studies into a document package that satisfies a commission's evidentiary standards — all within a compressed filing window. The gap between what the operations team knows and what the legal record must prove is where most rate cases lose time, and sometimes lose ground with regulators.

The problem compounds because modern utility operations generate data at volumes that traditional analyst workflows cannot fully process. A mid-sized electric distribution utility may track millions of meter reads, hundreds of capital project records, and thousands of customer complaint logs that are all potentially relevant to a rate case. The staff capacity required to synthesize that material into a coherent, citation-backed filing exceeds what most regulated entities can maintain in-house.

Autonomous agents are beginning to close that gap in a meaningful way.

What a Rate Case Filing Actually Requires

Before understanding how automation changes the process, it helps to be precise about what a rate case filing contains. At minimum, a complete filing before a state public utility commission includes a cost-of-service study, a rate design proposal, a revenue requirement analysis, prefiled testimony from company witnesses, supporting schedules, and responses to any standard data requests the commission has published. Larger cases may also include depreciation studies, infrastructure condition assessments, affordability analyses, and environmental compliance cost allocations.

Each component requires a different type of underlying evidence. The depreciation study pulls from asset management records. The revenue requirement analysis depends on test-year financial data that must be normalized for extraordinary items. The prefiled testimony synthesizes engineering judgment into plain language that a non-technical commissioner can evaluate. None of these can simply be exported from an existing system — they require interpretation, normalization, and cross-referencing before they are filing-ready.

The regulatory staff on the other side of the proceeding will scrutinize every figure. Inconsistencies between a schedule and the testimony that references it, or between this year's filing and the prior case record, invite data requests that can stall a proceeding by weeks. Filing accuracy is therefore not a courtesy; it is a strategic necessity.

How Do Agents Assist Utility Rate Case Preparation and Regulatory Filings?

The core question — how do agents assist utility rate case preparation and regulatory filings? — has a precise operational answer. Agents operate as persistent, task-specific processes that can be assigned to discrete subtasks within the filing workflow and then run continuously until those subtasks are complete, flagging exceptions for human review rather than requiring human initiation at every step.

A data normalization agent, for instance, connects to the utility's financial system, extracts test-year general ledger entries, applies commission-approved normalizing adjustments for weather, outages, and one-time items, and produces a normalized revenue requirement schedule that is reconciled against prior periods. It does not do this once and stop — it monitors the source data for subsequent corrections and updates the downstream schedules automatically. A human analyst would need to restart this process every time a prior-month journal entry is revised; the agent handles that revision propagation without manual prompting.

Similarly, a testimony drafting agent can ingest the normalized schedules, the engineering project summaries, and the prior case testimony from the same witness, then produce a structured draft that maintains internal consistency across all references to dollar figures and project descriptions. The draft still requires a sponsoring witness to review and certify it — no regulatory body accepts machine-generated testimony without a human sponsor — but the drafting agent compresses what is typically a multi-week manual process into days.

The Architecture of an Agent-Assisted Filing Workflow

Designing an effective agent architecture for rate case work requires mapping the filing's dependency chain before any code is written. Most rate case components have upstream dependencies: the cost-of-service study cannot be finalized until the revenue requirement is stable, and the rate design cannot be proposed until the cost-of-service allocation is complete. An agent architecture that does not respect this dependency chain will produce components that are internally inconsistent when assembled.

A well-designed workflow organizes agents into layers. The first layer handles data acquisition and normalization — pulling from financial systems, asset management platforms, and operational databases, then running standardized quality checks against prior filings and commission orders. These agents run continuously and produce a validated data layer that all downstream agents consume. The second layer handles analysis — depreciation calculations, cost allocation modeling, load forecasting reconciliation — and produces structured outputs in formats the testimony agents and schedule-formatting agents can read directly.

The third layer handles document production: testimony drafts, schedule population, exhibit numbering, and cross-reference validation. A cross-reference validation agent is particularly valuable because it reads the assembled testimony and compares every dollar figure and schedule reference against the actual schedule output, flagging any instance where a number in the testimony does not match the corresponding schedule. This type of internal consistency check is laborious to do manually and is frequently the source of deficiency notices from regulatory staff.

The architecture also requires an exception-handling layer. When source data is missing, a normalizing adjustment cannot be calculated, or a cross-reference fails to resolve, the filing cannot simply proceed with an error. The exception layer routes unresolved items to the appropriate human reviewer with enough context — the specific data point, the downstream dependencies, and the deadline impact — that the reviewer can make a decision quickly rather than spending time reconstructing what went wrong.

