*Date: April 28, 2025* I come from a family of linemen. Both my grandpa and my dad climbed the poles that hold our grid together and did so in the middle of fierce Texas thunderstorms and blistering summers. My dad traded the backbreaking work for the comfort of an office chair long ago and eventually the title of COO. He now manages the crews that build and repair the grid and plays a large role in deciding where proactive work gets done. Each time I’ve asked how he makes those decisions, I've received the same answer with the same gesture: “It’s all up here, " he says, pointing to his cowboy hat. He’s a human system of record. Ironically, systems of record combined with AI agents are now trying to become human-like systems of record. It's one of the next big battlefield in vertical SaaS. If you're building an agent. Chase the data. If you own the data. Fear the agent. The collision between these two will mint the next billion-dollar winners. Systems of record are an organization's single source of truth. Data enters, is queried, and provides the answer for dozens, if not hundreds, of other applications. For the last five decades of tech, building a system of record was one of the best ways to ensure enduring success, take a look at these founding years: - SAP: 1972 - Oracle: 1977 - Epic: 1979 - Quickbooks: 1983 From the looks of it, the tech landscape has decided the next five decades will be defined by the rise of AI. But, AI needs great data. That seems simple, but in energy, there are many segments where the objective truth doesn’t exist in a data lake housed in the cloud. It exists in the minds of people like my dad. It’s intuition around where to upgrade for resiliency or how to trade power given a looming heatwave. The limiting reagent for AI in these sectors is the complexity of the real world. Some will argue that AI can learn this intuition with enough training data, and that may be true. But, good luck getting the C-suite responsible to turn grid reliability over to AI. “My bot did it” doesn’t fly in a televised congressional hearing. However, we can’t deny that AI will play a role in the next-generation grid software stack. It will only be as good as the system of record it pulls from or that it becomes. There are three ways systems of record build moats they can control the production of data, control the movement of data, or control the usage of data. The latter is when a system of record becomes a system of action, the most powerful moat of all. ### Controlling the Data Controlling the data is most important where the last 10% of accuracy makes a meaningful difference. Think about a system of record that has parts wrong on a substation, rolling a crew to repair an outage with the incorrect part spells disaster. This is a natural path for AI agents evolving into systems of record and we’re already seeing it occur within vectors like customer service. AI agents can collect and analyze common problems where humans can then take action. (Hello, system of action.) The path from controlling data to the system of record begins with data capture. AI continues to increase and improve data capture. It’s now easy to take a photo of an asset or listen in on sales/service calls to collect data from the field. ### Moving the Data Controlling the movement of data occurs today via application programming interfaces (API) and robotic process automations (RPA) - Sabre (travel), Visa (finance), and UiPath are all examples of connecting data across disparate systems. They are the systems of record for transactions in their respective industries. Businesses built on controlling external data movement are lucrative but rare. The moats are small. UiPath’s growth has stalled to 5% in the age of AI, **except** in highly regulated industries - energy fits this description. Net, controlling the movement of data can be defensible in certain sectors but it’s highly prone to disruption as most software “moves” data in some way - look no further than how many now have API documents or build integrations with other key systems of their customers. ### Using the Data The final type of data moat is the most powerful: controlling the usage of data. These systems of action don’t just store data passively; they enable actions on top of it. For all its faults, this is where Salesforce excels. It is the system of action for revenue generation. When account executive logs in, they should, in theory be able to act upon the most lucrative and likely to close opportunities. But, it’s only as good as the data it contains. If you’ve ever worked in any sales organization, you know that reliable data in the CRM is a daily battle. Do you know what doesn’t forget to enter data? AI agents. AI agents also excel at multi-variate decision-making. In energy, that means things like dispatching crews based on outage reports and data from the field, executing trades based on wholesale market conditions, and maybe someday, deciding where to build the infrastructure itself. This combination of capturing data accurately and then acting on it, rather through an automated process or by enabling skilled humans, is powerful and defensible. ### Data Gravity - The Goal of Data Moats Browse any energy or industrial software website and the odds are high that you'll find the logo of Oracle Premivera or SAP somewhere. The odds are doubly high that you'll find out they are not user friendly and customer feedback will be "they are ripe for disruption." So why are they still in place? Data gravity. Data gravity in vertical SaaS means a tool that controls the most valuable data and whose importance tends to swallow less meaningful or productive systems. It means that many other (non-data) moats like workflows, revenue, and networks flow through the system. As we've noted, in energy, "the truth" isn't fully objective. It's defined in part by what people believe. The more people trust the system and its data, the more likely its data gravity will increase. Regardless of entry point and enduring moat, all systems of record should look for ways to increase their gravitational pull. ### So, what... As AI agents advance, they're creating a new opportunity for systems of record. The most successful systems won't just passively collect data—they'll actively interpret, learn from, and act upon it. Current systems of record have a massive leg up on the competition as they have the data needed to win these opportunities. In the energy sector, where intuition and expertise have traditionally governed critical decisions, the winners will be those who build systems that can translate human knowledge into structured data while maintaining the nuance and context that makes it valuable. The companies that understand this and position themselves at the intersection of data capture, movement, and action will build the enduring moats of the next decade.