Generative AI — the power behind ChatGPT and similar applications — is sparking a huge wave of interest, discussion, and scrutiny across utilities companies.

It’s not hard to see why. Utilities are going through the largest transformation they’ve seen in 50 years—perhaps ever. And business leaders have so many different irons in the fire it can be hard to keep up.

Consider the fact that utilities are actively driving the energy transition, while also optimizing their costs, maximizing profitability, improving capital allocation, enhancing customer experience, and managing a looming talent gap. The list goes on.

So, when a technology comes along that can support that transformation, help solve those challenges and accelerate so many core industry functions, it’s no great mystery why utilities leaders are keen to understand the potential.

Generative AI’s moment

Many of us have seen ChatGPT in action on our phones – or even our children’s phones! But it’s the large language models (LLMs) underpinning these applications that contain the real power.

What makes LLMs special is the fact they’re both very powerful and infinitely adaptable. So they can be applied to solve a whole range of business problems and customer pain points.

That includes everything from managing vast reams of documentation in areas such as regulatory filing, permitting and compliance to enabling customer self-service at scale to updating engineering specifications and accelerating green power.

This technology has the potential to truly reinvent how the industry operates, how employees work, and how customers are served.

In fact, Accenture’s analysis has shown that over a third of all working hours across utilities (37%) stand to benefit from either automation or augmentation through generative AI. And in some parts of the business the proportion will be much higher, including customer service (67%), human resources (66%) and finance (65%).

It’s not just traditionally office-based roles either. The same analysis finds over a third (38%) of electrical engineers’ working hours will be similarly impacted. This will likely be seen in activities such as developing plans, creating designs, monitoring operations, and preparing reports.  

Accenture has already helped one US west coast utility build a secure web platform powered by Azure’s GPT models. We educated employees about the technology and tested the generative AI application across 16 business groups with more than 50 users, developing 25 use cases to help iterate and improve the platform as it rolls out across the enterprise.

Early value on the table

Let’s consider some of the key use cases that can deliver value today.

Retail customer service is an obvious one. Companies can use generative AI to power a virtual agent that can support call transcription and summarization, detect customer sentiment, provide dynamic real-time suggestions to agents, and write semi-automatic follow up emails in everyday language.

When you consider the volume of customer interactions utilities deal with every day, the potential for efficiencies, productivity savings and operational cost reductions is significant, while also boosting customer satisfaction.

Naturally, this will be one of the first use cases to scale. And we’re already seeing examples of utilities within the retail space that have supplemented and augmented their contact centers and related customer interactions, as well as marketing and sales, with generative AI.

Creating new insights at scale

Other quick wins are likely to involve activities that can benefit from generative AI’s ability to deliver rapid insights at scale, but which don’t require significant changes to day-to-day operations.

That might include, for example, using generative AI to augment workforce capabilities across generation, transmission, and distribution. The technology’s ability to consolidate multi-modal knowledge embodied in instruction manuals and past interventions can assist field technicians as they carry out repairs or maintenance of equipment.

In this way, equipment downtime can be reduced, and asset utilization improved massively. It would also allow expertise on newer equipment to be rapidly democratized across the workforce.

Another good example involves power purchase agreements (PPAs). Imagine using generative AI to ask questions about a PPA portfolio – such as which clauses are driving profitability across customer types and time horizons, how a new regulation will impact the portfolio, or explain the broader market context at the time a certain clause was negotiated with a customer.

An LLM can significantly reduce the time taken to access this kind of information, linking it with market context and other relevant information to generate hyper-relevant insights. 

Similarly, the technology can be potentially invaluable when it comes to analyzing and summarizing regulatory changes. Applied to large volumes of regulatory data, it can allow utilities to better prepare their own operational systems, as well as third-party systems, to ensure compliance and support regulatory proceedings.

Permitting to win

That said, one of the use cases I’m personally most excited about is permitting and consenting. Not least because it’s critical to achieving net zero.

Whether it’s nuclear or renewables, transmission or distribution, a big part of what's happening with the energy transition is around capital build. All of which must be developed at an unprecedented pace within the required budget to meet net-zero goals.

One of the key bottlenecks in developing and commissioning these capital projects is permitting. In fact, by some accounts almost 80% of planned wind projects in the US are stuck in the permitting phase. And the EU has four-times more wind capacity in permitting than under construction.

Consider that permitting typically involves wading through thousands of pages of documentation and then developing drafts for submission to the relevant approving authority. That’s a perfect use case for generative AI.

We see interesting applications being implemented helping employees manage documentation and deal with consents, as well as interconnection and bidding requests.

Permitting isn’t the only way generative AI can support the energy transition.

The number of parts and components required in a large capital build such as a power plant, can be astronomical. And the complexity of the supply chain activity needed to get those components to the right part of the build at the right time from the right supplier equally so.

Imagine if you could use generative AI to process all that data and create insights to understand, organize and optimize all that activity — supplier/vendor management, contract management, spend management, service management, and so on. The potential for massively improved supply chain management — and all the associated benefits that come with it — is vast.

Clusters for extra value with Gen AI

Realizing some of the value on offer will need the industry to come together. That might entail pooling data or spreading the costs needed for fine-tuning and customizing an LLM for a particular use case.

Permitting, discussed above, is a great example. But there are others, such as asset management and predictive maintenance.

Understanding equipment failure rates is a key challenge right now for individual utilities, whether it’s wires, pipes, cables, transformers, smart meters, or any of the other hundreds of thousands of pieces of equipment they need to run their operations.

Because of the variety of different manufacturers, makes and models involved, most utilities don't have enough information on their own to properly predict the failure rates of these assets. But if they were to pool this data across the industry, they would do. And that would open up the possibility of using generative AI to create statistically relevant insights.

Time to get started

Let’s recognize that this is bigger than just a new piece of technology. It’s an opportunity to reinvent the enterprise — in talent, in the operating model, in governance, in the organizational structure, in legal and risk management, and more.

Nevertheless, there’s value on the table right now. And a willingness to dive in and experiment in a controlled, secure fashion will be critical to ensuring utilities aren’t left behind.

Let's switch on the insight and embrace the momentum of generative AI.