Utility AI is moving beyond general-purpose assistants toward applications connected to operating systems, analytical tools, and business processes.
As these applications take shape, teams have to decide which work belongs in the agent and which should remain in the systems behind it. Understanding how customers use energy, which technologies they have adopted, how load varies across populations and locations, where demand-side opportunities exist, and how programs perform requires in-depth analysis of interval meter data and related customer, site, and grid information.
Utilities own the unique applications, workflows, and customer experiences that empower their business. But the answers those applications give are only as good as the analysis behind them. To spot where EV adoption is straining local circuits or to estimate how much capacity a portfolio can deliver tomorrow, utilities need a trusted data foundation that captures how individual customers and populations actually use energy.
The Energy Data Gap within AI Architecture
A few questions your utility application may want to answer include:
- What is driving this customer’s energy use, and what technologies have they adopted?
- Where is demand-side potential concentrated relative to areas of grid need?
- How much capacity can our portfolio likely deliver tomorrow?
- Where are EVs, solar, and other distributed energy technologies being adopted, and how are they changing load?
- What programs are suited for my customers based on their usage characteristics?
Existing utility applications may already have access to customer, program, grid, and operational systems. However, connecting an agent to these systems does not provide the analytical layer needed to translate raw data into reliable energy intelligence.
When that layer is missing, the risk compounds. An AI application can deliver a plausible answer about customer usage or portfolio capacity that no one can trace back to defensible methods. Yet once that answer shapes a grid decision, a customer conversation, or a regulatory filing, the cost of getting it wrong falls on the utility.
FLEX: Your Trusted Energy Analytics Engine
Whether utilities build their own AI applications, adopt prebuilt agents, or run a complete AI operating system, those applications need a reliable tool for energy analysis.
FLEX provides that missing layer. With more than 55 million meters under management, Recurve's FLEX platform delivers specialized energy analysis to the utility's own applications (or any agent platform it chooses), drawing on interval meter data and related customer, site, technology, program, and grid attributes.
The agent can gather context, understand the user’s request, and call FLEX for the analysis it needs. FLEX performs the specialized analytical work, from understanding customer load and technology adoption to identifying opportunities, forecasting demand, and measuring performance. The utility controls the experience: how it presents those insights, what other systems and data it incorporates, and what actions users or applications can take.
This separation also makes the system easier to evaluate and govern. FLEX’s analytical results can be validated independently of the agent’s interpretation and use, giving utilities visibility into both the underlying energy analysis and the AI experience built on top of it.
What FLEX Contributes
FLEX turns interval meter data and related customer, site, technology, program, and grid data into intelligence about how customers use energy, where opportunities exist, and how demand-side resources perform.
Data Checks
That starts with the data itself. FLEX checks incoming data against documented schemas and flags issues such as duplicate readings, conflicting values, inconsistent intervals, and insufficient history.
Customer and Technology Insights
FLEX can characterize customer load and identify patterns such as heating and cooling use, peak-period consumption, load shape, weather response, and the presence and impact of technologies such as EVs and solar. These insights can be applied to an individual customer or analyzed across populations and geography, and combined with customer, site, program, and grid attributes to support customer interactions, understand adoption, identify opportunities, and define populations for particular programs or system needs.
Capacity Forecasting
For forecasting, FLEX uses interval data, enrollment and event history, and available technology and grid attributes to estimate the capacity demand-side resources are likely to deliver.
Measurement
For measurement, FLEX uses historical interval data, weather, and, where applicable, solar irradiance to model typical consumption at the meter level. Recurve applies proven open methods to estimate counterfactual consumption: what a meter would likely have recorded without an intervention or event. The platform also supports comparison group correction, meter-level qualification rules, and data quality checks, with results available from individual meters through customer segments, programs, and portfolios.
These are existing FLEX capabilities. Recurve is working with utility design partners to determine how to expose this intelligence to AI applications and which integration patterns are most useful.
A Representative Workflow
Consider an application used by a demand-response program team.
A user asks for the expected capacity of a portfolio during a three-hour event the following afternoon. The agent identifies the portfolio and event window, collects any additional parameters the application requires, and requests a forecast.
FLEX produces the forecast using the program data and configured analytical methods. The application can then present the result alongside previous event performance, portfolio composition, or other information the utility considers relevant.
The utility team controls the workflow. It decides who has access, how results are displayed, what supporting detail is available, and whether the application may initiate another action.
FLEX supplies one part of the system: the underlying energy analysis.
The same approach can work at the individual customer level. A customer contacts the utility about an unexpectedly high bill. A customer service application can use FLEX to understand the customer’s load shape, changes in usage, weather sensitivity, detected technologies such as an EV or solar, and suggest programs for enrollment. The agent can combine that analysis with billing, rate, account, and program information already available to the utility to explain what is driving the customer’s usage and identify relevant ways to reduce their bill.
These examples are starting points, not prescribed workflows. Each utility will have its own data, methods, operating requirements, and customer priorities.
Working with Existing Utility Capabilities
Many utilities already have mature data engineering and data science environments, including internal forecasting models, established measurement methods, and analytical services developed for specific programs. FLEX is designed to complement those capabilities, not replace them.
Where FLEX fits will vary by utility and use case. A utility might use FLEX to understand customer load and technology adoption while retaining its existing forecasting service. Another might incorporate FLEX targeting or grid intelligence into an internally developed planning application. Others may use FLEX forecasting and measurement together within a program management workflow.
In each case, the utility controls the application, workflow, and customer experience. FLEX provides specialized energy intelligence where it can add capability, reduce duplicate implementation, or provide consistent analytical methods across applications. Its documented methods and meter-level results also allow the underlying analysis to be reviewed independently from the AI application using it.
The architecture should follow the utility’s requirements. Recurve brings specialized models, methods, and experience for understanding energy use at the meter level and across millions of customers, while utility teams bring the customer relationships, local context, operational systems, and business requirements that determine how that intelligence should be used.
Recurve is currently working with utility design partners to define the highest value integrations and use cases. Interested in evaluating how FLEX fits within your AI architecture? Connect with our team to set up time to continue the conversation.