Where can AI and intelligent automation deliver practical value?
Intelligent automation (IA) combines artificial intelligence with technologies such as workflow management, robotic process automation (RPA), and system integrations to automate business processes. It connects the ability to interpret information with the ability to execute tasks, while retaining human oversight for exceptions and consequential decisions.
Simply put: AI helps interpret information; automation puts it to work.
Some of the strongest opportunities involve routine processes that require significant manual effort but follow predictable steps.
In accounts payable, a complete workflow can receive an invoice, extract information, validate it against business rules, match purchase orders and receiving records, route exceptions, obtain approval, and post the approved transaction.
AI interprets documents, business rules enforce checks, and workflow management routes exceptions to employees. Integrations or RPA bots move approved data between systems; employees retain payment approval authority.
For a flagged customer bill, IA could gather account history and meter readings, identify discrepancies, and prepare a case for specialist review. Corrections would require authorization; the workflow would not independently change rates or disconnect service.
In work order and field service administration, IA could check documentation, assemble supporting records, and route missing information to the appropriate team. An authorized employee would confirm completion before closing the order; the automation would not control physical equipment.
Similar opportunities exist in procurement and regulatory reporting.
Evaluate candidates by transaction volume, manual effort, data readiness, exception frequency, integration feasibility, and consequences of error. Standardize and simplify the process, clarify rules and ownership, then select the simplest suitable technology. Rules-based workflows or integrations may suffice; use AI where interpreting variable information adds value. Not every workflow needs an AI agent.
Utilities can also use AI-powered productivity tools to support research, document drafting, data analysis, and other knowledge-intensive work. These applications require different controls from systems that directly execute transactions.
The distinction matters because the consequences of an error change when a tool moves from generating information to taking action.
How do bots and AI agents work together?
Within IA, RPA bots and AI agents can play complementary roles. Bots execute predefined steps with predictable inputs and rules. AI agents can interpret information, use connected tools, and select actions within configured boundaries.
In the invoice workflow, a bot could transfer approved data while an agent gathers supporting information about a discrepancy and recommends a next step. The agent’s flexibility calls for controls suited to the actions it can take.
Combine these tools where each adds value, with permissions, approval requirements, exception handling, and monitoring matched to their responsibilities.
What does responsible AI governance look like for utilities?
AI governance begins with visibility into which tools employees use, what information those tools access, and what decisions they influence.
Unapproved AI applications can expose sensitive information or create inconsistent practices. Clear policies can establish acceptable uses, data restrictions, approval requirements, and accountability.
Governance also needs to address automated tools that interact with other systems. As utilities deploy more AI agents and bots, access management must cover digital identities as carefully as employee accounts.
A practical approach to AI adoption
- Select a defined use case. Identify the process, business problem, expected value, and measurable results.
- Assess the risks. Review data sensitivity, system connections, operational impacts, and potential errors.
- Establish controls. Define access permissions, approval requirements, monitoring, and escalation procedures.
- Test before expanding. Evaluate performance, exceptions, accuracy, and unintended outcomes in a controlled setting.
- Assign ongoing ownership. Identify who monitors the technology, manages changes, and responds when performance falls outside expectations.
Formal frameworks can support this work. ISO 42001 provides requirements for AI management systems, while the NIST AI RMF offers guidance for identifying and managing AI risks.
As adoption grows, utilities may also benefit from a centralized process for evaluating use cases, approving tools, monitoring performance, and managing automated identities.
The opportunity is substantial, particularly in administrative processes where manual work consumes valuable staff time. Utilities that establish clear controls and accountability can pursue those efficiencies while keeping reliability, security, and public service responsibilities at the center of technology decisions.