The extraordinary time, effort, and billions of dollars spent annually on data discovery should go toward what actually matters: improving the grid.
General-purpose AI tools, such as ChatGPT, Copilot, and Claude, are making the data request process for regulated energy proceedings more arduous and tedious. The increased volume and reduced quality of data requests are making it harder for rate case discovery to serve its valid purpose of informing the parties and helping the regulator make the best decision.
In response, utilities are seeking to modernize their side of the data request process, and are working to standardize discovery across multiple operating companies and regulatory jurisdictions.
How? In a recent roundtable with HData and seven large power utilities, participants described what their ideal solutions might look like, and what problems they need to address:
- “Everybody touches discovery. Everybody across a utility is a big part of answering the questions, even if you’re not assigned to it. There are other parts of the organization that have to provide the support to get the answers.”
- “The biggest efficiency gain we’ve gotten in the last two years is all getting on the same page with our rate case, our resource plan, and our certifications all following the same process.”
- “What we would want in a perfect world is to have an end-to-end solution where we would be able to just review the answers and have the human in the loop to approve the questions and move it to the next step.”
Participants shared they’re not seeking to just generate faster responses. They want to better manage the entire process: receive a request, identify the responsible experts, locate relevant information, prepare a response, coordinate reviews, secure approvals, and maintain an authoritative record of what is submitted. That requires way more than a chatbot. It requires workflow management, access to regulatory history, enterprise coordination, confidentiality protections, legal privilege preservation, and AI that understands the lingo and substance of utility regulation.
Regulatory Innovation: Piloting a Data Request Management System with Utilities
To address these issues, several of our roundtable participants are piloting HData’s Data Request Management application, which allows teams to run incoming requests across all their priors and their public docket context, generate proposed answers so subject matter experts aren’t faced with a blank page, redline every change, and manage the complex workflow of responding.
HData uses purpose-built AI which combines multiple large language models with specialized architecture battle-tested by thousands of regulatory professionals. Unlike general-purpose AI, HData’s system is fully grounded in the source content, including the public docket context and each customer’s history of previous proceedings, and can show the sources behind every answer, linking to the page and paragraph level.
Some utilities are also combining HData with internally-developed tools, while some are integrating shared drives and document management systems with HData.
HData Can Help Utilities Manage Data Requests More Efficiently, But That’s Not Enough
HData’s operating system contains a comprehensive collection of authoritative regulatory information (dockets, laws, testimony, IRPs, RTO materials), all pre-processed for citations. This means we can help professionals find the relevant precedents, understand the legal and strategic context behind requests, and develop responses that are grounded in both the public record and the proprietary history.
This is a huge improvement over asking a general-purpose AI tool to improvise an answer based on a few PDFs, and it can help utilities handle growing discovery workloads without having to proportionally grow their regulatory teams.
However, we know that other parties have general-purpose AI tools that instantly generate ten times as many questions as before, and the questions are lesser in substance and quality. Even if HData enables the utilities to answer ten times as fast, has anything actually improved in the overall process? Not really. At best, we’re at equilibrium.
In this scenario, perhaps AI has made all sides more prolific. Everyone is generating more documents, reviewing more documents, and spending more time managing the machinery of discovery. Meanwhile, the underlying information hasn’t improved, and the regulatory decisions haven’t necessarily gotten any better. That is not the future we should build.
The Real Answer is a Shared Platform
Energy regulation is a shared challenge. Utilities, regulators, and intervenors all have different responsibilities, different incentives, and fundamentally conflicting interests.
But they’re working on the same problems. They’re examining the same investments, evaluating the same costs, and participating in the same proceedings. Much of the information they need is the same, too. So why should each participant have to assemble the record independently, apply its own disconnected AI tools, and spend so much effort on data requests whose answers might already be in that record?
Let’s imagine a different approach.
Imagine that a regulatory commission, regulated utility, and intervenors all share the same digital platform. This platform contains all of the public record. Not just the docket we’re working in, but every docket, from every commission, across the United States and Canada.
When a utility files an application or a resource plan, that content goes into the platform, too. Confidential materials are routed to the appropriate parties who are entitled to see them. When commission staff or intervenors begin working on data requests against the utility’s submissions, they start by querying the available content, using specialized AI grounded in that content.
Only after the commission staff or intervenors have exhausted what’s answerable from the available content do they then issue their formal, litigation-style data requests.
The utility can object to requests whose answers are already evident from the content available to the requestor. The utility’s side of the platform automatically knows which information is available to which requestor, and can instantly determine which requests are objectionable on these grounds.
The utility uses the same platform to generate responses to legitimate requests. Since the platform has access to all the information the utility has provided in the past, it can automatically ensure new responses are consistent. When the utility responds to the legitimate requests, the new responses go right into the platform, and thus themselves become part of the queryable corpus, with access properly restricted for anything confidential.
Imagine what this could mean for all our parties in a rate case:
- The regulatory commission and the intervenor get the information they need, faster.
- The utility doesn’t need to spend time on answers to duplicative requests.
- Constructive requests are prioritized, and redundant ones are eliminated.
The participants in the rate case still won’t agree on the underlying disputes. Nor should they. Regulatory proceedings are supposed to accommodate competing interests, vigorous scrutiny, and disagreements about the public interest
Instead of using AI to generate more questions and more answers, in this approach, we are using AI to generate better questions and better answers, all informed by a shared record that doesn’t need to be reconstructed for every new request and response.
A shared platform is a fundamentally different proposition from scattered AI tools. We should be building a shared operating system for energy regulation, rather than separate productivity suites for each participant.
The HData operating system is already used by nearly 200 organizations across the energy ecosystem, and it’s accessed by both the regulatory commission and the regulated utility in numerous states, creating unique potential for a shared data environment.
Better Information, Better Decisions, Better Outcomes
Last month at HData’s Let’s Go! regulatory innovation conference, Commissioner Ana Ortega of the Florida Public Service Commission delivered a keynote address which explained to our audience how retrieving the right information is different from retrieving the highest volume of information. She stated that if the different parties in rate proceedings have a shared view of the same record, then their disagreements and negotiations can become more productive. The regulator’s ultimate decisions will be better-informed, thus driving better outcomes for the grid and its customers.
There’s a temptation to think of AI in energy regulation as a fight between opposing sides. If regulators get AI, then utilities need AI. If intervenors automate their questions, then utilities must automate their responses. That temptation is understandable, but the long-term answer cannot simply be to automate an increasingly dysfunctional adversarial process. The purpose of regulatory discovery is not to produce questions and answers. It’s to establish the facts needed for good decisions.
We can partially solve the problem of bad AI-generated data requests by giving utilities better AI to answer them.
But the real solution is to bring the people asking the questions and the people answering them together, on the same shared platform. Not so they’ll agree. Rather, so that they’ll disagree, more constructively, and that the extraordinary time, effort, and billions of dollars spent annually can go toward what actually matters: improving the grid on which we all depend.
About HData
As the operating system for energy regulation, HData serves the largest customer ecosystem in regulated energy, helping enterprises navigate regulatory complexity. Through centralized data, powerful AI, and workflow automation, HData accelerates the research, analysis, and insights critical to deciding the future of energy. Learn more at hdata.com.
