What the rise of voice assistants, chatbots and other AI technology means to improving service and reducing bills

Using AI technology to ensure interactions with customers are simple, intuitive and engaging is a hot topic for water and wastewater company Welsh Water.

The not-for-profit water utility is placing customer involvement and innovative communication at the heart of its business strategy. Morgan Lloyd, head of marketing at the company, says: “Welsh Water, like many utilities, are looking at how we can develop bots and AI to give our customers a better and quicker service.

“In the short to medium term, this is likely to focus on using automation to allow customers to do the simple transactional tasks – from submitting a meter reading through a Facebook Messenger chatbot to finding answers to the more ‘transactional’ queries like checking your bill balance or getting information on water supply interruptions more quickly.”

Like many firms in the sector, Welsh Water is beginning to harness so-called machine learning technology, software that can crunch through vast amounts of data to ‘learn’ more human responses based on patterns of the data.

While observers say AI has been slow to enter the utilities world, it is now waking up to the potential for AI technologies, including predictive analytics, chatbots and voice assistants.

Welsh Water’s foray into AI has seen it develop a Facebook Messenger chatbot to engage with customers. In 2017, the company launched a consultation, called Have Your Say, which gathered the views of thousands of people to help the business plan for the future and improve services for customers. The campaign’s key focus was to build customer trust.

Encouraging customer involvement in utility consultations can be notoriously challenging, so the company knew it had to take a different approach. The Facebook Messenger chatbot offered an opportunity to deliver the consultation in a quick and interesting way.

Alongside a bespoke website and various events during Wales’ summer festivals, the chatbot was one of the new ways the company hoped to reach out to customers. The bot, developed with digital agency Coup Media, allowed Welsh Water to engage with a potentially larger audience and reach demographic groups that are traditionally less likely to engage. Customer engagement surpassed the company’s expectations, with more than 2,500 people taking part through the chatbot and a further 12,500 through the website within just the first few weeks of the campaign.

“There are FMCG [fast moving consumer goods] companies that are using artificial intelligence to predict consumer trends and accelerate the development to market of successful products. Utilities should be looking at the same techniques to identify the consumer disposition for new bundled products and services,” says Toby Siddall, managing director and UKI utilities lead at Accenture.

“This will require best-in-class capabilities in customer analytics, customer insight and service design.”

As well as helping companies identify new services and predict customer trends, AI can help utilities drive behavioural change as they seek to help their customers better engage with their energy consumption and increasingly help them find the right tariffs and services.

Sandra Schroeter, senior international product marketing manager, customer engagement and support at LogMeIn, explains: “We see companies using AI chatbots in customer service a lot. In the utility space, a chatbot can evolve further by integrating with data from smart meters, for example, and would allow customers to ask the chatbot questions about their usage in real time.

“AI can also help agents to be more efficient by monitoring live chats and suggesting answers to the agent, eliminating or minimising the time required for agents to search for answers.

“That said, there will still be a need for human agents, whether that is to deal with more complex customer enquiries; serving customers who don’t want to interact with AI; or handling emotional or high-value interactions that companies don’t want to leave to the AI.” This is a theme we will return to later.

Voice automation

Linking with voice assistants is seen as another avenue for optimising energy consumption.

Octopus Energy, which recently acquired AI and machine learning capabilities from failed supplier Usio, has partnered with Amazon Alexa to allow customers to benefit from real-time energy pricing using voice automation.

“Our mission is to make buying energy as simple as buying cornflakes, in an energy market riddled with complexity and customer confusion,” says Greg Jackson, founder and chief executive of Octopus Energy.

“The rise of voice-enabled technology is fascinating, and we wanted to be an early adopter in understanding how that technology can support in the drive to engage consumers in their energy.”

The partnership allows consumers to use Alexa to adjust energy usage based on half-hourly price changes offered by Agile Octopus, a smart time-of-use tariff. Agile Octopus users can directly engage with their energy by asking Alexa a range of questions about their energy use, such as when electricity is cheapest or more expensive, and plan accordingly – saving money and reducing carbon emissions.

Alexa is enabled in more than 100 million devices globally that manage all facets of the smart home, including lights, door locks, heaters and more, allowing customers to seamlessly marry Octopus-provided rates with a range of smart home capabilities.

