South Carolina's state-owned electric and water provider Santee Cooper announced Thursday it's partnering with Google to deploy artificial intelligence tools that could save the utility millions of dollars annually through better weather prediction and financial planning. The collaboration, revealed during a Wednesday press call by CFO Tami Wilson, marks what she described as a "major digital transformation" aimed at moving away from outdated processes. The new system will help the utility manage forecasting for major initiatives, including a $10 billion grid-expansion budget.
The utility is rolling out Google's WeatherNext 2 technology, which delivers probability-based weather forecasts up to 15 days ahead, and Google's Gemini Enterprise Agent Platform for speeding up financial scenario planning. Santee Cooper is also distributing several hundred Gemini Enterprise licenses throughout the organization to modernize workflows across the board. Wilson explained the utility will use generative AI for what she called "low-cost initial drafting," while agentic AI will handle multi-step processes that run on their own. The custom forecasting model Google is building will account for local weather quirks in Santee Cooper's service area, including Lake Marion and Lake Moultrie—two artificial lakes big enough to generate their own microclimate.
According to Wilson, a single degree of error in temperature prediction can shift the utility's power demand by roughly 100 megawatts per hour, enough to power tens of thousands of homes. "If we're caught short on a freezing day in the winter or a scorching afternoon, it could cost us up to $100,000 an hour to go out and procure emergency power," she said. Wilson emphasized that the utility wants to "automate the routine" while keeping accountability and risk oversight with human operators. Raiford Smith, global director of power and energy at Google Cloud, noted that Gemini Enterprise was built "specifically to serve as that intelligent operating layer for modern utilities" with capabilities designed for complex infrastructure.
The push for AI reflects a gap in conventional forecasting methods that Santee Cooper aims to close. Smith explained on the call that traditional weather models rely on physics and perform well for one- to two-day forecasts, but their accuracy drops sharply beyond 48 hours as the calculations grow too complex. WeatherNext 2 uses an inference-based approach instead, which the utility expects will "dramatically improve" its ability to predict power load and avoid costly emergency purchases when demand spikes or dips unexpectedly. Better medium-range forecasts mean the utility can plan generation and purchases more efficiently, cutting waste and trimming expenses tied to last-minute power procurement during extreme weather.
Santee Cooper didn't immediately share the cost of the new tools or specify exactly which processes will be automated. Wilson said the utility is taking a cautious approach to the rollout, ensuring that employees using the AI systems receive training on handling confidential and restricted information. The goal is to blend automation with human judgment—letting machines handle routine tasks while people remain responsible for high-stakes decisions and risk management.