We sat down recently for a question and answer session with HData Chief Technology Officer, Yuval Lubowich. Here’s what the AI and industry expert had to say.
Q: What’s your background, Yuval? How did you get started in this industry?
A: It all started with my passion for computers and what people can do with them. I was curious at an early age. I got my first computer at the age of 8 and I never looked back. It seems like a million years ago now, but my dad bought me an Atari 800XL. I used it to teach myself how to code. I guess I was one of the lucky people who knew from a young age what I wanted to do when I grew up! I went on to get my Bachelor's and Master's in computer science and have worked in small and large tech companies for the better part of 30 years. Fast forward to today, I am the co-founder and CTO at HData.
Q: How do you describe HData to people who ask?
A: We’re a RegTech company that focuses on Regulated Energy companies in the US. We’re transforming the entire energy industry by using AI and automation to make it easy for anyone to file, explore, analyze, and leverage regulatory data. It’s huge and exciting work I’m proud to be part of.
Q: Can you share a moment in which you realized AI could fundamentally change the way we handle regulatory data in the energy sector?
A: Yes. Absolutely. Several weeks ago, I met the general counsel of one of the biggest utilities in the country. We were there to demo our Regulatory AI solution. The meeting started out a few minutes late. We found out it’s because the team we were demoing to just received a new rate case. They’d spent a long night reviewing it.
If you’re not familiar with rate cases, they’re a formal process used to determine the amount of money regulated utilities can charge customers for electricity, natural gas, private water, and steam service. Rate cases are a primary instrument of government regulation of these industries.
The general counsel asked me, “Are you a betting man?” and I said, “Usually no. But I know what you are really asking, so let’s do it!” He sent me two questions his team had answered during the night and asked me to pose them, word-for-word, to our generative AI solution.
The system took a few seconds to answer the questions, but when he read them, the look on his face was priceless. He was floored.
He told us that his top lawyer, who has over 30 years of experience handling rate cases, would have taken 30 minutes to draft a response to one question. Our system did that in seconds. The quality of the answer was on par with what his lawyer would have written, too! This team we were talking about handles more than 10,000 questions per rate case on an annual basis. The ROI was a no-brainer.
That’s when I knew AI could fundamentally change the way we handle regulatory data in the energy sector!
Q: Can you discuss a specific instance where HData's AI provided a solution to a regulatory challenge that seemed insurmountable?
A: One of our customers has an internal subject matter expert who manages the company’s rate cases. Their whole job is to review all of the documents and testimony in each of the company’s rate cases—and similar rate cases across the industry. Nobody knows as much about rate cases relevant to the company as this person. Their knowledge is crucial to the company’s success in every rate case they’re involved in.
The subject matter expert is getting close to retiring.
The rest of the team does not have the capacity to go back and read the hundreds of thousands of pages from multiple decades' worth of rate cases from the past, much less understand them well enough to duplicate the subject matter expert’s capability.
Fortunately, the subject matter expert recently started using HData’s Regulatory AI. Our team worked with the customer to load the most recent rate case into a Private Catalog. This means every document can be instantly and accurately summarized for any member of the team. And not just summarized but interpreted.
Thanks to our tool, it’s going to be less disruptive when the rate case expert retires.
Q: Could you delve into how HData's AI impacts decision-making processes in energy companies?
A: We don’t get to see specific decisions because our platform maintains each of our customers’ confidential information. But we do think the most forward-looking utilities are using Regulatory AI because of the raw usage figures. Regulatory AI was released just last summer, and in that time, our query volume has grown by a factor of ten.
Q: How do you ensure that HData's AI remains adaptable and up-to-date with constantly evolving energy regulations and policies?
A: Regulatory AI is able to stay up-to-date because HData feeds it updated intel every single day. Every ten minutes or less, every new item published on a regulatory commission docket is fed into our platform. Our platform preprocesses all official regulatory documents so that they are ready to serve as the basis of queries.
Preprocessing ensures that Regulatory AI’s answers won’t hallucinate because every query is automatically accompanied, within our back-end, by the relevant “chunks” of the document or documents (or a whole file) and only the relevant chunks.
Q: How do you see HData's role in the broader context of the global energy transition towards more sustainable and renewable sources?
A: Well, you can’t manage what you can’t track, and until Regulatory AI, the energy transition was too big to track.
There are just too many decisions, reports, actions, regulations, and documents related to the energy transition being released by regulatory agencies on the federal and state level in the United States alone for any one user to read them all.
But we don’t need users to read them all anymore.
Regulatory AI users can set alerts for terms or concepts they need to track, and they’ll receive notifications whenever a new docket item that pings those concepts is released. They can feed every docket item automatically through Regulatory AI for an instant summary and interpretation.
Imagine if you hired a subject matter expert in distributed energy resources or electric vehicles to monitor regulatory releases and give you the bullet points on every release relevant to their topic. It’s like that, but much, much cheaper.
Q: In terms of data privacy and security, what measures does HData take to protect sensitive information within its AI-driven platform?
A: First, we don’t need to train Regulatory AI because the real value of Regulatory AI is not in the Large Language Models (LLMs) but in the technology we built around them. In fact, the LLMs are almost interchangeable; we could provide the same value using either ChatGPT or Gemini.
HData’s technology built around the LLMs is where the regulatory information is preprocessed. This is the technology that stops the LLMs from hallucinating and allows Regulatory AI to provide links back to the original source documents of every answer. The LLMs we embed within this technology are already trained, and it’s the preprocessing that counts, so we don’t need to train them on our customers’ data.
Second, although we do allow our customers to upload and query their own internal documents, those are completely separate and secure, accessible only to the company that uploaded them. Most of the work that Regulatory AI does focuses on publicly-available information. We have the best, most comprehensive data store of energy regulatory content, both structured and unstructured, and we serve all of that content to Regulatory AI so that it is ready to be queried.
Q: How do you envision this technology evolving over the next five years?
A: Our next frontier is to improve how Regulatory AI reaches across the boundary between unstructured and structured content.
Imagine asking Regulatory AI to describe the Return on Equity of every utility company with revenue over $1 billion in the U.S. Southeast, for example. That query requires a response with both structured and unstructured components.
Before long, HData’s Regulatory AI will be able to handle that query. But that’s not a five-year prediction. That’s a one-year prediction.
Q: What advice would you give to young entrepreneurs or technologists based on your experiences with HData and AI?
A: You do your best work under pressure. Don’t seek freedom from pressure. Instead, seek ways to deal with pressure productively.
Also, and no less important, read. Spend at least an hour a day educating yourself, seeing what’s out there and what others are doing. Work and meetings will wait till you are done.
Are you interested in learning more about artificial intelligence in the utilities sector? Curious to see first-hand what HData products can do? Request a copy of our ebook to learn more.