“The Alexa relationship also builds on a prior integration we built with smart device pioneers If This Then That (IFTTT), where any IFTTT-enabled hardware can be linked to Agile Octopus to turn on when cheapest,” adds Jackson.

Allowing customers to use voice controls to conduct a range of actions, from optimising home heating to charging an electric vehicle, is just the start of where the technology could develop. Octopus is currently collaborating with a number of start-ups and technology companies taking different approaches to driving a smarter energy future. “Fresh innovation that shifts demand away from peak times and maximises flexibility is central to building a smarter grid, and the businesses that can unleash that will build the market of the future,” says Jackson.

Chatbots versus real people

For utilities, it will be essential to manage customer expectations while the sector’s use of the technology remains in its infancy. Susannah Richardson, director for field service and contact centre solutions at IFS, admits that a major barrier to AI technology at this stage is its self-learning capability. Current forms of AI need large quantities of data for algorithms to learn, with an average 200 variations of data needed for AI to seamlessly answer a customer’s request without review. Consequently, there is a bottleneck around the amount of data needed for each specific use case.

“We recommend that you don’t expect the bots to answer every question or resolve all the requests, instead analyse your most common use cases or frequently asked questions and train your bot to resolve these, while seamlessly handing over more unusual or specific questions to a human agent,” says Richardson.

The real evolution of the chatbot will be its ability to employ improved self-learning capability to optimise itself. When the bot needs to transfer to an agent, it copies and learns next time how to recognise, respond and process this request automatically. “This is possible today, but with improved machine learning this will become more and more powerful,” says Richardson.

As an enterprise software business, IFS’s solutions aim to help companies offer more connected services for their customers. By incorporating data from different front and back office systems, the IFS Customer Engagement (CE) platform operates as a centralised interface to manage requests from utilities customers. The platform uses AI chatbots and natural language processing (NLP) to offer self-service options and automatically retrieve the necessary information to address the customer’s need.

Welsh Water’s Lloyd agrees; the company hopes that its fresh approach to customer engagement and innovations such as the chatbot will continue to increase customer trust. But despite successfully using AI to facilitate the consultation, the technology is not without its limitations, he says: “The challenge is that customers expect these tools to provide better service than a human agent.”

Customers increasingly expect AI-enabled utilities services to match and even exceed their experiences with other sectors that are early adopters of the technology, whether it is chat boxes on retail websites or smart recommendations on social media platforms.

AI can undoubtedly help customer service centres reduce incoming enquiries as people increasingly resolve simple requests for themselves. But simply reducing the number of customer service staff should not be the ultimate goal for businesses. “We might be able to employ AI to reduce the number of agents processing address and subscription changes, but we still need to invest in customer service staff to deal with complex, emotional issues and differentiate our brand,” says Richardson.

“Today’s smartphone empowered customer is more intelligent, they are more informed as they research issues, they are more complex, emotional and social. As a business we need to respond by ensuring that we have humans to provide empathy and have the time to resolve complex issues and go that extra mile to ensure that the customer is not just satisfied but a social advocate.”

Field marshals

Beyond using AI to improve engagement with customers, there is scope for the technology to help utilities deliver more responsive repairs in the field.

While the adoption of AI in the field is in the early stages of development, there are already a few common use cases. “The simplest is to help customers help themselves: before they make that call to the support centre, a chatbot or online triage employing NLP can be used to diagnose and frequently resolve the issue,” says Susannah Richardson, director for field service and contact centre solutions at IFS (pictured). “Research shows that customers would much rather use self-service to resolve most issues if given the option.”

If an issue remains unresolved, or the customer chooses to go straight to the support centre, the agent is empowered with the knowledge base and tools to diagnose and remotely resolve the issue, potentially negating the need for an engineer. “In the background, AI can be employed to increase the accuracy and probability of resolution, analysing the likely failure modes for a specific asset,” says Richardson.

If the fault requires a field technician to repair the issue, then the data from the triage, plus data on the in-field asset, can be analysed to ensure that the technician dispatched has the appropriate boot stock to provide a first-time fix, she adds.

This article first appeared in Flex, issue 3. Read the full issue of Flex here