Generative AI for energy and utilities: 5 surprising facts

Fact No. 1: GenAI can help deliver near-term value

GenAI can offer far more “quick wins” than traditional AI — largely because you don't build your own model. You leverage a foundation model, use GenAI embedded in enterprise software, or, most commonly, both. You do need to stand up usage policies and other guardrails, get some data in shape and provide upskilling. But with a number of use cases, you can accomplish that relatively quickly.

For utilities, for example, GenAI can transform call and digital contact centers. It can automate responses to routine customer needs. For more complex needs, GenAI can connect customers to the right specialist — and provide that specialist with insights and guidance. GenAI can also analyze call center data to help improve operations, reducing average handle time (AHT), abandonment and transferred or misrouted calls. That can add up to lower costs and happier customers.

And this is just the beginning. GenAI can then act as the glue between operations and maintenance and asset information systems, so that customer interaction, field operations and asset information can be brought to bear more synchronously and with greater impact. GenAI can be used to unlock the “language” of smart meter data as well as aid field operations in planning asset maintenance or upgrades.

GenAI is like many other technologies in that you can start small, achieve cost savings or revenue growth, then use your “winnings” to fund next steps.

Fact No. 2: You don’t need to complete data modernization to get started

If you don’t have the data or cloud infrastructure to pursue GenAI, don't worry. You can still get value from GenAI, if you select use cases in part based on data: Pick a pilot for which you already have (or can quickly gain) access to the necessary clean, well-organized, compliant data.

If you already have, for example, Internet of Things (IoT) sensors on key assets — such as pipelines, wells or transmission lines — GenAI can track asset health, help correct performance deviation and drive better-informed decisions. As GenAI makes these data sets more useful and useable, you can use this value to win buy-in and budget for initiatives to accelerate your cloud modernization journey.

GenAI also offers a bonus: It can speed up data modernization — and help cut costs — since it can often read, analyze and summarize even highly unstructured data. In other words, GenAI isn’t just a data consumer. It can help organize data for you.

Fact No. 3: You can scale quickly — if you choose

GenAI is remarkably scalable. With traditional AI, each new use case usually requires building a new model. But a single GenAI model — with appropriate customization — can address multiple use cases in multiple functions. In our experience at PwC (and where we have over 3,000 identified internal use cases), roughly 80% of GenAI use cases fall into one of six patterns. Each can scale rapidly.

As one example, once you have GenAI conducting “deep retrieval” (extracting data and insights from documents) on contracts, it can do the same for requests for proposals (RFPs), invoices, customer communications, financial reports, tax regulations and more. This potentially could help to streamline legal language related to utility easements and rights-of-way. Add other GenAI patterns, like summarization, and you can increase overall back-office productivity by 20% to 40%.

To help achieve scale safely, at PwC we (and many of our clients) use an “AI factory,” an operating model designed to help set priorities, share and allocate resources, and support rigorous oversight and governance.

Fact No. 4: GenAI can help reduce your risks

Poor decisions or operational failures could put vital energy supplies or even your employees’ safety at risk. Here too, GenAI can be an ally. It can help your people analyze more data to better anticipate supply interruptions, prevent accidents and more. GenAI’s boost to back-office capacity can help reduce risks too. For example, it could help review more contracts in more detail, looking for potentially troublesome clauses.

GenAI can also help fight fraud. A well-trained GenAI model, with reliable data pipelines, could analyze (for example) data from customer meters. It could spot anomalies, generate reports, follow escalation protocols, and help human specialists take corrective action. Energy trading desks can benefit too: GenAI can provide your traders with easy access to load forecast data, weather data, consumption data, as well as analysis of their past trades. That can help reduce the risk of poorly timed trades.

For GenAI to reduce risks — without creating new ones — its outputs need to be reliable, relevant and compliant, with privacy protected and bias minimized. Responsible AI, based on a tested toolkit, that covers the full AI life cycle and includes specialized GenAI governance, can help you design trust into GenAI from day one.

Fact No. 5: You may not need to hire a lot of new talent

The talent exists to help you tap into GenAI, if not within your organization, then with the right third party. Since you don’t need to build a GenAI model, you probably won’t need to bid for scarce AI talent. Your cloud provider or AI vendor does that. That makes many GenAI projects feasible and affordable, whereas traditional AI often isn’t.

Yet GenAI isn’t plug-and-play — not if you want high-value, risk-managed outputs. At PwC we have been upskilling our workforce on principles of responsible AI, effective GenAI prompting and more. You may need to do the same, in stages. You may also need to cross-skill some of your current technology team to help customize and oversee GenAI. In the end, you still may need to make some new hires — although, most likely, only a few.

It won’t just be your workforce that may need upskilling to use GenAI well. You and your fellow industry leaders may also need to learn how to use GenAI well — to help you make better decisions and adjust long-term plans. Eventually, GenAI could make your knowledge workers 20% to 50% more productive. It might be wise to start considering the likely impact on operations, your workforce strategy and more.