The utilities aren’t alone in seeing the potential of generative AI in customer service. A recent cross-industry survey of more than 1,000 businesses found that nearly half of organizations are planning to use generative AI for sales and customer service such as optimizing support chatbots and boosting self-service capabilities. Among utilities and energy companies, that figure was higher yet – 52%. What’s more, 39% of energy and utility companies surveyed had established a dedicated team and budget to integrate generative AI into product and service development plans.

There are grounds for such enthusiasm. Unlike existing utilities-focused AI-assisted functions such as predictive maintenance, generative AI will complement and augment existing customer service approaches without the need to reconsider fundamental business processes. AI-driven solutions can speed response times and cut customer service workloads. They can also help steer customers – and, when necessary, customer service representatives (CSRs) – to the right answers, quickly.

Those AI-assisted answers contain the seeds of lower operational costs and higher revenue through tailored cross-selling and upselling products and services such as heat pumps and energy audits. Generative AI will be able to reconcile the minutiae of utilities’ and partners’ vast product and service offerings with real-time insights into the specific consumption patterns and demographics of inquiring customers. That will open doors to improved service and better and more profitable customer relationships. That, at least, is the vision.

But how might utilities best go about bringing generative AI into the customer-service fold? Here are a few considerations for those starting down that road.

1. Choose your target. Generative AI will ultimately handle a wide variety of customer service tasks. But you have to start somewhere. Perhaps that somewhere is the management of billing complaints. Or maybe you go with appliance rebates or tariff recommendations. Pick a target and pilot it.

2. Think data. Choosing a target means narrowing the focus. On the data front, do the opposite. The more data a generative AI system has available, the better it can be. Lay the data groundwork early. Among the data sources potentially in play include meter-to-cash-related data, operational data, network-related mapping data, demand-response data, SCADA data and smart-meter data.

3. Get the data tools. Utilities live in many places. Generative AI wants it all together. That takes an analytical strategy to aggregate diverse data for the 360-degree view AI needs. Combining diverse data sources for generative AI then requires data-management solutions to handle the data retrieval and aggregation. Such solutions must include safeguards for data privacy and compliance.

4. Go real-time. Customer acceptance of generative AI chatbots will vary based on demographics, technological predilections and other factors. But even digital natives most likely to embrace them will have high expectations. The limitations of today’s chatbots will likely fuel initial skepticism of generative AI versions, so the data they work from must be timely and accurate – whether the topic is order status, usage forecasting, or billing-related. In-memory computing is a good way to ensure that queries tap real-time data.

5. Keep your LLM options open. Different large-language models have different strengths. Your platform should be able to work with a variety of LLMs to ensure the best fit.

6. Test responses and establish guardrails. Building generative AI for utilities’ customer service involves a collaborative effort combining the strengths of solution providers with those of the utilities who know their customers and business goals best. The process involves guiding AI toward the proper responses to the real-world questions, as they’re likely to be asked and ensuring that the answers are accurate based on the underlying data sources. It can be a painstaking process, but it’s one that’s critical: A utility can ill afford to have its customer-service AI hallucinate the answers to billing-related or outage-related inquiries, among others.

7. Consider the human role. People will remain vital to utilities’ customer service for the foreseeable future, but the human role will evolve. One envisions AI handling increasingly complex inquiries independent of human oversight, but also enabling higher-level decision-making among lower-level customer-service staff. Operationally, the interactions and handoffs between AI and human users must be delineated, and there are obvious human-resource implications in terms of onboarding and training.

Utilities and their solution providers have a long way to climb on the generative AI learning curve. Given the technology’s tantalizing potential, it would be unwise to bet against a rapid ascent. Utilities – and their customers – simply have too much to gain.