in association with
1. Utilities are nervous about AI as tighter regulation looms.
The head of operations at an energy retailer says they are already using natural language processing with telephony and looking at using generative AI to create customer emails. “But there is a nervousness,” they admit. “We have a mix of people who are risk-averse and others who are excited by anything new and shiny.
“When you are a customer-facing business, it is easy for things to go wrong and get shared widely, so we are treading with caution.”
The head of digital at another foreign-owned energy retailer expects EU regulations and strict governance standards to apply to its use of AI. “We want to encourage colleagues to use AI, but what are the guiderails? What are the principles?”
Across the world, governments and regulators are asking themselves these same questions. The EU passed the Artificial Intelligence Act in March, while regions like the US and China are concentrating on their own regulatory frameworks to address ethical concerns, privacy issues, and the potential misuse of AI.
There’s movement too in energy, with Ofgem recently closing a call for input from the industry on the impact of AI back here in the UK.
2. AI is only as good as the data it is trained on…
The former digital lead at a water company overseas explained some anglophone large language models (LLMs) struggled to successfully decode strongly accented French. “They could throw up all kinds of weird results, so you need to consider the model you use, because it is only as good as the data you are training your library on. You can’t just go to the LLM everyone uses and hope it will give you the right answers.”
AI may also fail to detect verbal irony or sarcasm, which attendees pointed out is a frequent feature of customer comments – especially on social media.
The water company digital lead explains: “People talk about AI as if it’s a homogenous tool, but you need different tools for different things. So, I suspect we’re going to end up with a utility-specific LLM.”
3. …and that data needs to be in great shape.
Before implementing digital solutions, you need to “understand why you do the things you do, closely followed by sorting your data,” says the former water sector digital guru. The energy company operations director points out that in an ideal world the business should have an in-depth understanding of a customer’s billing cycle, propensity to contact, and whether they use a smart meter. They add: “That’s the dream, but we need to make sure this background data is perfect before we let the AI do its stuff.
“I don’t think we’ll ever be in the position where we are so comfortable the data’s good that we don’t need someone there, reviewing what the AI is doing. It may be the case that rather than writing an email or social post, the agent is the reviewer. But they still need to know what brilliant looks like.”
Staying with the theme of data quality, a consultant instrumental in developing an online electricity and gas service for business cautions there is a danger of a vicious circle developing where “you haven’t got the confidence that the data is right to be able to enable AI to come in and tidy it up, and as a result the data is getting worse”.
4. There isn’t a one size fits all solution when it comes to AI.
“Every supplier wants something different, every supplier thinks they know better, and every supplier is trying to build something for the worst-case scenario.” That was the view of the consultant on our panel. He points out that a digital platform targeted at half-hourly market settlements and billing was inappropriate for his needs. “It slowed everything down, and we were all about speed.”
He adds that the power of AI should be in enabling patterns to be spotted that might otherwise be missed. “If a business customer hasn’t cancelled their direct debit, for example, we might not take on board that their consumption has dropped by 90% – no one looks at it, because we are still being paid.
“But that might be a sign that the business has left the site or ramped down production, and the chances are that the company is moving or going bust – something fundamental.”
5. AI has the potential to disappear altogether within applications…
The digital expert at the energy retailer says: “I think AI will go native: it will be inside applications, and you won’t know. A board member will open a lovely dashboard that has everything they need, and you will have no idea it was generated by AI. Or a customer query will come in and an email will be generated for you: as an agent, you will look at that email and say ‘yes / no’, and you won’t think.”
A former wholesale leader at a water company adds that they foresee AI ultimately fixing problems and responding to customers without human intervention, or even awareness. “Errors will be fixed without people knowing about it,” they believe. The head of digital believes this eventual ‘invisibility’ of AI is analogous to debates about computer processing power 20 years ago.
“Nobody talks about that anymore. All of our laptops have more power than anyone needs. Similarly, people won’t realise that something is generated or created by AI.”
6. …but that won’t happen yet.
The energy retailer digital specialist says that although it is “not currently mandated” that content generated by AI is flagged as such at their company, they expect that to change. “Internally, it is recommended. It’s my personal choice and I flag it up when I use AI to generate content for our intranet.
“It’s also important that code that is generated by AI is flagged as such when it is uploaded. The developer – the individual uploading it to the system – should be accountable for it.”
7. Specific customer journeys require special care.
There are some customer journeys that shouldn’t be automated – or only automated very carefully. “Bereavement is a journey across all energy suppliers that you just don’t mess with,” points out one panellist. “Because it if it goes wrong, it’s on the front page of the Daily Mail.
“It’s the story where someone says, ‘I’ve lost my husband’, and the AI says, ‘I hope you find him soon’.”
“The AI might be like, ‘have you looked under the bed?’” adds another attendee. “And it’s not a hallucination, it’s a legitimate response.”
But our consultant pointed out that AI also has the potential to improve processes such as this by learning from experience. “The bereavement process from 40 years ago is probably not suitable for today. But there are also probably decisions made 40 years ago that have not been undone.
“That is where AI would constantly learn and update.”
8. AI could make customer service much more personal.
“If you were designing the market from scratch,” says the former water sector digital leader, “you wouldn’t build the market we have. It’s like they say about the camel being a ‘horse built by committee’. It’s the same with the energy market.”
He adds: “If you were starting again, you wouldn’t think about half hourly or daily settlements, you’d think about the customer expectation, and tailoring the journey to them. And while it’s impossible to go through each call they have made, AI has the potential to plough through all that data and understand it.
“So, I think the potential for extreme levels of personalisation is huge. Some businesses are already good at that. But there’s a long way to go.”
So, panellists at the Utility Week Intelligence / EXL event agreed AI is already transforming utility customer service. But if the next phase of the technology’s development is to live up to expectations (and not bump along the bottom of Gartner’s Trough of Disillusionment), a combination of regulatory rigour, high-quality, well-structured data, the right tools for the job, and a cautious approach to sensitive customer journeys will all be needed.
Get those elements right, and the hype could be justified.
“The potential of AI for the energy and water industries is massive,” concludes Mitch Donnelly, head of utilities at EXL. “It could transform customers’ experience of utilities in a way that’s different to anything we’ve seen.
“It’s time to embrace the opportunity.”
in association with