AI in the Utility Industry

AI in the Utility Industry

Utility 2.0

How AI can transform everything from utilities’ customer engagement to their operations

It’s a funny punchline because it’s true: The goal of most utility executives is to lead the second-most innovative company in the industry.

To be fair, we should all be happy that utilities are more risk-averse than most companies in other industries. When your fundamental mission is to literally keep the lights on and provide the power modern society needs to function, dabbling with untested technologies and business models can have profoundly bad outcomes.

But anyone who has watched a football or basketball game recently knows that artificial intelligence (AI) has long since crossed the chasm from promising idea to foundational business tool. Indeed, all those ads about how AI is transforming industries from logistics to healthcare to professional sports reflects the critical role companies see AI playing in driving efficiency, competitiveness and customer value.

Despite its ubiquity and hype, not everyone understands what AI actually means. At the biggest-picture level, AI is simply the capacity of machines and computers to mimic human behavior. Underneath that big umbrella definition, though, are machine learning technologies and sophisticated algorithms that help machines and computers work smarter and more effectively than us mere mortals.

It’s why AI’s ability to identify trends and anomalies in huge data sets is such a potentially powerful tool for detecting diseases. What’s even more exciting is that AI thrives on data: As the volume of data gets larger, AI’s ability to translate that mountain of information into meaningful insights gets better.

The simple fact that utilities are now evaluating and testing AI to help transform their operations, customer relationships and business models is a testament to how mainstream it has become.

For example, led by the Electric Power Research Institute (EPRI), the industry’s premier research organization, a broad swath of utilities has been investigating the use of drones and AI to improve the inspection of transmission and distribution assets. The underlying rationale is simple: Unmanned drones equipped with cameras can collect massive troves of images of conductors, transformers and other equipment that can then be analyzed quickly and thoroughly by AI. There is already a broad agreement in the industry that this approach has the potential to identify equipment at risk of failure in a manner that is much faster and safer than a current method that relies on manual inspections.

While transmission and distribution infrastructure inspections may be the most well-known application of AI in the utility industry, it’s hardly the only one. AI has the potential to manage the already- large and accelerating influx of distributed energy resources like solar and battery storage, which has triggered a substantial increase in the bi-directional flow of energy and the creation of a whole new class of so-called prosumers – residents and businesses that generate electricity themselves.

“Why not just do business-as-usual? I think utilities are starting to understand that, like a lot of other industries, if they do that their business could be affected in a significant way,” said Abhay Gupta, co-founder and chief executive officer at Bidgely, a software company that works with utilities around the world to harness the power of AI. “North America is currently, for the most part, a regulated environment. But there’s no guarantee that’s not going to change in the future. There are companies out there, particularly in the tech world, that would salivate at having the captive audience or the revenue stream that utilities have from consumers.”

Adding to those competitive pressures is the increasing expectation all consumers have around hyper-personalization, thanks mainly to their day-to-day experience with companies like Netflix, Amazon and Google. Fortunately, utility customers are largely receptive to improved engagement and support from utilities. Indeed, a 2017 Deloitte survey of residential utility customers found that a large majority of people ranked their utilities as the preferred provider of distributed energy resources, like solar. Importantly, the Deloitte survey also revealed an openness among residential customers to additional services from their utilities, including improved monitoring and management of their energy usage. It’s why Deloitte recommends utilities work to not only become more customer-centric, but also consider partnering with companies outside the industry to develop and deliver new data analytics products and services.

Thanks to AI, utilities are enviably positioned to take advantage of customer interest in deeper engagement. In fact, utilities already have what AI needs to thrive. “They have a mountain of customer data. For machine learning to operate or to be successful, it has to have a cache of data to learn against,” said Gupta. “This is an unnatural advantage that they have that they are currently not really exploiting or leveraging.”

But that lack of usage won’t last. Perhaps the most obvious area where utilities can marry their vast collections of data with AI is to improve customer engagement. It’s about taking a Netflix kind of approach to personalization to forge a mutually beneficial relationship with utility customers. And to deliver the equivalent of what Netflix does with personalized movie title and trailer suggestions in the utility world requires genuinely understanding how consumers use energy on the individual appliance level. This scenario is possible only through the energy disaggregation that can be achieved by applying machine learning algorithms to monthly and smart meter data reads.

