A customer whose house heating stops working on a winter night calls the gas or electricity provider, and what happens on that call is the utility customer experience for that customer. Before hanging up, the customer expects to know whether they can fix the fault over the phone; if not, they want to know when a technician will arrive.
When looking for ways to improve utility customer experience with AI agents, you’d look for ways to answer every call without a queue and resolve the fault on the same call. An AI agent holds the conversation with the customer and completes the task the call creates, such as booking a technician in the utility’s field service system.
The ten ways in this guide follow the order in which a customer deals with their utility, starting with the outage alert sent before anyone calls. Way 4 covers the fault call that needs a technician and gets the most detail, because the customer’s problem remains unsolved until that visit is booked.
What Is Utility Customer Experience?
Cognizant defines utility customer experience as how satisfied a customer is with every interaction they have with their utility provider, whether that is a gas, electricity, or water company, from paying a bill online to reporting an outage by phone.
Utility customer experience covers the self-service channels and the phone line together, and an AI agent can work across both. Connected to the utility’s CRM, the agent that answers the phone can see what the customer already reported online.
From the customer’s first message to the update in the utility’s systems
Three customer experience trends in utilities show why customers and regulators watch out for each of those interactions closely:
- Rising prices: The U.S. Energy Information Administration expects the average residential electricity price to rise from 17.3 to 18.2 cents per kilowatt-hour between 2025 and 2026.
- Customers expect a working app: J.D. Power’s 2026 study of 46 large U.S. utilities scored customer satisfaction with each digital channel on four factors: information, navigation, speed, and visual appeal. Mobile apps scored 654 out of 1,000, ahead of desktop websites at 617 and mobile websites at 591, yet 28% of those utilities still offer no app.
- Regulators pay for satisfaction: Ofgem, Great Britain’s energy regulator, rewards or penalizes each electricity distribution company by up to 1% of its base revenue based on customer satisfaction surveys.
1. Warn customers about outages before they call
Proactive outage communication tells customers what’s happening before they think to call, which keeps restoration questions off your support lines. Once the outage is confirmed, an AI agent sends the notice by SMS or phone call and keeps sending updates until service is restored, which is one example of high-volume communication operations.
McKinsey’s 2021 analysis of a survey consisting of nearly 20,000 North American utility customers found that about half of satisfaction during an outage depends on factors the customer-service team controls, such as how timely and clear the information is. Customers in J.D. Power’s 2026 study rated current outage information among the hardest things to find online.
HappyRobot is an AI agent platform for enterprise operations, and HappyRobot calls its AI agents AI workers. Our AI workers for utilities start once the utility’s own systems confirm the outage, then send each update and confirm restoration with the customer.
2. Answer every call at once, day or night
An outage alert keeps customers off the phone only when their question is about the service restoration timeline. But when a customer has a power outage while the neighbors’ lights are still on, they'll eventually call for a utility update.
You can have a voice AI agent respond to such calls on the first ring, at any hour. Three parts of the AI agent’s phone-line setup make it possible to answer every call at any hour:
- Parallel calls: An AI agent can answer many calls at the same time, so callers do not wait for a representative to become free.
- The same handling for calls outside office hours: A call at night goes through the same diagnostic as a daytime call, and the AI agent writes the outcome to the CRM. That way, the morning shift starts with that call already handled.
- The same phone numbers: An AI agent can answer calls on the utility’s existing phone number, so customers keep calling the number they already use. Our guide to replacing a legacy IVR explains how.
Naturgy runs this setup in production on its technical service line.
Naturgy, one of Spain’s largest gas and electricity suppliers, runs its technical service line on an AI worker built to answer 100% of incoming calls simultaneously, with no wait times, 24 hours a day. Naturgy also adds call volume in stages and checks quality at each stage before routing more calls to the AI worker.
3. Fix technical problems on the first call
Once the call is answered, the customer wants to know whether the fault can be fixed by phone. For that, the AI agent runs a structured diagnostic and walks the caller through each check during the call.
When a customer’s home has lost power supply but the rest of the street still has it, the agent starts at the home’s breaker panel. It asks whether a breaker has tripped, then talks the customer through resetting it. If the power is still out after those checks, the agent books a technician, so the utility provider sends technicians only to faults the customer cannot fix by phone.
