Helen Friedland is Chief Technology Officer, Halcyon

A new way to understand complex dockets
If you’re in the energy information business, you’re probably pretty well versed in docket diving. As with other legal proceedings, public utility commissions (PUCs) organize their work via dockets, which requires key stakeholders (like regulated utilities, parties and intervenors) to share evidence that informs PUC decision making.

While these dockets contain incredibly important information about the future of our energy infrastructure, they are dense, and the volume of documents within can be overwhelming. When you open a docket, you're greeted with a list of documents — sometimes thousands long — which, if you’re lucky, may have a one-sentence description that might be useful. There is little organization; there are no summaries. For those of us who build things for a living, it's like being handed a jigsaw puzzle without the picture on the box. 

This lack of clarity materially impacts our understanding of the docket. How do you figure out what's really going on? Who are the parties involved? What opinions do they hold? What’s the current status of the docket, and, more importantly, where is it headed? 

Understanding these elements is essential. But even for a single docket, it can be painfully onerous and, in an industry where better information access is a competitive advantage, simply takes too long.

Summarizing dockets with LLMs and RAG
Docket analysts navigating this never-ending labyrinth of lengthy PDFs typically have two use cases:

1. They are in the discovery/early-stage research phase and need to identify, organize, and analyze one or multiple dockets. During this information collection phase, analysts need a way to quickly grok a docket’s high-level information without going page by page.
2. They are already tracking specific docket(s) for specific projects and are well versed in docket status and details. These analysts need search capabilities more sophisticated than CTRL+F to search for specific terms over thousands of pages, and they need monitoring systems that do more than just tell them when there’s a new filing or update.

Right now, we’re focused on solving for use case #1 (with some exciting stuff in the works for #2). Using Retrieval-Augmented Generation (RAG), we can tap into a well-defined set of documents, pose clear questions, and extract meaningful answers — all without having to click through endless files. RAG is particularly well-suited for docket analysis: the set of documents (while admittedly quite large!) is defined; the questions asked of these dockets are clear and largely extensible across dockets; and the overarching context (in our case, energy) of each docket is well-defined and consistent. 

Teaching LLMs to think like energy analysts
While the idea of using LLMs and RAG for docket analysis sounds promising, there was a significant challenge: these documents are steeped in energy sector jargon, acronyms, and complex technical language. To get the most out of our technology, we needed to ensure it understands the intricacies of the energy industry.

We tackled this by combining our proprietary knowledge graph with custom prompting strategies. Our knowledge graph leverages our energy subject-matter experts’ decades of experience to encode industry knowledge into our platform and help surface connections between fragmented, opaque data; our prompting strategies rely on the deep understanding of analyst workflows and business needs we’ve gained by working closely with our customers over the last two years. 

Essentially, we’ve given the LLM the very extensive, very nuanced context it needs to decipher the energy industry’s unique language and think like an energy analyst.

Building Docket profiles 
We're thrilled to introduce the first version of Docket profiles - here are a few examples (must have a Halcyon account to view):

• Colorado Public Utilities Commission 24A-0296E
• California Public Utilities Commission C2411013
• New York Public Utilities Commission 21-E-0580

 Here’s how it works:

• Automatic data collection: We crawl, process, and index PUC documents daily, so the docket information we have is always up to date. Notably, LLMs are not involved here — we don’t need to train a model to get the latest data (and we include timestamps that refer to when we most recently updated a given docket’s information). 
• Uniform presentation of metadata: Docket profiles aggregate valuable metadata scraped from various PUC websites. No more hopping from site to site — the key information is unified in one place.
• RAG-powered queries against all documents within a given docket: Beyond the basic metadata, Docket profiles include smart queries that provide comprehensive summaries and answers to the most pressing questions about the docket.

Here’s what they look like - current alerts subscribers can use our free docket search capability to access Docket profiles as well. 

Our goal with Docket profiles is to provide a comprehensive overview of any PUC docket at a glance. We want you to know who the players are, what the key issues are, and where the docket might be headed — all without having to manually comb through a mountain of PDFs.

Help improve Docket profiles!
Right now, the best way to access Docket profiles is by signing up for free Halcyon Alerts. Once you do, every email with an update to a docket will have a link to the corresponding Docket profile:

As mentioned, alerts subscribers can also use docket search to access all of our Docket profiles from directly within our platform. 

While we’re excited about this initial release, we know there’s always room for improvement. We’re already brainstorming ways to make it even easier for you to "dive into" or "dissect" a docket. If you have any ideas or wish-list items for what you’d like to see in future versions, please let us know. We’re building with — and for — energy professionals who make their living wading through these dockets every day. 

We'll be back next week with part 2 that will share first-hand perspective from one of those energy professionals on just how useful these Docket profiles can be! 

Happy docket exploring!

Comments or questions? We’d love to hear from you - sayhi@halcyon.eco, or find us on LinkedIn and Twitter