The difference between an understanding of appliance by appliance energy use in a household and educated guesses about typical household energy usage that are based on neighboring homes is significant. And that difference is even more glaring when the information is used to determine not only which customers utilities seek to enroll in demand-side management programs but also the sorts of messages that are crafted to communicate with them.

“Utilities are still sending offers to upgrade a pool pump to people who don’t own pools,” said Gupta. “If you want to build a relationship with a customer and the customer’s current reference point is a highly personalized interaction, it’s a huge lost opportunity that can be remedied using AI on customer data.”

There’s plenty of reason to believe that an increased focus on customer engagement is good for business. A 2018 report by audit firm KPMG LLP argued that companies delivering the most personalized experiences to their customers ultimately reap higher revenue growth and increased brand standing and loyalty. KPMG’s top-ten ranking of companies providing the most personalized customer experience was led by Navy Federal Credit Union and included three grocery store chains. No utilities made the top ten.

But the consulting firm EY envisions a host of ways that AI can elevate the personalized experiences utility customers receive. The automation that AI delivers to customer service can let a utility know when they need to deliver more personal attention to a customer. “We could use AI to identify patterns of behavior that indicate customer dissatisfaction - perhaps tone of voice or choice of words or questions about energy usage or tariffs - enabling intervention and remediation to reduce frustration,” wrote Thierry Mortier, EY Global Innovation Lead for Power & Utilities, in a blog post.

The use cases for AI in the utility industry extend well beyond improved customer engagement and include improved regulatory compliance, better transmission and distribution planning, and new electrification opportunities that can bolster revenues.

In other words, AI can be a fuel that leads to a new kind of utility.

“A lot of utilities are looking to transition from selling kilowatt-hours to selling products and services,” said Gupta. “In as much as a utility can get more visibility about their consumer’s lifestyle and the profile of their usage and what’s important to them and motivates them, the more they can ensure that what they offer is valuable and useful to them.”

Our Electric Future

How electrification benefits customers and utilities – and how AI can help

There are few things in this complex, opinionated and messy world that can be categorized as universally positive. There’s a strong argument to be made that increased electrification is one.

Obviously, one of the primary beneficiaries of increased electrification would be the utilities supplying the electricity. The Electric Power Research Institute (EPRI) projects that efficiency gains will lead overall electric loads to decline in the absence of what it terms “efficient electrification” initiatives. But EPRI calculates that pursuing electrification would lead to cumulative load growth of between 24% and 52%.

But by no means would utilities be the only winners if electricity were more ambitiously harnessed to do everything from power cars, buses, heat pumps and warehouse equipment, or to grow crops in large indoor warehouses.

In short, a lot of good things would happen. Last April, EPRI released a National Electrification Assessment that outlined the societal, customer and utility impacts of electricity providing up to 50% of final energy consumption by 2050. The results: Increased grid efficiency and flexibility, improved human health thanks to better air quality, reduced energy consumption and consumer costs, as well as significantly lower greenhouse gas emissions, even in the absence of climate policy.

How does increased electrification deliver such wide-ranging benefits? The example of electric transportation is illustrative. Since 2000, the electric power sector has reduced its emissions of carbon dioxide by 20% and sliced criteria air pollutants by 80%. Which means that vehicles with batteries that are charged by the grid are becoming less damaging to air quality and the climate every year that the overall grid gets cleaner – which is every year.

But capitalizing on the promise and benefits of increased electrification requires new tools to help utilities robustly manage, optimize and ultimately monetize the opportunities that come from a modernized and transformed grid. Artificial intelligence (AI) is one of the most essential tools for utilities to embrace.

The use of AI to encourage and optimize the expected influx of electric vehicles provides a powerful example of how load-level information can be used to deliver benefits to EV drivers and utilities alike. Already, momentum around EV adoption is accelerating. According to Bloomberg New Energy Finance (BNEF), sales of EVs will increase from 1.1 million in 2017 to 11 million in 2025 and 30 million in 2030.