Naturgy’s AI worker runs the diagnostic on every call before asking about contract or coverage, so a non-customer gets the same troubleshooting as a customer. Many issues are resolved entirely on the call.
Customers scored the AI worker 9.4 out of 10 in its first weeks. One caller who rated the call 10/10 in a post-call survey wrote:
“The care and attention they showed me — so thorough and so kind.”
4. End fault calls with a booked technician visit
Some faults cannot be fixed over the phone, such as power that stays off after every switch on the breaker panel has been reset. For that customer, the open question is when a technician will arrive.
On a ticket-based fault call, the representative logs the fault as a ticket, and the call ends there. The ticket waits in a queue until a dispatcher assigns a technician, and only then does someone call the customer back with an appointment. The customer has no answer until that callback.
How an AI agent books a technician during a fault call
An AI agent answers the question during the same call by doing the dispatcher’s field service scheduling in four steps:
- Take the fault report: The agent records the fault type and the checks the diagnostic already ran, so the technician arrives knowing what was tried on the phone.
- Find the best-fit technician: The agent checks the field team’s schedule and picks a technician qualified to repair this type of fault who is available to take the visit.
- Confirm the technician: The agent calls the technician directly to confirm they can take the job, because an open slot in the schedule does not guarantee the technician is free.
- Give the customer a name and an arrival time: The agent returns to the customer with the technician’s name and an estimated arrival time before the call ends.
HappyRobot AI workers for utilities can complete all four steps because HappyRobot’s orchestration layer connects the customer’s call to the field team. Our utilities page puts the result in one line: “The customer hangs up knowing the job is booked. Not ticketed. Booked.”
Naturgy’s AI worker arranges a technician whenever the diagnostic cannot fix the fault by phone, and Naturgy promises to have a technician on site anywhere in Spain within three hours.
5. Explain bills and settle meter and billing disputes
Customers call when a bill is higher than they expected. They want to know why the charges are high and may request another meter reading to see if a correction is needed.
The AI agent looks up the account during the call and handles three types of requests:
- Explaining the charge: The agent explains the bill line by line, including the billing period and the meter reading used.
- Meter reading disputes: The agent compares the disputed reading with past readings on the account. The customer can give their own reading on the call. If the readings still do not match, the agent books a meter re-read.
- Billing disputes: The agent can also check the disputed charge. If the charge is wrong, the agent corrects it and arranges payment of the corrected amount.
Naturgy tested invoice explanation and meter-reading submission in its proof of concept, as two of four call types. The test used a copy of Naturgy’s systems with realistic account data, so it didn't affect live customer accounts.
6. Move customers in and out without delays
When a customer relocates to a new address, the utility has to start service at the updated location on the moving date. If the request waits in a queue, service may not be ready on that date.
The AI agent handles the move on the first call by collecting the moving date and the new address. The agent checks the address before creating the new account, so it fixes a wrong street name or unit number before it enters the account record.
A wrong address on the account can send the first bill or a technician to the wrong home. The customer ends the call with the new account confirmed and the service start date set for the new address.
Our AI workers for utilities handle service connections and moves, and a fixed step in the workflow can require the address check to pass before the agent creates the account.
7. Turn required inspections into confirmed visits
Utilities have to inspect some equipment at customers’ homes on a fixed schedule. For example, federal pipeline safety rules require gas utilities to inspect the exposed parts of each service line for atmospheric corrosion at least once every 5 years, with intervals not exceeding 63 months.
Where the meter or pipework is inside the home, the technician needs the customer present to access it. If nobody is home, the visit produces no inspection and has to be rebooked before the due date.
The AI agent books and confirms the visit:
- Before the due date: The agent reminds the customer by SMS or phone call and books a time that suits them.
- Before the visit: The agent confirms that the customer will be home, so the utility knows which visits will go ahead.
- If the time no longer works: The customer can choose a new time, and the original slot becomes free for another visit.
8. Offer payment plans before a bill becomes a debt
In AI in debt collection, the AI agent contacts customers before a payment is due, so a customer who cannot pay the full bill can agree to a repayment plan that splits the amount into smaller installments before the account falls into arrears. On that call, the customer either confirms when they will pay or chooses from the repayment options they are eligible for, and the agent records the agreement on the account.