Regulatory Data Requests and Agent-Driven Response Management

One of the most resource-intensive phases of any rate case is the data request response period. Opposing parties — consumer advocates, industrial customer groups, or commission staff — submit formal interrogatories that must be answered on defined timelines, often within ten to fifteen business days. A contested case may generate hundreds of data requests across multiple rounds. Managing that volume while also continuing to prepare for hearings strains even well-staffed regulatory departments.

Agents address this challenge by maintaining a structured index of all evidence in the filing and mapping each data request to the indexed evidence it touches. When a data request arrives, a classification agent reads the question, identifies the relevant schedules and testimony sections, and drafts a response that cites the appropriate filing exhibit and page. The draft goes to a human reviewer who verifies accuracy and adds judgment that the agent cannot supply — for instance, whether a particular question is objectionable under the commission's procedural rules, or whether a sensitive cost item requires a confidentiality designation.

The indexing function is equally important outside of formal data requests. During hearing preparation, attorneys and witnesses need to locate specific figures across a large record quickly. An agent-maintained evidence index that links every number in every schedule to its source data, its normalization methodology, and its testimony citation gives the litigation team a searchable map of the entire filing. That capability reduces hearing preparation time substantially and makes cross-examination responses more reliable because witnesses can verify figures against indexed sources rather than memory.

Cost Allocation Modeling and Multi-Jurisdictional Complexity

For utilities with operations across multiple jurisdictions — common among gas pipelines, regional transmission organizations, and multi-state electric distribution companies — cost allocation is one of the most contested elements of any rate case. The question of which costs are properly allocated to retail versus wholesale customers, or to one state versus another, is rarely settled by formula alone. It involves regulatory precedent, negotiated allocation factors, and commission-specific interpretations of cost causation principles.

Agents can maintain a living allocation model that incorporates all applicable commission orders, prior settlement agreements, and pending allocation factor updates. When a capital project is added to the utility's asset base, the allocation agent can automatically apply the relevant allocation factors based on project type and service territory, flag any project whose cost characteristics do not fit neatly within the established allocation methodology, and route those exceptions for regulatory counsel review before the filing is assembled.

This continuous modeling approach is more reliable than the traditional practice of building allocation models from scratch for each case. When prior case allocation factors are embedded in the agent's reference data, consistency with prior filings is automatic rather than dependent on an analyst remembering to check the prior case record. That consistency matters because regulatory commissions give weight to prior treatment, and departures from prior methodology require explicit justification.

Depreciation Studies and the Role of Persistent Monitoring Agents

Depreciation studies are required in most jurisdictions for cases involving significant plant additions or when a utility has not updated its depreciation rates in several years. The study requires actuarial analysis of asset service life, net salvage expectations, and the appropriate depreciation method for each asset category. This work is typically done by specialized consultants and takes months. Agents do not replace the actuarial judgment at the center of a depreciation study, but they substantially compress the data preparation that precedes it.

A depreciation data agent connects to the utility's asset management system and produces a clean, audited extract of plant in service by account, vintage year, and service territory. It flags assets that appear in the financial records but lack corresponding engineering records, identifies account classifications that differ from prior study assumptions, and reconciles the plant extract against the prior depreciation study's beginning balance. That preparation work, done manually, can consume several weeks of analyst time before the consultant can begin the actual analysis. An agent collapses it to days.

The monitoring function extends beyond the filing period. A persistent depreciation monitoring agent can track asset retirements and additions against study assumptions throughout the rate cycle, alerting the regulatory team when actual retirement patterns deviate significantly from study projections. That signal can trigger a decision about whether to seek an interim rate adjustment or to document the deviation for the next rate case — a proactive posture that is difficult to maintain without automated monitoring.

Load Forecasting Reconciliation and Rate Design Evidence

Rate design — the structure of how rates are set across customer classes — must be grounded in load forecasting data. The forecast supports the utility's claim about how demand will change over the rate effective period and justifies any changes to the fixed-variable split in customer charges. Commissioners and intervenors scrutinize load forecasts carefully because they directly affect both revenue recovery and the distribution of costs across customer groups.

Agents assist by maintaining a continuously reconciled load forecast that draws from the utility's metering data, its econometric model outputs, and external reference forecasts from the regional grid operator or state energy office. When the external reference forecasts are updated — typically quarterly — the reconciliation agent re-runs the comparison and reports on whether the utility's internal forecast remains within acceptable variance bands. If variance exceeds a threshold set by the regulatory team, the agent generates an alert and a summary of which forecast assumptions are driving the divergence.

For the filing itself, the rate design agent takes the reconciled load forecast and applies the proposed rate structure to each customer class, producing projected revenue recovery under each rate design option. Those projections feed directly into the revenue requirement comparison that the filing must include. The agent also checks that the modeled customer bills under the proposed rates are consistent with the bill impact analysis that the filing's customer affairs section must contain — a cross-document consistency check that is easy to miss when multiple teams are building sections independently.