“A lot of that is being driven by the fact that EVs are becoming more and more affordable,” said Abhay Gupta, co-founder and chief executive officer at Bidgely. “Internal combustion engines will start to become more expensive than electric vehicles in short order.” In fact, BNEF research in 2018 found that lithium-ion battery pack costs averaged around $208 per kilowatt-hour in 2017. By 2030, predicts BNEF, the cost will have dropped to about $70 per kilowatt-hour. The same report projects that EVs will reach price parity with internal combustion powered cars by 2024.

AI can be used to accelerate the adoption of EVs and, once people have them, ensure that drivers and utilities are getting the most out of them financially. Here’s how it works: AI allows for energy disaggregation, which is simply a complex way of saying that it provides visibility into the fine details of the minute-by-minute energy usage of critical loads within a house. Visibility into EV charging – 80% of which takes place at home – provides information that can be leveraged to assist the homeowner, the utility and the grid as a whole.

One of the biggest problems is that the grid was never designed to have that high a capacity of demand at the same time. For example, if 10 homeowners on a street start charging their vehicles with fast charger at the same time, it is likely that we will surpass the capacity of the distribution transformer for that street. Every car manufacturer is about to release long range battery cars that even with fast charging would take 8-10 hours to charge overnight. This means that the option of staggering the car charging by a few hours would even not work for the grid as the adoption of EVs take on. The role of AI is critical at every stage. The first stage of growth is to identify which homes have EVs and help them move to time of use pricing. As the number and battery size of EVs increase - the role of AI may become arbitraging and forecasting who or how many homes in the street or on a transformer will charge when and combine with super dynamic pricing to offer super low pricing at times when there is excess capacity and super high pricing when you have higher demand. An example of this is Uber pricing - when demand is high, uber prices go up and users can decide to pay that price, or take alternate transport or wait for 30 minutes for demand to go down. This is just one proposed solution - there may be many other solutions - combining generation and battery storage with charging is another one that further complicates the equation of who can charge when.

For instance, AI can alert a utility that a home has an EV using a level one charger that typically begins charging a battery at 6 p.m. each night. If you know that as a utility, you now have the option to encourage that user to put a timer on their charging so they can charge at midnight when electricity is cheaper instead of 6 p.m. Or, you can even offer them a level two charger at a discount, which allows the utility to have control of the charger. Then the utility can send price signals or maybe even control that charger to help manage grid load.

These are all options that are financially advantageous to both utilities and EV owners, but they are options that utilities can’t confidently present to customers without the sort of granular information that AI provides. Utilizing AI-enabled visibility to offer EV-friendly rates or discounts on chargers helps utilities economically manage what can often be costly peak demand.

But it also is a way to ensure that utilities are relevant and resonant in their communication with customers – it’s the kind of data that utilities need to help foster everything from interest in EVs to participation in utility demand side management and energy efficiency programs.

By taking this information, you’re able to have targeted offers that make sense for people. It’s not just an offer. You become that trusted energy advisor that utilities want to be.

Happy Customers Equals Happy Utilities

How utilities can leverage AI for customer satisfaction and engagement

Every company in every industry cares deeply about cultivating highly satisfied and engaged customers. It’s the sort of statement that seems so obviously self-evident that it’s not worth uttering.

In the utility industry, though, the importance of customer satisfaction has not always been recognized. Case in point: A few years ago PricewaterhouseCoopers (PWC) released a research report titled “Beyond the hype: What is the value of customer satisfaction to a regulated utility?” Far from just assuming everybody knew the answer, PWC deemed it a question worthy of research.

PWC ultimately delivered a host of answers about why customer satisfaction should be a top priority for regulated utilities. For example, PWC’s researchers found that customer satisfaction is an important factor influencing the outcomes of regulatory initiatives. Though they didn’t claim a one-to-one correlation between high customer satisfaction and the approval of rate increases, PWC’s researchers did argue that a minimum level of customer satisfaction was vital.

As the energy system has undergone dramatic change in recent years with the influx of distributed energy resources (DER) and the potential threat of competition from non-utility energy providers, the importance of customer satisfaction and engagement has only increased. In fact, the PWC report also found that a focus on customer satisfaction is a way for utilities to protect their core business from disruptive entrants.

