See our views on AI in energy and utilities featured in Utility Dive here

"There are many different technologies and use cases, and it is important to be specific about what the technology is we're discussing," said David Groarke, managing director at Indigo Advisory Group. “Utilities have been using AI for years, in terms of machine learning."

"You can already see some of the technology in the home," said Groarke. There is a level of AI in home energy management at play, and it will improve. A learning thermostat can identify comfort levels, but the next step will be combining that with the ability to keep bills lower. "All of your devices will be making decisions around prices and tariffs," he said.

Where utility use of AI will take off first, said Groarke, is in data centers and with commercial and industrial consumers.

Google, now Alphabet, acquired artificial intelligence-focused company DeepMind in 2014 and turned the system on itself in an effort to reduce costs. Last year, the company revealed it had slashed its data center cooling bill by a whopping 40%. Now, DeepMind is working with the UK's National Grid to help make the country's electric system more efficient.

DeepMind CEO Demis Hassabis told the Financial Times earlier this year that a double-digit demand reduction was possible--with no additional infrastructure. "It would be amazing if you could save 10% of the country’s energy usage without any new infrastructure, just from optimization,” he told the paper.

How is that possible? Data centers have levels of redundency built in for cooling and backup power, and large systems face inefficiencies as well. Reserve power resources are ubiquitous across power grids, says IEEE Senior Member Shawn Chandler, whether for backup or renewable integration. "AI can be used to better predict these changes, and craft a response that is tuned to expected system behavior over time," he explained here.

The extra energy used for cooling, the standby generation on the grid, all of it can be better optimized.

"Redundancy is a huge buffer," said Groarke. "But if you have the right algorithms, you can effectively figure out the minimum requirements of both systems."

But right now there is a tendency for service providers to describe an even wider swath of technologies as AI. Groarke said that in the last 12 months, companies have latched onto AI as a branding opportunity.