As gas utilities look at how to bring the power of artificial intelligence into their operations, customer service is showing the way.
That’s because the insights made possible by AI automation make it easy to demystify the complexities of natural gas for a public not steeped in the regulatory affairs and spot pricing behind their monthly bills.
“Customers want to understand their energy usage,” explained Christopher Swartz, senior group product manager in customer systems at San Diego Gas & Electric. “They want to understand how to save money on their bills. Having that level of engagement and being able to properly analyze customers’ accounts and provide them with options or solutions is why the customer space is a great place for automation.”
SDG&E is among the natural gas utilities showcasing the possibilities of automation and AI, from analytics to message delivery. The results include not only satisfied customers, but also satisfied employees who are empowered with the tools and time to deliver customer experiences that delight and inform.
E-commerce has attuned today’s consumers to expect personalized, friction-free service. SDG&E strives to exceed those expectations.
Changing customer expectations occupy “the heart of everything we’re doing” to embrace AI and automation across all SDG&E customer service channels, said Swartz. “SDG&E is serving as the trusted energy adviser, giving the customer the information they need to make decisions in this space,” he said.
SDG&E began its AI journey by partnering with AWS Cognitive Services to centralize data in a secure, interactive customer cloud. As a result, time-consuming analyses once drawn from scattered systems can be completed with lightning speed and precision targeting.
Now, SDG&E has the capability to detect patterns and historical usage, perhaps revealing previously overlooked customers who might be suited for a money-saving program or billing options. “It’s about using the information we have in our controlled centralized platform and being able to leverage that information,” said Swartz.
The process also intersected with the work of SDG&E’s IT innovation team, which immediately began exploring proofs of concept when AI’s commercial viability became known. Ideas offered to Swartz’s customer service technology team pointed toward “a lot of great opportunities,” he said. “We started realizing we have some great use cases that might fit well.”
Each use case involves a “take it slow” approach. First, data derived from analytics must show potential for driving value for sizeable numbers of customers. “When ideas come in and we work with different teams, we might see one issue that’s generating 4,000 calls a month and another that’s bringing in 400,” Swartz said. “We’re going to focus our time on the 4,000. The data gives us the ability to make those decisions versus having to estimate and figure out which one makes the most sense.”
Then, concepts are tested internally and, if the results justify, are carefully rolled out. With each “small chunk,” as Swartz calls them, customer feedback is collected, and tweaks are made to continuously improve the customer experience. Each step must justify further investment of time and money on the way to scaling up.
“This is part of our DNA now,” said Swartz. “We’ve had such great success that we’re continuing this approach.”
SDG&E applied four leading AI technologies to extract telling insights from data and improve the efficiency and effectiveness of the customer experience.
Natural language analytics: Working with the SDG&E Customer Care Center, Swartz’s team applied natural language analytics to transcribe and summarize hundreds of thousands of customer phone calls. Reviewing these transcripts line by line, the program teased out customer intent and sentiment. It also spotted agent phrasing that elicited common customer reactions, from delighted and comprehending to frustrated and confused.
Customer intent emerges from themes threading through the calls—for example, tens of thousands of customers asking about net metering for solar energy. “We then can use that data to design other technology to meet customers where they need it,” said Swartz.
Optical character recognition: The ability to scan written documents through OCR is well established, but incorporating AI escalated SDG&E’s ability to efficiently process large volumes of low-income program recertifications.
The former system of manual processing was slowing down the discounts that low-income customers needed, in part because handwritten documents are often peppered with anomalies that boggle old-style scanners. An illegible signature or an acknowledgement box marked with an X instead of a checkmark could send an application to a human for time-consuming review.
Under the new system, the OCR software and associated AI have scanned thousands of applications, learning to spot the shared meaning behind differing markings. “The model has been trained to recognize that these are good, acceptable answers,” said Swartz. “It always keeps learning and keeps improving.”
At its launch, the system automatically handled about 80% of applications. Within about five months, the rate rose to 97%, saving hundreds of hours of processing every week.
Customer notifications: AI-enabled analytics revealed that low-income customers frequently call about the status of their applications for assistance. That insight led in December 2023 to the first round of AI-inspired text messages that give customers updates on their applications.
In keeping with the “small chunks” approach, those initial outbound notifications were simple, but SDG&E envisions a buildout “to make it more of a proactive approach with our customers versus reactive,” said Swartz. “If we see from the analytics that these types of notifications are beneficial for customers, we want to invest there.”
Examples could include notifying customers who might benefit from budget billing or reminding them about money-saving rate options. The team is exploring the possibility of reducing call-ins by crafting notifications with built-in self-service options or aligned with information posted on the SDG&E website.
Virtual assistant: Before AI, call center agents had a feeling that a substantial number of calls were coming from customers seeking help with their password resets in SDG&E’s online My Account system. Analytics revealed the hard truth—customers were indeed struggling with the steps needed to reset or recover lost passwords.
The large numbers made those customers an ideal target for SDG&E’s new virtual assistant, which initially focused on their specific dilemma. Since a planned rollout in January, customers experiencing issues with their passwords online have been proactively prompted by a virtual assistant with a list of choices and step-by-step instructions.
Within the first half of 2024, the virtual assistant will expand to include responses to the frequently asked questions revealed through data analytics of incoming requests. If large numbers of customers are asking about a certain topic, the virtual assistant will be loaded with predefined responses on that subject. From their My Account pages or the website, customers can click on the virtual assistant widget for a direct connection to the exact information they need.
When customers experience the benefits of automation, employees enjoy the ripple effects. For some employees, the AI-assisted OCR frees time to address problematic applications for low-income assistance, instead of wading through piles of handwritten forms to spot trouble areas.
As the virtual assistant was pilot-tested internally, staff gained instant access to information about SDG&E initiatives and programs. AI can even summarize regulatory filings into trusted, easily digestible information for employees to absorb quickly and apply during customer interactions. “It’s a tool for employee empowerment and education,” said Swartz. “This will enable them to have the information they need to do their jobs.”
The insights gained through natural language analytics are also being infused into agent training and scripts to minimize frustration and bridge gaps in customers’ understanding.
Then, as AI-driven initiatives expand gradually and demonstrate success, the customer service unit can generate examples to share enterprise-wide. “That’s where the true innovation comes into play,” said Swartz. “Now that we have functioning processes, we can show our colleagues how it works. They start thinking they have a process that might be a great fit. Someone sees the capabilities and says, ‘We have a great use case for this.’”
The work is exciting and the potential limitless.
“Obviously, you have to start small, but it’s really exciting to see employees embrace the technology and to see that innovation from a use-case perspective and where we can make things better for our customers,” said Swartz. “It’s been really good for our team all across the company to start seeing this new technology and the problems and challenges it can solve.”
As customer service grows increasingly intricate and sophisticated, CS Week empowers utility professionals to deliver consistent, customized excellence with every customer contact.
Attendees will gain tools to help solve the “CX Puzzle,” the interlocking series of CS functions that shape and define the customer experience. CS Week Conference 48 delivers educational content and experiences designed to elevate attendees’ knowledge of each piece: billing and payments, contact centers, credit and collections, digital engagement, disruptors, field services, sponsor solutions, and strategies and analytics. This year, a new puzzle piece charts the pathways to leadership development.
Other highlights include the annual Women in Utilities Sunrise Event, elective Deep Dives into key areas, the Expanding Excellence Awards, and exhibits spotlighting the technology and products essential for meaningful customer connections.
CS Week Conference 48 takes place April 30 to May 2 at the Fort Worth Convention Center. For registration and details, visit csweek.org.