It’s also important to remember that utilities and the customers they serve do not exist in a vacuum. Consumer expectations about how their utility should engage with them are heavily influenced by their experiences with other companies, especially those delivering a fully digital experience like Netflix, Google, and Instagram, all three of which engage customers with highly personalized offers and instant responses.

Though utilities may be starting from a shortfall when it comes to engaging and satisfying customers, they also have significant advantages compared to companies in other industries. In particular, utilities have access to massive amounts of data about their customers - data that can be used to improve customer engagement and satisfaction.

And the recipe for improving customer satisfaction is not exactly a guarded secret. In fact, J.D. Power conducted research that highlights the importance of pursuing a strategy of proactive customer engagement. For these efforts to be successful, J.D. Power says they must be personalized, actionable and timely.

For utilities, the ability to do this rests on the robust, precise data and insights that artificial intelligence (AI) can deliver. “It’s the ability to process massive amounts of data in an intelligent way and learn from that data,” said Josh Gleason, Head of Product Marketing for Bidgely, a Mountain View, California-based software company that uses AI to help utilities improve customer satisfaction and engagement.

But what kinds of data can actually be translated into tangible actions that utilities can take in order to improve customer engagement and satisfaction? The answer to that question begins by being clear-eyed about how people actually engage with energy. While it’s true customers depend on the grid, the grid is at least one step removed from their actual experience with energy – a distinction that is very important if the aim is to genuinely understand what customers will value.

“Engagement at the consumer level and education at the consumer level has to start with appliances,” said Gleason. “People don’t interact with the grid, they interact with their light switch, and they interact with the start button on their washing machine. If you, as a utility, don’t know what those interactions are, how do you know how people are using your service? How can you personalize that? You can’t.” It’s similar to asking people to cut down their expenses even though they have no idea how much they spend on travel, groceries, eating out and other monthly bills. Only awareness and education about their individual circumstances can begin to change behavior.

AI provides a highly personalized view of those household interactions with different appliances through sophisticated energy disaggregation and customer segmentation. It’s an approach that leverages lessons learned from a few homes around behaviors and preferences and projected bill amounts and the propensity people have to call a utility’s customer service line.

It’s not easy to pull off. To achieve this appliance-level view of energy usage, data scientists must have massive amounts of data that can inform the algorithms they or computers write to identify which distinct usage patterns equate to an air conditioning unit, a pool pump or an electric heater, for example. While us humans tend to get less efficient at picking out patterns and performing well as the volume of information increases, AI is just the opposite. “It’s coming to conclusions humans cannot make because it’s processing so much data,” said Gleason. “And the beauty is that it gets more powerful and more accurate the more data it obtains.”

The more significant point here is that AI can provide utilities with the highly personalized, actionable and timely information they need to communicate with customers in a way that drives enhanced engagement and satisfaction.

Here’s what that looks like in action: One large utility recently used energy disaggregation data in bill alerts it sent digitally to customers. “We are sending everyone in the program what we call a mid-cycle alert that says, ‘Hi, John Doe, it looks like your bill is going to be $20 higher this month and here are the likely culprits,’” said Gleason.

In this particular scenario, not only was the alert sent about a bill that was projected to be higher than usual, it also includes actions customers can take to address the problem. Millions of these alerts have been sent, and over 90% of customer reviews about them have been positive. This is just one example of how AI can be used to improve customer engagement.

It’s also just the beginning. Utilities can leverage the power of AI in much the same way as Amazon and Netflix. For example, Netflix eschews the use of traditional demographics and instead replaces them with so-called “taste clusters.” These clusters are based on the actual viewing habits of Netflix customers, which becomes the data that allows the company to do such a good job of recommending movies and TV shows its customers might enjoy. Not only is the use of clusters scalable, it becomes more precise, relevant and powerful with every extra piece of information it can harness.

A similar clustering approach is possible in the utility industry. Data provided via clusters and energy disaggregation becomes like the viewing habits Netflix uses to deliver personalized suggestions. Utilities can use the data to not only present their customers with programs and information that is relevant and useful, but clustering also can help utilities present that information in a way that is more likely to elicit a response.

