What happened
Utility Dive reports that Santee Cooper is working with Google to modernize forecasting and financial planning. The state-owned South Carolina power and water utility plans to use Google’s WeatherNext 2 for medium-range weather forecasting, a custom model tuned to local conditions, and Gemini Enterprise Agent Platform for financial-planning workflows. The utility also plans to deploy several hundred Gemini Enterprise licenses, while retaining human responsibility for accountability, risk management and oversight.
Utility Dive reports that Santee Cooper announced the Google partnership on Sept. 3, 2026, following comments by CFO Tami Wilson during a press teleconference. The utility is using Google’s WeatherNext 2 to improve medium-range weather forecasts and load forecasting, while Google’s Gemini Enterprise Agent Platform is being used to speed financial forecasting.
According to Utility Dive, Google is helping Santee Cooper build a custom forecasting model that accounts for local weather factors, including possible microclimate effects from Lake Marion and Lake Moultrie. Wilson said the utility previously relied on traditional meteorological models and is pursuing a broader digital transformation that automates routine work while leaving accountability and risk management with human operators.
Utility Dive reports that Santee Cooper distinguishes between generative AI, which it plans to use for low-cost initial drafting, and agentic AI, which it intends to use for autonomous multi-step processes. Google said the agent platform can support scenario-based financial planning connected to calculations for major projects, including Santee Cooper’s $10 billion grid-expansion budget. The utility is also deploying several hundred Gemini Enterprise licenses across the organization.
Utility Dive reports that WeatherNext 2 provides probabilistic forecasts for lead times of up to 15 days. Wilson said a one-degree forecast difference can shift the utility’s power position by about 100 megawatts per hour and that emergency power procurement during extreme conditions could cost up to $100,000 per hour. These figures and the potential for millions of dollars in annual savings are claims reported from the utility, not independently verified results.
Source details: utilitydive.com ↗
Why it matters
The reported deployment connects AI forecasting to decisions involving grid reliability and a $10 billion expansion budget. More accurate medium-range weather and load forecasts could help Santee Cooper plan electricity purchases and operations, but the article does not independently establish that the tools will deliver the claimed savings or accuracy gains.
The reported use case places AI inside operational planning for a public utility rather than limiting it to office-productivity tasks. Forecasts influence how much electricity a utility expects to need and whether it must procure power under time pressure, making accuracy and uncertainty important for reliability and costs.
Santee Cooper’s stated rollout also illustrates a division of labor between generative AI and agentic AI. Drafting assistance generally leaves a person to review an output, while autonomous multi-step processes can affect several connected systems or decisions. That difference makes access controls, auditability and human approval boundaries practically significant.
The article gives concrete stakes but not independent validation. Utility Dive reports the utility’s estimates about forecast sensitivity, emergency procurement costs and possible savings, while the source does not provide comparative test results, a baseline error rate, an audited savings figure or evidence that the custom model has improved operations in production.
What to watch next
Watch for evidence from Santee Cooper on forecast accuracy, emergency-power costs, realized savings, the specific workflows automated by agents and how confidential data is governed. Utility Dive reports that the cost of the tools and the exact automated processes were not disclosed.
Santee Cooper did not provide Utility Dive with the cost of Google’s tools or the precise processes that will be automated. Those details will determine whether the deployment is a limited pilot, a broad internal rollout or a larger operational change.
The utility said employees involved in the rollout will receive training on confidential and restricted data. Future reporting should examine what data can be sent to the systems, retention and access controls, human approval requirements and whether agent actions are logged and reversible.
The most useful follow-up evidence would include out-of-sample forecast performance against Santee Cooper’s traditional meteorological models, changes in load-forecast error, emergency-power purchases and independently documented savings. The source does not establish general public access to any of these tools; the reported use is an internal enterprise deployment with several hundred licenses.