Prefiled Testimony Management and Witness Coordination

Prefiled testimony is the direct evidence through which a utility makes its case before a commission. Each witness sponsors specific schedules, and the testimony must be internally consistent — the engineer's testimony on capital project costs must match the financial witness's revenue requirement schedules, which must match the accountant's cost-of-service testimony. When these sections are prepared by different teams working on parallel tracks, inconsistencies accumulate and must be caught in a final review that is often compressed against filing deadlines.

An agent-driven testimony management workflow assigns each testimony draft to a version-controlled document environment where every cross-reference is tagged to the schedule or prior testimony section it references. When a schedule number changes during the filing preparation — a common occurrence as the cost-of-service model is revised — the testimony management agent identifies every testimony citation that references that schedule and flags it for update. This eliminates a class of deficiency notices that stem purely from clerical inconsistency rather than substantive disagreement.

Witness coordination agents also track the signature and certification workflow for prefiled testimony. Most commissions require that testimony be signed under oath or contain a sworn certification. Managing that process across multiple witnesses, some of whom may be in different locations, requires tracking which drafts have been reviewed, which have been certified, and which are pending revision. An agent-managed certification tracker reduces the risk that a filing deadline is missed because a witness's certification was overlooked in the final assembly process.

Hearing Preparation and Real-Time Record Navigation

The evidentiary hearing phase of a rate case places different demands on the data infrastructure than the filing phase does. During cross-examination, a regulatory attorney needs to locate specific figures in the record within seconds, not minutes. A witness needs to answer questions about the source of a specific number on a specific schedule without hesitation. These demands require a searchable, indexed record that is organized around the questions an opposing party is likely to ask, not just around the filing's own organizational structure.

Post-filing, agents shift into a record navigation and hearing preparation role. A hearing preparation agent can generate witness preparation summaries that map the most likely cross-examination topics — derived from the intervenors' data requests and any pre-hearing briefs — to the specific exhibits, schedules, and testimony sections that the witness will need to reference. This is not a simple search function; it requires the agent to understand the relationship between the questions and the evidence, and to surface the most relevant exhibits rather than returning every document that contains a search term.

Real-time record navigation during hearings is a newer application, where an agent monitors the transcript feed and automatically updates a hearing room display with the relevant exhibit or schedule when a witness references a specific figure. This reduces the time consumed by exhibit retrieval during testimony and allows the proceeding to move more efficiently — a practical benefit that commissions increasingly appreciate given the volume of complex cases on their dockets.

Compliance Monitoring After the Rate Case Closes

Rate cases do not end when the commission issues its order. The order typically includes conditions: compliance filings to be submitted within specified timeframes, reporting requirements for new programs, and tracking mechanisms for capital projects whose costs were included in the approved revenue requirement. Managing compliance with a complex commission order requires the same kind of persistent monitoring that agent infrastructure provides during the filing process.

A compliance monitoring agent reads the final order, extracts all compliance requirements with their deadlines, and creates a structured compliance calendar. It monitors the utility's internal systems for the data that will be needed to satisfy each requirement and generates draft compliance filings when the required data becomes available. When a compliance deadline is approaching and the required data has not yet been received from an internal department, the agent sends a structured reminder that specifies exactly what is needed, by when, and what the consequence of a missed deadline is under the commission's procedural rules.

This post-order compliance function is where many utilities experience regulatory friction that damages their credibility with commissions over time. Missed compliance deadlines attract scrutiny in subsequent rate cases. An automated compliance monitoring capability addresses that risk directly, and the same infrastructure that supported the rate case filing can be extended to support the post-order period without additional system architecture.

Building Production Infrastructure Versus Running Point Projects

The distinction between production infrastructure and point projects is nowhere more consequential than in rate case work. A one-time automation engagement that is scoped to a single filing and then wound down produces a temporary efficiency gain. The data normalization logic, the cross-reference validation rules, and the commission-specific configuration all expire when the project closes. The next rate case starts from the same baseline, and the same build cost is incurred again.

Production infrastructure compounds. The data normalization agent that processed the test-year financials for one case carries that normalization logic — and the exception history it accumulated — into the next case. The cross-reference validator that was configured against one commission's procedural requirements retains that configuration and applies it to the next filing without re-scoping. The compliance monitor that tracked post-order requirements for one docket transfers its framework to the next order without manual reconstruction. Each successive rate case benefits from everything the infrastructure learned in prior cycles.