In a relatively short period of time, many utilities have made improving customer engagement and satisfaction a higher priority - don’t expect too many more research reports investigating the value of customer satisfaction to regulated utilities. It’s more likely, thanks to the improved use of utility data with AI, to see researchers probe this question: Which utility is most like Netflix?

Right Person, Right Message

The value of AI for optimizing utility programs

Smart thermostats, efficient appliances, distributed energy resources and other advanced technologies are rapidly transforming the global utility industry. Yet the approach many North American utilities take to get customers to sign-up for various demand side management (DSM) programs doesn’t take advantage of cutting-edge technology and data analytics.

“Right now, the approach is typically mass marketing,” said Matt Hale, a product manager at Bidgely, a Mountain View, California-based software company that works closely with utilities around the globe to better engage with their customers. “It’s billboards, radio ads and a fair amount of dabbling with social campaigns.”

Some utilities have also utilized home energy reports and online surveys as a lure to draw customers into DSM programs; the idea being that showing their customers how their energy consumption compares to their neighbors will motivate them to embrace programs that can help slice their bills. But even these techniques don’t have highly personalized data about a homeowner’s energy use and needs. That is only possible when a homeowner fills out an online survey, something that only around 10% to 15% of the people actually do. What about the vast majority of people who don’t fill out the surveys?

An even more fundamental challenge around the use of mass marketing approaches to get customers to sign up for DSM programs is that they often don’t accurately identify the customers who would benefit the most from increasing their energy efficiency. Of course, targeting the customers who will reap the biggest savings and benefits from the use of technology like a smart thermostat also helps utilities run more cost-effective programs, which opens up the possibility of expanding successful programs or innovating even more with new pilot projects and initiatives.

While it’s important to remember that targeting the right utility customers with the right offers can stretch marketing dollars a lot farther, that’s not really where you find significant savings. “The much more compelling argument is to stop wasting your rebate dollars on people who don’t save much energy,” said Hale. “Because marketing spend might be 10%, maybe 15% of program cost, but rebate spend is typically 70% to 80%.”

But highly-personalized and data-driven marketing simply hasn’t been the norm at most utilities. In a blog post for the consulting firm Accenture, Mark Sherwin argued that even though utilities were comparatively slow to embrace the benefits of data-driven marketing, doing so in the future was important to their success. “In fact, data-driven marketing holds the key to success across customer acquisition, retention and service,” wrote Sherwin, managing director of Accenture Digital. “This is because data-driven techniques enable utilities to profile the customer on an individual level, helping utilities better target their campaigns.”

Accenture research underscores the importance of being relevant to individual customers. According to the company, the typical customer devotes about 10 minutes of their attention per year to their utility. Taking advantage of that tiny window requires impactful communication.

But how can utilities more effectively target customers who can most benefit from DSM programs? One tool utilities now have at their disposal is artificial intelligence, or AI. Leveraging AI, utilities get a level of transparency and intelligence about the household energy usage of their customers – down to the individual appliance level – that has never been possible before. In other words, AI provides visibility into what appliances customers have, and how and when they use them. It’s highly individualized information that can be used to understand the savings a customer might reap in a utility program without them having to provide any information themselves.

Put another way, AI is the equivalent of sending out an army of energy auditors to do a deep dive into how all of a utility’s customers use energy.

These precise household and appliance-level insights allow utilities to more effectively target customers and encourage them to participate in DSM programs. It starts with where to focus their outreach and marketing efforts. Indeed, utilities know exactly the level of savings they want to achieve with each of their DSM programs. With that strategy in mind, tapping the power of AI to do a house-by-house and appliance-by-appliance analysis of which customers can achieve the highest potential savings by replacing an inefficient air conditioning unit or installing a smart thermostat provides a blueprint for marketing efforts.

Let’s say, for example, that a traditional utility program sets a goal of enrolling 1,000 homes in order to reach its savings goal. But with better targeting - and marketing messages tailored to individual homeowners - a utility would only need to sign up 700 homes. That’s because each of the targeted homes would reliably reap higher savings. This means that the cost of the program is smaller even though the overall savings are about the same.

There are plenty of reasons to believe that the targeted approach to DSM participation that is enabled by AI is set to increase. In part, it’s simply due to the fact that more utilities are looking to non-wires alternatives – including energy efficiency and customer-sited DER – as a tool to meet peak demand that doesn’t involve substantial investments in new substations and other infrastructure.

