As utilities begin to explore multiple use cases for GenAI, the lowest-hanging fruit and most obvious use cases are related to the customer experience. GenAI isn’t quite ready to engage in unmediated customer-facing touchpoints, but now is the time to simultaneously improve the customer and employee experience by empowering and supporting customer service agents with GenAI's revolutionary capabilities.
GenAI for Customer Support
Eventually, most simple customer support interactions will be fielded by GenAI-enabled chatbots. As these supercharged chatbots answer customer queries, they will lower utilities’ cost-to-serve and improve customer experience by delivering clear, fluid automated responses in real time.
For now, utilities are not ready to experiment with using GenAI models to communicate directly with customers. However, utilities are already becoming comfortable with providing GenAI tools to their customer service agents. GenAI can help agents on two fronts: AgentGPT (e.g., helping agents craft emails and other customer-facing communications) and agent search tools (e.g., helping agents find and understand assets like training material and standard operating procedures). GenAI outputs can be exact and meaningful in both contexts, enabling utilities to realize benefits like improved staff efficiency and reduced training times.
AgentGPTs are much easier to implement than full-blown CustomerGPTs. In an AgentGPT scenario, a GenAI-powered chatbot prepares an answer to a customer query. A human agent then quickly reviews this answer before passing it along to the customer. Human agents play a failsafe role: When a GenAI tool “hallucinates” or seriously malfunctions, the human operator can step in and prevent that information from reaching customers. For utilities that lack complete confidence in their GenAI models or are midway through elevating their data ecosystems, this approach can safely advance a scalable, value-additive GenAI use case.
GenAI tools can also help agents locate information and perform their jobs more efficiently. Instead of searching for documentation in complicated nested repositories, agents can pose questions to a GenAI tool with a single click. With a single action, an agent can have a comprehensive library of services, products, SOPs, and troubleshooting best practices at their fingertips. Pairing human customer service agents with properly trained GenAI tools makes their jobs more accessible and enjoyable while speeding up their ability to solve customer problems. When well-supported customer service agents are able to spend their time addressing the more complex and challenging customer service needs, the overall customer experience improves by leaps and bounds.
Following a recent use case discussion of a GenAI tool for agents, one of Wipro’s utilities clients found that they save time during customer conversations and also realize other benefits. They are finding that agent GenAI tools will allow them to reduce the time they spend training agents and that GenAI engines can also automate post-call documentation. While agents may only spend one or two minutes documenting each call, removing this workload from the customer support function will create tangible cost savings and performance improvements.
Use case discussions like this have convinced us that utilities should go full steam ahead using GenAI for indirect customer and agent support. They’ll need to move more slowly when rolling out customer-facing GenAI tools, but these agent support projects will allow them to build their underlying GenAI capabilities. Within a year or two, as the GenAI ecosystem matures, we expect leading utilities to be comfortable putting GenAI-enabled chatbots in front of customers.
Guardrails for GenAI Tools
Agent-facing GenAI tools are inherently lower-risk than customer-facing GenAI tools. Even so, all GenAI tools require appropriate guardrails, particularly around safety/security and domain-specificity. Security guardrails are critical. For GenAI tools trained on customer support materials, the risks of data leakage may be moderate. The risks become more significant as GenAI tools access customer information to summarize customer conversations.
In addition to security risks, utilities need guardrails to ensure the domain-specificity of GenAI responses. Given the physical realities of utilities products like natural gas and electricity, some customer support questions may have significant safety and regulatory implications. If a GenAI tool presents inappropriate information from other domains in its responses, the negative consequences can be substantial. Domain-specificity can be achieved by training standard GenAI large language models on domain-specific content, including pre-defined policies, guidelines, and safety rules.
Utilities Agent Support: A Springboard for GenAI Excellence
With GenAI, utilities can reimagine operations, grid management, asset management, renewable energy integration, forecasting, and planning. All of these use cases are worth exploring, and the best way to get a head start is to begin implementing GenAI in the context of agent support. The AgentGPT/agent support context is ideal because it’s low-risk and high-impact, and will deliver concrete benefits to both employees and end customers. Once utilities are comfortable with this GenAI use case, they can solve the more complicated legal, regulatory, and technical challenges, enabling GenAI ROI in other critical business areas.