Machine learning in the energy sector is not a new concept. The grid’s medium and high voltage networks already rely on algorithms to recognise patterns and predict outages.
Behind the meter, however, there is less maturity, says David Groarke, Managing Director of New York-based Indigo Advisory Group, which researches emerging technologies in energy.
The drawing of inference from large data sets allowing “a machine to perform at the level of a human expert” - a well used definition of artificial intelligence (AI) - is in its infancy in customer-facing applications.
But the potential of combining machine learning with natural language processing and pattern recognition to improve customer experience and the customer-utility relationship is significant.
An infographic on Artificial Intelligence in Energy and Utilities produced by Indigo for example, highlights how energy companies could apply AI and game theory to customer recruitment for a demand response pool through reward and penalty mechanisms.
The suggestion here is that AI-enabled customer applications can help to win trust and therefore unlock consumer flexibility.
Groarke believes the full adoption of AI in energy is about five to ten years away but he encourages energy companies to “touch” this new technology and try it out while it continues to evolve.
A key finding, however, is not the consumer distrust of AI but the utility’s. Groarke believes there are “three buckets to be addressed” in the utility industry for AI to succeed in customer applications.
Energy consumption algorithms - remove bias
The first is to remove bias from training data for AI algorithms. Groarke explains: “I was at a conference recently where the head of an energy storage company said ‘we’re not just building fancy boxes for dentists, we want to appeal to all classes of customers’.”
The same is true for AI in customer energy solutions. “Not everyone can afford a Nest smart thermostat so consumption patterns should be based on the broadest datasets available to make it equitable.”
AI and energy decision-making - where is the trust?
Energy companies seeking to win consumer trust with AI-enabled energy and customer services will also need to consider transparency.
Groarke explains: “Some customers will want to know how the algorithm arrived at a particular decision. Explanations of why energy consumption may have changed year on year are better than automation at building trust. You can’t just set it and forget it.”
An example of a transparent tech-focused energy supplier is US startup Drift. Describing itself as a “power company built on modern technology”, Drift claims to make “energy easier and more efficient for users by introducing intelligent software and automation to the energy market.”
The company, which launched in New York in Q2 2017, is buying power from its network of peer-to-peer energy producers and selling it to end users.
Drift says it uses artificial intelligence and machine learning “to determine how much energy will be needed the next day by factoring in things like the daily energy usage of individual customers over time and zip code by zip code microclimate weather data estimates.”
From a consumer perspective, Drift claims to offer savings on energy bills of up to 20% in exchange for a $1 weekly subscription fee to cover costs.
Customer-centric solutions - how AI can help
Drift’s concept is also a good example of Groarke’s third point about achieving success with AI in energy retail - the need for customer-focused design of products and services.
In the start-up’s case, the offering is built around putting “your power in your control”.
In another context, more established energy suppliers should develop AI-enabled services based on actual customer behaviour, such as how they make decisions. This will allow machines to better help consumers and subsequently automate the process, says Groarke.
One tool in the customer-experience AI toolkit is conversational interfaces or chatbots.
Software and solution provider Ferranti Computer Systems has developed a cognitive customer service assistant called MIA as part of its MECOMS solution.
In partnership with Microsoft Cognitive Services, MIA can reportedly analyse language patterns and keywords to detect when customers are unhappy or frustrated.
Leslie De Cuyper, Product Manager MECOMS at Ferranti Computer Systems, told Engerati that he used design thinking when developing MIA. “I considered what I liked and disliked about interacting with companies. One of the first things that came to mind was how much I disliked searching through FAQs for an answer to one of my questions.
“Although our smartphones’ screens and mobile websites are now almost as good as their desktop siblings, browsing on the move is still far less efficient than quickly asking your phone a question and have it come back with exactly the information you need.”
De Cuyper foresees chatbots like MIA helping to build consumers’ digital trust. He says: “The fact that a bot is less ‘personal’ than talking to an actual person will lower the barrier for customers to approach the company with questions, and getting answers in turn lowers insecurity and doubt.
“Customers are far more likely to trust information coming from a bot as they are not perceived as likely to lie to get rid of you and they have infinite patience with no ‘bad days’.”
Getting consistent information when you need it and through the channel of your preference, is the ultimate catalyst for trust, says De Cuyper.
Retail future: Slovenian energy supplier Gen-I explains how it has adopted MIA as part of its customer care solution.
In a recent blog, he also highlights the positive impact a chatbot can have on an energy company’s cost to serve by allowing it to scale the size of the business without doing the same to the call centre.
One benefit of an emerging technology is it is “all to play for in the residential and industrial markets,” concludes Groarke who cites only 10-15 vendors offering targeted energy-related AI products. And the question of trust? “It takes time to build consumer trust,” he says. “It will come with the incremental adoption of new technology.”
Artificial intelligence - webinar
David Groarke and Leslie De Cuyper will both present in the Engerati webinar 'Emerging technologies: Artificial intelligence in energy retail'.