There is also an increasing body of evidence pointing towards the need of utilities to embrace a pay-for-performance (P4P) approach, particularly to their energy efficiency programs. A research paper released by the Energy Policy Institute at the University of Chicago examined what it termed “the conventional wisdom” about energy efficiency policies leading to investments that both pay for themselves financially and reduce greenhouse gas emissions. “However, this belief is primarily based on projections from engineering models,” wrote the paper’s authors, Meredith Fowlie, Michael Greenstone and Catherine Wolfram. In other words, the designers and implementers of energy efficiency programs could reach their goals based on assumed savings rather than actual meter data.

By examining around 30,000 Michigan homes enrolled in the Department of Energy’s (DOE) Weatherization Assistance Program, the researchers came to a very different conclusion. “The findings suggest that the upfront investment costs are about twice the actual energy savings,” wrote the authors. “Further, the model-projected savings are more than three times the actual savings.”

Findings such as these have elevated the interest among policymakers in adopting P4P metrics to evaluate energy efficiency program success. A report released by the Natural Resources Defense Council (NRDC) detailed the rationale for policymakers to pilot P4P energy efficiency programs. “The need to further ramp up EE to avoid greenhouse gas emissions from energy production, along with an interest in better use of digital energy meter data and analytics to encourage efficiency, has led policymakers in states like California and New York to consider expanding the use of pay-for-performance,” said the study’s authors. “P4P programs reward energy savings on an ongoing basis as energy savings occur, often by examining data from a building’s energy meters, rather than providing up-front payments to fund energy-saving measures.”

As utility program implementers adapt to the P4P approach, it will become critical to target their outreach to the customers who will achieve the most significant savings. That will increase the need to better understand individual customers. "Pay-for-performance approaches are mostly in pilot phase, but as the industry shifts toward that model, it’s going to be really critical to have that targeting,” said Hale.

An Expert in Every Home

Using AI to Improve Customer Support

Each year the consulting firm Deloitte surveys hundreds of companies about their customer service strategies and future plans. In its most recent survey of 450 customer contact leaders around the world Deloitte found that artificial intelligence is very much in their future plans: indeed, fully 56% of respondents said they have plans to invest in AI.

The reasons for this high level of interest in AI to help bolster customer service and support are many. One is cost. According to a recent study by LOMA, a trade association for the insurance and financial services industry, the average cost of training a call center employee is $7500. High turnover rates mean that companies have to make those investments frequently. A report by the Quality Assurance & Training Connection (QATC), a call center trade group, turnover among call center employees is twice as high as other industries. There are plenty of other reasons for the increasing use of AI in the form of chatbots and other tools. The software company Autodesk has implemented AI to aid customer service, a move that has improved its customer response time by 99% and lowered its per query cost from between $15 and $200 when handled by people to $1 when virtual customer service agents respond.

The combined power of AI, natural language processing, chatbots, smart speakers and other tools is now making it possible for utilities to achieve some of the same benefits other industries are seeing. At the same time, AI provides utility customers a profoundly knowledgeable energy advisor to answer their questions 24 hours per day. Just as people can currently ask their smart speakers about what’s on their daily schedule or to recommend a local restaurant, technology advances mean it’s now possible for those same devices to be savvy and always-available utility customer service representatives.

“You can ask questions like, ‘Why is my bill so high? How can I save? How can I lower my bill?’”

It’s just one example of how the possibilities around enriched utility customer support are advancing and improving. These enhancements provide utilities with important tools to not only improve customer service and satisfaction but also fend off competitive pressures, including customers generating their own electricity through the embrace of distributed energy resources (DER) like solar and batteries.

Competition isn’t the only thing that’s focused the attention of utilities on improving their customer service. Today’s consumers simply won’t give utilities – or anybody else – a pass on shoddy customer support. And in many ways, customers expect more from their utilities because they already have so much personal information available to them.

“Everyone uses Google, Amazon and these other companies who just offer incredible, personalized, effortless customer service,” said Matthew Hale, product manager at Bidgely, a Mountain View, California-based software company that helps utilities better engage and support customers. “When utilities aren’t giving me useful tips and advice, and personalized recommendations it’s surprising because they have all of my data.”