TFSF Ventures FZ LLC approaches agent deployment in regulated industries — including energy utilities — as production infrastructure built to run continuously across rate cycles, not as a consulting engagement or a platform subscription. Every line of code produced is owned by the client at completion. There is no ongoing license dependency on any vendor for the infrastructure to operate. The rate case agents that handle data normalization, testimony cross-referencing, and post-order compliance run on systems the utility controls, which means the operational value carries forward into every subsequent filing without additional build cost.

For organizations evaluating whether TFSF Ventures FZ LLC pricing is within reach, deployments in the regulatory affairs vertical start in the low tens of thousands for focused builds, scaling by agent count, integration complexity, and operational scope. The Pulse AI operational layer runs as a pass-through at cost based on agent count, with no markup applied. Every line of code produced in the deployment is owned by the client at completion — there is no ongoing license dependency on TFSF for the infrastructure to operate.

Questions about whether TFSF Ventures is legitimate or whether documented deployments support the production claims are answered by the same verifiable fact: TFSF Ventures FZ LLC operates under RAKEZ License 47013955, with a 30-day deployment methodology across 21 verticals including energy operations. The company was founded by Steven J. Foster with 27 years in payments and software, and every deployment produces owned infrastructure rather than a subscription to a third-party platform.

Integrating Agent Infrastructure With Existing Regulatory Systems

Most utilities have existing document management systems, financial reporting platforms, and commission filing portals that are not designed with agent interoperability in mind. The integration question — how agent infrastructure connects to these systems without requiring the utility to replace them — is often what determines whether an automation initiative produces real operational change or remains a pilot.

Effective integration relies on agents that operate at the API and data export layer of existing systems rather than requiring those systems to be rebuilt. An agent that pulls normalized financial data from an existing ERP via scheduled export, processes it, and pushes outputs to a document management system that the regulatory team already uses produces immediate operational value without disrupting established workflows. The utility's staff continues working in familiar systems; the agents handle the transformation and cross-referencing work that previously fell to analysts.

TFSF Ventures FZ LLC's 19-question Operational Intelligence Assessment is specifically designed to map this integration landscape before any deployment begins. The assessment identifies which existing systems hold the data that rate case agents will need, where the data quality gaps are, and which integration points require custom handling versus standard connectors. That scoping precision is what makes a 30-day deployment realistic — the build phase starts with a clear integration map rather than discovering data quality issues mid-deployment.

The Regulatory Credibility Dividend

There is a dimension of agent-assisted rate case preparation that does not appear in any cost-benefit model but matters substantially to regulated utilities: commission credibility. A utility that files consistently clean evidence, responds to data requests on time and with precise citations, and meets all post-order compliance deadlines builds a record of regulatory professionalism that influences how commissioners and their staff approach the utility's subsequent filings. That credibility is difficult to quantify but is recognized by any experienced regulatory attorney as a real factor in proceeding outcomes.

Agent infrastructure contributes to that credibility by eliminating the inconsistencies and lateness that damage it. When cross-references are validated automatically, when data requests are indexed and responded to within deadline, and when compliance filings arrive before the due date, the commission's staff develops a different relationship with the utility's filings than they have with parties whose records require extensive correction. The operational discipline that agent infrastructure enforces is indistinguishable, from the commission's perspective, from the organizational discipline that regulatory credibility requires.

That alignment — between what good agent infrastructure produces and what good regulatory practice demands — is why energy operations is among the 21 verticals in which TFSF Ventures FZ LLC deploys its proprietary Pulse engine and 30-day deployment methodology. The nature of regulatory work rewards the kind of persistent, precise, exception-aware processing that autonomous agents are specifically designed to deliver, and the infrastructure-first deployment model means that advantage accumulates across every rate cycle rather than expiring when a project closes.

About TFSF Ventures FZ LLC

TFSF Ventures FZ-LLC (RAKEZ License 47013955) is an AI-native agent deployment firm built on three pillars, all running on its proprietary Pulse engine: autonomous AI agents deployed directly into the systems a business already runs, a patent-pending Agentic Payment Protocol licensed to enterprises and payment networks globally, and a Venture Engine that compresses the full venture lifecycle from idea to investor-ready. Founded by Steven J. Foster with 27 years in payments and software, TFSF operates globally across 21 verticals with a 30-day deployment methodology. Learn more at https://tfsfventures.com

Take the Free Operational Intelligence Assessment

Run the Operational Intelligence Diagnostic — 19 questions benchmarked against HBR and BLS data. Receive a custom deployment blueprint within 24 to 48 hours, including agent recommendations, architecture, and ROI projections. Start at https://tfsfventures.com/assessment

Originally published at https://www.tfsfventures.com/blog/utility-rate-case-preparation-agents-building-the-filing

Written by TFSF Ventures Research