utility customer service: every channel went digital, satisfaction did not move

Digital channels now manage more than half of utility customer interactions. The migration to self-service has already occurred. But it has not led to a long-term measurable improvement in customer experience, as measured by net promoter score, customer effort score, satisfaction, loyalty, retention, and lifetime value.

Yet industry benchmarks indicate these metrics have all remained flat for three consecutive years despite sustained digital investment. The lack of significant change can be attributed to energy providers digitising the channels without redesigning the model. Consequently, the gap is widening between what customers expect—shaped by the personalisation standard set by streaming, e-commerce, and digital search—and what utilities actually deliver.

The deployment of AI in utilities has the potential to change this. Research by National Grid Partners reveals that AI has moved to the top of the agenda, with 96% of utility industry leaders identifying it as a strategic imperative. The opportunity requires a service model built on proactive, predictive, and personalised engagement.

The utility service model is evolving

The trajectory from reactive call handling to AI-led customer intelligence has four discernible stages.

  • The traditional call centre resolves most contacts through human agents, with limited AI or self-service.
  • The contact centre handles customer care across digital and human channels, with AI managing a substantial share of routine contact.
  • The experience management hub focuses on proactive service with most interactions AI-enabled.
  • The insight powerhouse, which is the most advanced stage, anticipates customer needs, with the large majority of interactions managed through AI, supported by specialist human agents for the cases that genuinely require them.

Most energy providers are still in the earlier stages of this evolution.

How is AI improving utility customer operations?

The transformation AI makes possible in utility customer operations touches both the service interaction itself and the intelligence available to every participant in it. These transformations include the following:

Agentic AI and real-time resolution

Generative AI creates content and insight from data, while agentic AI acts on that intelligence. Virtual agents that execute tasks under human oversight convert resolution timelines from weeks to hours. Complaints that previously required manual escalation through multiple teams can be handled by an ecosystem of AI agents, with human review only at certain defined points.

A significant performance differential can be observed between AI-assisted and conventional service operations. In a controlled study of more than 5,000 customer support agents, access to a generative AI assistant raised issues resolved per hour by 14% on average, and by 34% among the least experienced agents. This capability is most directly responsible for the satisfaction gap between utilities and the consumer brands customers use as their reference point.

Smart meter support and energy intelligence

Smart meter data, read at 15- or 30-minute intervals, generates a volume of consumption information no human support team can process at the individual customer level. Energy disaggregation techniques apply machine learning to interval meter data to infer household-level appliance usage, giving utility customer operations an appliance-by-appliance picture of how each customer uses electricity, without requiring any input from the customer themselves.

This intelligence is the foundation of modern smart meter customer support. A customer whose bill is trending upward can receive a proactive alert mid-cycle that identifies the likely appliance or behavioural cause and presents corrective actions. This naturally deflects inbound call volume. AI-driven smart meters are also being integrated with distributed energy resources, enabling real-time demand and supply balancing and providing the data foundation for time-of-use tariff guidance, EV charging optimisation, and demand response programme participation at the individual account level.

The proactive model: from outreach to insight

Build AI-enabled utility customer operations | Modernise your energy operations

Build AI-enabled utility customer operations | Modernise your energy operations

The shift from reactive service to proactive engagement requires reframing customer service as a value driver rather than a cost line, and investing accordingly. Utilities moving toward the insight powerhouse stage use AI to recognise opportunities continuously, reducing energy costs, deepening product adoption, and reaching segments that mass outreach programmes have historically underserved.
One dimension of the available operational value is AI-driven field productivity improvement. Another is personalised demand-side management outreach targeted at households whose consumption profiles make them most likely to benefit.

Building the workforce alongside AI

AI in utility customer operations does not reduce the importance of skilled human agents. It changes what skilled means. As AI handles routine enquiries, billing explanations, smart meter queries, and standard complaint pathways, customer service representatives become available for the interactions that require judgement, empathy, and authority. The capabilities that matter in this model are data literacy, analytical thinking, and empathy-driven communication, not script adherence.

Energy providers that approach workforce development as a parallel programme to AI deployment consistently outperform those that treat it as a downstream step. Electricity demand in the US is expected to grow by 120 GW by 2030, and residential electricity prices have already risen over 13% since 2022. The service organisations that hope to handle this growth sustainably must scale AI capability and workforce capability together.

How can Infosys BPM help organisations leverage AI in utilities and energy sectors?

Infosys BPM energy and utilities services help energy providers redesign customer operations for the proactive service model. Infosys Topaz-powered AI platforms and smart meter analytics transform utility customer operations. We help enterprises combine AI in utilities deployment, smart meter support operations, and customer analytics to grow along the service maturity spectrum toward genuinely anticipatory customer engagement.

Frequently asked questions

Energy disaggregation applies machine learning to interval meter readings to estimate which appliances are driving a household's consumption, without any hardware in the home or input from the customer. It matters for service because it turns a single bill figure into an explanation: the agent, or the automated alert, can point to the likely cause of a rise instead of simply confirming the amount.

Three things, usually in this order: interval meter data that is complete and reliable enough to model consumption; a customer record that joins billing, contact history and premises data to the same account; and interaction data from digital channels. AI deployed on top of fragmented records tends to automate the existing experience rather than improve it.

Through the same service model, with a narrower set of inputs. Billing history, tariff, property characteristics, payment behaviour and contact history still support proactive outreach, segmentation and automated resolution, just without appliance-level insight. Treating unmetered customers as out of scope creates a two-tier experience precisely among the households least likely to have opted into digital channels.

Yes. Interval consumption data reveals occupancy and daily routine, so it is treated as personal data in most jurisdictions and its use for anything beyond billing generally needs a defined lawful basis, clear customer notice, and retention limits. Utilities building personalisation on meter data should settle the consent and governance model before scaling the use case, not after.

Start with the highest-volume, lowest-judgement contact reasons (billing explanations, meter reads, payment arrangements) and fix the data behind them before adding automation. Deflecting those contacts creates the capacity and the evidence needed to fund the next stage. Attempting predictive, proactive engagement before basic resolution works reliably usually produces outreach customers do not trust.