The changes are taking place thanks in large part to utilities’ increased use of AI. It’s the combination of AI and natural language processing and smart speakers that allows a savvy, automated energy advisor to take up residence in just about any home in America. While impressive, the fact that a utility customer can ask a smart speaker about a high bill and what can be done about it is just one of the breakthroughs. What’s arguably more powerful is that AI enables smart speakers to provide answers that are highly personalized and relevant.

How does it work? AI delivers granular visibility into a utility customer’s energy usage at the appliance level. Sophisticated algorithms help make sense of the data and translate it into information and offers that a utility can make to a customer that can help solve their problems. In other words, it’s like having a massive army of super smart data scientists constantly poring over every utility customer’s energy usage.

“What if you could have someone look at everyone’s bill every month and see who got a high bill? Then you could look into their consumption data, look into their consumption patterns, look at what you know about that home and try to figure out what caused that high bill,” said Hale. “Using AI, we can run that process automatically for every home.”

It’s easy to see how this kind of information can benefit a utility’s customers and also significantly improve customer support. While important, awareness and education about a customer’s energy usage is merely one aspect of customer support. Equally critical is reaching out to customers proactively when they are on track for a high bill and also presenting them with options that can help them do something about it. This approach can take place via email before a customer has even noticed a high bill. “Alerting a customer that they’re on track for a higher than normal energy bill is just the start. It’s also important to provide tangible actions they can take to avoid that high bill, either this month or for next time. It’s about putting the customer back in control,” said Hale.

This type of proactive customer support – either through an email or through an opt-in text message – can solve a lot of issues before anybody is motivated to reach out to a utility call center. In fact, proactive, high bill alerts have been shown to reduce high-bill related calls into a call center by 50%, which is a significant cost savings for utilities.

Even in cases when customers need to speak with a customer service representative, AI can help a utility arm those reps with the information they need to offer helpful insights and recommendations. “They have the insights that AI is recommending for the customer,” said Hale. “They don’t have to walk through a script or try to analyze raw consumption data, which is hard to do, and can be frustrating for both people on the call.”

Tapping AI to reduce call center costs is compelling. "I think the call center cost reduction is pretty real and the utility can probably justify investing based on that, said Hale. "But the customer satisfaction benefit is probably the more compelling one in the long run, and it comes from offering a more modern, up-to-date experience.”

The Power of AI to Democratize Energy Savings

Better data visibility can help ensure utility programs make an impact across the economic spectrum

A recent study by the Environmental Defense Fund (EDF) put hard numbers to a conclusion many utilities and regulators have already reached: the potential savings offered by utility-funded energy efficiency programs are not reaching the customers who could benefit the most.

Indeed, although the grid as a whole is becoming increasingly efficient, the ever-expanding array of energy efficiency programs and improvements available to utility customers simply aren’t reaching those who spend a disproportionate percentage of their income to pay their monthly bills.

According to the study’s authors, the 36 million-plus American families with incomes below twice the federal government’s poverty level – just under $50,000 for a family of four – use over 30% of U.S. residential electricity. Despite that, only 6% of all energy efficiency spending is dedicated to low-income programs.

This is a huge missed opportunity for utilities, low-income families and society. EDF calculates that the right menu of policies and approaches could slice the energy use of low-income families by 20%, delivering hundreds of dollars of savings each year per family and a total reduction in electricity expenditures of $7.4 billion. That also equates to the elimination of 48 million tons of carbon dioxide emissions.

Though many obstacles have to be overcome to achieve these benefits, one important challenge the authors highlight is the lack of comprehensive data – historically, at least – on everything from utility customer demographics and energy usage to program statistics.

The Value of a Data-driven Approach

This dynamic is beginning to change for two very powerful reasons: an uptick in regulators considering rules or implementing programs designed to ensure that all utility customers benefit from programs promoting efficiency and distributed energy resources (DER), and the power of data to help utilities achieve those compliance targets producing more compelling results. For example, California has long promoted the availability of solar through its Single-family Affordable Solar Homes (SASH) program, which provides generous incentives to encourage the installation of photovoltaic (PV) panels on low-income homes. New York has followed suit as part of its ambitious Reforming the Energy Vision (REV), which has made expanding access to renewable energy and energy efficiency to all New Yorkers a cornerstone of its efforts. Federally, the Low Income Housing Energy Assistance Program (LIHEAP) helps fund weatherization efforts and solar installations that reduce energy costs.

