LET'S CONNECT

AI in Utilities

Accelerating Common Information Model (CIM) Adoption with AI

December 2025
Challenges of Data Interoperability in Utilities
  • Utilities rely on dozens, often siloed, information systems, including SCADA, AMI, GIS, CIS, DERMS, OMS, and more. Even when two or more of these systems are designed to communicate, their underlying data structures are frequently legacy formats that do not easily align together for interoperability.

  • In today’s utility environment, seamless data exchange is critical for reliability, forecasting, operational planning, customer programs, and regulatory compliance. When cross-functional teams collaborate to interoperate, they often spend significant time discussing models and data structures, not the business solution. Even when organizations invest in an enterprise service bus or integration platform, they still need to answer the question:

  • What data format should be used to ensure everything speaks the same language?
  • To answer the question, we can look at the Common Information Model (CIM), the global standard for utility data exchange (IEC 61970/61968). CIM greatly improves interoperability across heterogeneous information systems. However, converting data from existing systems to CIM is often time-consuming, complex, and prone to error when not planned from the start.
Difficulties with Manual CIM Conversion
  • The approach to mapping a legacy data model into CIM resembles any other data model mapping:
  • Inventory – create an inventory of the legacy schema, if one does not already exist
  • Mapping – map the legacy data structures to a CIM equivalent
  • Transformation – build ETL scripts to perform the data transfer
  • Validation – manually check or build complex tools to verify data transfers correctly
  • However, mapping has challenges such as:
  • Syntactic differences – differing naming conventions (such as TransformerID vs. xfmr_id)
  • Semantic differences – variations in data types, abbreviations, and formatting
  • Missing or inconsistent metadata – assists in the mapping process, both schema and data ETL
  • Complex relationships – multiple legacy fields map to one CIM attribute, or vice versa
  • Knowledge gaps – teams may not fully understand both data models
These issues slow down projects, increase integration costs, and require repeated iterations with teams that may already be time constrained.

What is CIM

At the 2025 Utility Analytics Week conference, a presenter asked how many people knew what CIM was. In a room of more than 700 attendees, only a handful raised their hands. CIM has been used primarily by operations groups, but its scope spans every part of a utility. At its core, CIM defines a shared, standardized data model that allows systems to communicate consistently using a common vocabulary. It includes models for assets, networks, measurements, customers, and operations. For areas which data structures do not exist, CIM has ways to extend its model to cover schemas not in the specification. The specification for CIM is a UML-based information model. The official serialization formats are RDF/XML and XML, though developers and modelers are increasingly using JSON due to its ease of integration with modern applications. No matter the message or storage format, a CIM-based model provides a unified way to represent any utility domain. For more information, the CIM User Group provides extensive resources and videos.

How AI Can be Used

Utilizing a vast corpus of previous data, AI can augment the CIM conversion process by acting as an assistant during mapping and validation. The following tasks can be assisted with AI:
  • Schema Mapping – AI can suggest CIM-equivalent classes and attributes based on legacy schemas
  • Pattern Recognition – By analyzing sample data, AI can infer field meanings and improve mapping accuracy
  • Transformation – AI can generate mapping logic or ETL code
  • Validation – After the ETL process, AI can detect missing translations, incomplete relationships, and inconsistent mappings
As with other use cases in our AI in Utilities series, AI does not replace utility experts and instead, expands the toolkit available to them. This allows people to focus on higher-value analysis and engineering work.

Considerations

Using AI in CIM conversion brings significant benefits but also requires planning at all stages:
  • Data Governance & Security – know what data can be provided to AI systems. If using external AI systems, legacy data should be anonymized and randomized. CIM models are public domain and can be passed safely to AI models, but legacy data models may not be publicly available
  • Human Oversight – AI can assist with mapping, but human experts must remain in the loop for validation and approval
  • Start Small – begin with a limited dataset or domain to identify quick wins and refine the process
  • Integration with Existing Tools – consider how generated ETL logic will be incorporated into your existing integration platforms or CI/CD pipelines

What's Next?

AI can accelerate the first-time conversion of a legacy data model into CIM. By reducing manual mapping time and enabling teams to focus on higher-value tasks, utilities can save both time and resources. While CIM conversion is often viewed as a one-time exercise, AI enables it to become an ongoing process. Focus areas include:
  • Continuous validation of data exchange, both batch and real-time
  • Automatic documentation of mappings and transformation
  • Predictive impact analysis as systems change and process more information.
The possibilities for using AI in data mapping, not just for CIM, are limitless. By taking steps towards small integration of AI into your organization, AI can be used as a powerful tool for utility data modernization.