Once a plan is in place, the agent tracks each installment and contacts the customer when one is missed. If the customer does not respond, the agent passes the account to the utility’s team.
Each contact starts with the customer’s payment history and preferred channel, because our AI workers for finance keep that context from earlier contacts.
9. Personalize usage insights and rate-plan advice
Smart meters record a home’s electricity use at least every hour and send the data to the utility daily. By 2022, about 72% of U.S. electric meters were smart meters, according to the U.S. Energy Information Administration. Using that data, a utility can warn customers that usage is running high before the bill arrives, and show them which rate plan would cost less.
Customers struggle to find this information on their own. In J.D. Power’s 2026 study, pricing plan information was among the hardest things to find in utility apps and websites.
Usage analysis and rate-plan advice are general utility practice, and HappyRobot does not provide them. When a high-usage warning leads a customer to call, the AI agent explains the bill line by line on that call.
10. Give human agents the full story when they take over
Some calls need a person, such as a complaint that requires judgment. The representative who takes that call should know what the customer already told the AI agent, so the customer does not have to repeat it.
Agent-assist tools suggest answers to representatives during calls. The AI agent briefs the representative before the transfer and records the call afterward:
- Transfer on request: A customer can ask for a person at any point. In a warm transfer, the AI agent explains the call to the representative before connecting the customer.
- Summary in the CRM: After every call, the AI agent writes a summary to the CRM with the issue type, the diagnostic result, whether the contract covers the repair, and the next step. Any representative who opens the account later can read it, so no one has to write notes.
Naturgy’s AI worker writes that summary after every call. Terrell Sheridan, Naturgy’s Head of GenAI Transformation, describes the result:
“Our AI agent handles the most routine tasks with precision and consistency, which frees our teams to focus on the situations that require judgment, empathy and a more personal relationship with the customer.”
What AI agents need to improve utility customer experience
An AI agent can complete the work described in the 10 ways only if it can access the customer’s account securely and act on it during the call.
An AI agent understands the request, decides the next step, and acts on it during the contact
A utility should confirm the agent has each of the six capabilities below before it takes live calls:
- Verify the caller: Let the agent send a one-time passcode to the phone number on file before discussing account details.
- Check the contract during the call: Enable the agent to pull the customer's contract to see whether the reported fault is covered.
- Write results to existing systems: The agent should be able to update the CRM and field service tools the utility already runs. Where a system has no API, HappyRobot navigates the system the way a human operator would, as our guide to integrate AI agents with legacy systems explains.
- Follow fixed steps: Caller verification and AI disclosure run as fixed steps in the workflow logic, so the agent cannot skip them.
- Disclose the AI: The agent tells customers they are speaking with an AI agent and lets them ask for a person. In the EU, Article 50 of the AI Act has required AI systems to tell people they are talking to an AI since August 2, 2026.
- Monitor live calls: The system tracks call outcomes in real time and flags unusual call patterns before they turn into operational problems, as our guide to AI agent observability explains.
Naturgy did the groundwork first. It spent three years cleaning, simplifying and consolidating hundreds of products and processes, then migrated its entire customer base to a new platform under the rule “you have to clean before you move.” It then took its proof of concept from kickoff to delivery in three months, before putting the AI worker on its technical service line.
Read Naturgy’s customer story for the full timeline.
In a project like this, being embedded isn’t a nice-to-have; it’s what turns a promising demo into a solution that actually runs in production.
How AI agents improve utility customer experience
AI agents improve utility customer experience by completing the work a call creates while the customer is still on the line, so the customer hangs up with the fault fixed or a technician booked.
To start, pick one call type that ends with a ticket or a callback today, such as a fault report that needs a technician. Measure how many of those calls end with the work completed, then move that call type to an AI agent and compare the two numbers.
Naturgy followed this order. It started with its technical service line and adds call volume only after quality holds at each stage. Once the first call type performs, the same steps apply to the next one on the list, such as meter-reading disputes or moves.
See a HappyRobot AI worker take a fault call from the first ring to a booked technician.
