Regulatory teams in energy face two pressures at once: the volume of filings keeps growing, and the deadlines don't move. Researching dockets, analyzing testimony, and turning it into a position still takes days the calendar doesn't always allow. AI promises relief, so the question comes up fast: should we build our own tool or buy one? But it's the wrong place to start.
At LET'S GO! 2025, HData's annual customer conference, Leo Trudel, Principal at the energy advisory firm Realize 2050, walked the audience through a framework for thinking about that question more rigorously. According to Trudel, while there is no universal right answer to build vs. buy, there is a repeatable way to reach the right answer for your organization and your use case. In regulated energy, where the work is complex and the cost of a mistake is high, that discipline matters more than in most fields.
Utilities run pilots constantly, but very few of them become production systems.
“The pilot success rate is somewhere around 5%,” Trudel said, because “the capabilities don’t match the requirements.”
That gap shows up when a team brings in technology before it understands three things: the problem it's solving, what every affected team needs, and whether the technology even fits the job. Quarterly pressure makes it worse, pushing teams toward quick wins when real progress depends on foundational work. And the people in regulatory, compliance, and legal are often consulted last, even though a tool built for one team frequently touches their work too.
For regulated energy organizations, that group of stakeholders is usually larger than it first appears: a workflow tool for engineering may touch rate base or a document management tool may touch a regulatory recordkeeping requirement. The cost of discovering those overlaps late in a procurement cycle can be high.
“The only way you can fully understand the requirements of a problem is to talk to all the groups that are going to be affected,” Trudel said.
Teams that struggle usually start with the technology. They shop for a tool, then look for a problem to justify it.
The better order begins with an honest look: Where are you behind your peers? Which metric actually matters? What would “better” look like in concrete terms? From there, identify the use cases tied to that goal, gather requirements, shortlist tools that could meet them, refine those requirements with stakeholders, and only then write the request for proposal.
Skip those steps and you get long procurement cycles that collapse at the finish line over a missing security certification, a data residency gap, or no audit trail. For regulated organizations, those late disqualifiers are exactly the requirements that are easiest to overlook early.
Building usually costs more than a first quote suggests, especially when a purpose-built platform already exists. The upfront cost is only a kind of down payment: an in-house system has to be maintained for as long as you run it, and regulatory work makes that steeper because sources, dockets, and data formats change constantly.
Time is the other cost. An existing platform is ready now, whereas a built one arrives in development cycles. For teams whose core problems are deadlines, waiting a year for a tool to mature defeats the purpose.
In essence, the prerequisites are real: reliable data sources, a plan to collect, label, and maintain it, scarce AI talent, and meaningful work to keep that talent. Trudel described a utility data science team that dissolved within three years: not over pay, but because the work itself did not match what the team was promised.
“Homegrown solutions are twice as likely to fail as off-the-shelf solutions,” Trudel remarked.
Build only when you must, and only when you’re ready for what it takes.
For regulatory teams, the question is not build vs. buy. It is whether you’re willing to change how you make the decision: starting with clear objectives, cross-functional input and alignment, and more disciplined evaluation. When those are in place, the build vs. buy answer usually reveals itself.
For regulatory work in particular, the right tool needs to understand not just data but the regulatory ecosystem that data lives in. That discipline is what separates technology investments that produce durable value from the ones that quietly stall.
Trudel’s full session and other videos at LET'S GO! 2025 are available for download. For more on AI built specifically for the regulatory work of utilities, regulators, advocates, and other energy professionals, visit hdata.com.
As the AI-native operating system for energy regulation, HData serves the largest customer ecosystem in regulated energy, helping utilities, regulators, advocates, advisory firms, corporates, and energy technology companies navigate regulatory complexity. Through centralized data, domain AI, and purpose-built applications, HData accelerates the research, analysis, and workflows critical to how the future of energy is decided.