By: Susanne Bottemanne

The systems electric utilities rely on provide fertile ground for agentic AI. In ADMS systems alone, they promise to enable intelligent outage restoration, coordinate distributed energy resources, speed up storm response, optimize volt-var control, improve load forecasting, and more.

But while AI agents operate in the digital world, utilities and their customers depend on prodigious amounts of hardware across vast expanses of land – and, given offshore wind, sea. Wherever hardware may be, it wears out and/or breaks. Intelligent, autonomous field robots are understandably attracting serious attention and investment in this industry.

Autonomous robots promise to reduce operation and maintenance costs and extend equipment life amid mounting pressures from aging infrastructure, workforce gaps, grid-modernization needs, and massive demand from data centers. These robots, which span from the four-legged to the winged, can inspect hardware, detect anomalies, and trigger responses without human involvement. Utilities now recognize that, while autonomous robots can add value in isolation, they work best when they interface with AI agents that are integrated into operating and business systems.

Utilities are adopting physical AI in a variety of contexts, with asset health and management emerging as potentially high-ROI application areas. That adoption has come with a recognition that smart field robots must act upon intelligence from enterprise workflows and then feed new information from their tasks back into those workflows. That way, staff and agentic AI working in operations, human resources, finance, and procurement can act and react to asset-related developments quickly and efficiently. Here are four key steps to achieving the integration essential to exploiting the true potential of physical AI.

1)      Let the utility’s specific needs drive physical AI implementation strategy

Use cases should dictate a utility’s autonomous robot implementation path. Many are already working with autonomous drones in areas such as automating vegetation management and maintaining utility-scale solar arrays.

Texas-based Percepto’s autonomous patrolling drones present an example of unmanned aerial systems going beyond protecting power lines from branches. The company’s distribution-grid inspection systems can identify broken components, rust, rot, and corrosion, as well as spot overheating components with AI-driven temperature assessments.

Israel-based Ecoppia has deployed water-free autonomous systems to clean solar panels across dozens of utility-scale projects across the Middle East and India. (In dusty regions, keeping solar modules clean can boost efficiency by 20% to 35%, thereby starkly improving the levelized cost of ownership.)

Consider also an offshore wind producer that’s deploying a four-legged robot to keep tabs on an offshore wind platform. The inspection robot, made by Swiss-based ANYbotics, eliminated the need to send techs to a remote platform for months. When human intervention was eventually required, the maintenance team already knew what was wrong, which expert to send, and what equipment and spare parts to bring, avoiding costly and risky trial-and-error platform visits.

These sorts of asset health and management applications beg for agentic AI combined with autonomous robots. Such integration enables robots to inspect assets continuously with sensors that can spot anomalies based on visual, chemical, acoustic, and thermal signals; send data directly to field service management systems; have those systems assign work orders (or suggest teeing up work orders for human decision-makers); and propagate the impacts of the work to business systems (operations, procurement, finance, and so on). That gives the organization an immediate understanding of operational status and the business impacts related to that status.

If autonomous robot integration can boost the uptime of an asset portfolio by a percentage point or two, investments in physical AI will have paid for themselves many times over. But utilities are recognizing that, to properly harness such capabilities, it also often takes substantial work on the data side.

2)      Get the utility’s data cleaned up and harmonized

A utility’s goal should be to incorporate a portfolio of autonomous-robot inspection into a maximally closed-loop system in which inspections trigger automated work orders for repairs and that automatically cascade through field service management, warehousing, procurement, finance, and beyond. Such data flows aren’t possible if a utility has inconsistent, duplicative, stale, or siloed master data. That means bringing together SCADA and other operational data with a field service and asset management system’s planning and scheduling solution, as well as with core business systems. The result is harmonized data across the utility’s diverse IT landscape.