Research released last year by the Smart Energy Consumer Collaborative (SECC) underscores why incentives and programs are not enough by themselves to expand the benefits of energy efficiency and renewable energy to all utility customers. SECC researchers found that low-income households tended to know less about energy efficiency measures than the general population and were less likely to respond to utility outreach efforts. There is a significant need for personalized education and outreach, though utilities have not traditionally leveraged the kind of data that can make those efforts a success.

“Data has not been a focus or in the forefront of energy efficiency program planning, design, and implementation,” said Jordana Temlock, director of regulatory affairs for Bidgely. “It hasn’t been a priority because its value has been misunderstood within the traditional program paradigm, where utilities and their partners have been content with marketing the same rebate programs to everyone to be able to hit their program goals.”

While this approach can be effective for achieving program goals, it tends to favor higher-income households with large energy bills that could produce greater savings. It also helps explain why Home Energy Reports (HERs) from utilities have typically only been sent to homes with the highest bills. The return on the HERs could be much higher if targeted based on data that shows which homes have the highest inefficient usages instead of targeting just based on consumption.

Connecting Utility Programs with Customers

The combination of data and the emergence of sophisticated artificial intelligence (AI) algorithms can provide granular, appliance level insights making it far easier for utilities to democratize the participation in and benefits of customer funded utility programs. A host of companies have recognized the value of using data to make these connections. Bidgely’s energy disaggregation is the most sophisticated available today. Here’s how it works:

Through patented AI products and services, utilities are able to disaggregate the energy usage in a home appliance-by-appliance and load-by-load. The insights produced from this data provides utilities with information on who to target for a given program but also how to engage with customers in ways more likely to resonate with them. For example, this kind of data could help utilities identify a low- or moderate-income home with an extremely inefficient HVAC system; the customer could then be approached with a tailored offer to replace the system through a utility program that would fund the swap.

“As an AI company partnering with utilities, Bidgely produces disaggregated appliance-level insights so that utilities can personalize and target efficiency programs to really help with program uptake,” said Temlock. “It helps the utility understand the consumers and how they're using energy. From that point, you figure out which program might be the most useful for them and craft the engagement going forward.”

In other words, it’s all about utilities using hyper-specific information about household energy usage to guide their outreach and messaging to customers, ensuring that all customers can take advantage of utility incentives and programs that would be most appropriate for them. This is already possible when utilities use data and AI to examine how customers have used energy in the past, helping the utility to formulate a plan for customers to use it more efficiently in the future.

Another example are so-called seasonal alerts that enable utilities to tell a customer who had exceptionally high energy usage in the winter or summer about programs to upgrade their HVAC system. “Sending out seasonal alerts is a great way to get people to be thankful that their utility is looking out for them,” said Temlock. “It’s a really great way of marrying proactive, insightful and personalized messages to influence customer behavior.”

Potential Impact on Ratemaking

When data leads to improved uptake of programs across customer segments, utilities are able to comply with their regulatory obligations. The importance of granular data about household energy usage becomes even more important as regulators ponder the use of performance-based regulations.

In recent years, the traditional cost-of-service ratemaking approach, in which utilities earn profits based largely on capital expenditures, has come under greater scrutiny for disadvantaging DERs. By contrast, regulators in states such as Hawaii, Minnesota and Rhode Island have been analyzing performance-based regulations that incentivize utilities to meet customer demands for lower-cost, more sustainable and more efficient energy.

Insights delivered by AI and data could become a critical tool for achieving compliance of performance based metrics as customer engagement and satisfaction become included in this approach to ratemaking. “It’s going to vary state-by-state, but data and AI should be part of the solution and strategy for helping utilities meet some of these evolving performance-based ratemaking efforts,” said Temlock. “It can be difficult to meet these targets and to prove you’re meeting these targets to regulators. But if you have information and visibility, it makes it that much easier to comply.”