AI Agent Operational Lift for Florida Public Utilities Company in Fernandina Beach, Florida

Why AI matters at this scale

Florida Public Utilities Company operates as a small, community-focused electric, gas, and water utility serving a defined territory in northeast Florida. With 201–500 employees and estimated annual revenues around $65 million, FPUC sits in a challenging middle ground: large enough to generate meaningful operational data from smart meters and SCADA systems, yet small enough to lack dedicated data science teams. This size band is precisely where targeted, cloud-based AI tools can deliver disproportionate returns by automating decisions that currently rely on tribal knowledge and manual spreadsheet analysis.

For a utility of this scale, AI is not about building custom deep learning models from scratch. It is about leveraging pre-built solutions that ingest existing data streams—AMI interval reads, outage management system logs, GIS vegetation layers—and turn them into actionable work orders. The goal is to do more with the same headcount, improving reliability metrics that directly impact regulatory standing and customer satisfaction.

Three concrete AI opportunities

1. Predictive vegetation management. Florida’s storm season makes tree contact the leading cause of outages. By running satellite imagery and LiDAR data through a machine learning classifier, FPUC can rank circuit segments by trim urgency. This shifts crews from fixed-cycle trimming to risk-based scheduling, potentially reducing vegetation-related outage minutes by 15–20% while avoiding unnecessary truck rolls.

2. AMI-based load analytics. FPUC’s smart meter network produces interval data that remains largely unused beyond billing. AI-driven disaggregation can identify inefficient HVAC units or pool pumps driving peak demand. Proactive customer alerts with tailored efficiency tips reduce peak load, deferring costly capacity upgrades. Even a 2–3% peak shave translates to meaningful savings on wholesale power purchases.

3. NLP on work order history. Years of unstructured crew notes contain hidden failure patterns. A natural language processing pipeline can extract common root causes—such as specific transformer models failing after voltage sags—and feed those insights into preventive maintenance schedules. This turns reactive repairs into condition-based maintenance without adding engineering staff.

Deployment risks specific to this size band

Mid-sized utilities face unique AI adoption risks. First, data quality is often inconsistent; SCADA historians may have gaps, and GIS records may not reflect field reality. Any AI model is only as good as its input data, so a data cleansing sprint must precede any pilot. Second, IT/OT convergence is a real cybersecurity concern. AI tools that touch operational networks must be rigorously segmented to avoid introducing vulnerabilities into critical control systems. Third, change management is harder in a small organization where field crews and dispatchers have deep institutional knowledge. AI recommendations will be ignored unless presented as decision support rather than replacement. Starting with a single, high-ROI use case championed by operations leadership is the safest path to building trust and demonstrating value before scaling.