Loading
Skip to content
Miracle Minds AI
  • Try “HIPAA”, “model risk”, “FinOps” or “agentic”.

Customer operations and reporting

Customers are angry about bills, so the gain is in resolving billing and outage contacts correctly the first time, not in deflecting them. The binding constraints are consent and accuracy: usage data, AI voice calls and regulatory filings each carry rules a generic contact-center deployment ignores.

Part of AI for Utilities & Energy. 7 sources, checked .

J.D. Power 2025 residential electric utility satisfaction score (1,000-point scale), the lowest across its residential utility studies
499
J.D. Power 2025 residential electric utility satisfaction score (1,000-point scale), the lowest across its residential utility studies
J.D. Power, 2025(opens in a new tab)
Average monthly residential electric bill in 2025, up 34% since 2020 (J.D. Power)
$189
Average monthly residential electric bill in 2025, up 34% since 2020 (J.D. Power)
J.D. Power, 2025(opens in a new tab)
Smart (AMI) meters at U.S. electric utilities in 2025, out of about 170 million meters (EIA)
147 million
Smart (AMI) meters at U.S. electric utilities in 2025, out of about 170 million meters (EIA)
U.S. Energy Information Administration, 2026(opens in a new tab)

The problem

Satisfaction has hit a floor. J.D. Power's 2025 residential electric utility study scored overall satisfaction at 499 on a 1,000-point scale, the lowest across its residential utility studies, and found average monthly residential electric bills up 34% since 2020 to $189, with rising costs the primary driver.1 EIA expects the average residential price to reach 18.2 cents per kWh in 2026, up from 17.2 cents in 2025.2

The hardest contacts are about money. In 2020, 34 million U.S. households, 27% of the total, reported difficulty paying energy bills or kept their home at an unsafe temperature because of cost, and 10% reported receiving a disconnection notice.3 Payment arrangements, assistance programs and disconnections are where an AI mistake costs the most, for the customer and for your standing with the commission.

The data to explain a bill now exists. U.S. electric utilities had about 147 million smart meters in 2025 out of roughly 170 million meters in total.4 Interval data lets an agent show a customer which hours drove a bill, but it is personal information. California's privacy rules for its large utilities bar using usage data for a secondary purpose without the customer's prior, express, written authorization.5

Two more rules shape the design. The FCC has confirmed that TCPA restrictions on artificial or prerecorded voice cover AI technologies that resemble human voices, so automated AI voice calls need consent or an exemption.6 And major electric utilities must file FERC Form 1 every April 18, properly completed and verified.7 AI can draft the work, but every number in a filing still has to trace to a source system.

What we build

  1. 01

    Bill explanation from interval data

    An assistant that uses smart meter data, the customer's rate schedule and weather to explain why a bill changed, shows the hours behind it, and hands disputes to a person.

  2. 02

    Agent assist for the contact center

    Real-time retrieval of tariffs, payment arrangement rules and assistance program eligibility during a call, plus call summaries and notes drafted for the agent to confirm.

  3. 03

    Outage communications

    Restoration estimates and status updates generated from outage management data and sent through each customer's consented channels, with a consent check before any automated voice call.

  4. 04

    Billing exception triage

    Models that flag estimated reads, meter anomalies and likely rate misassignments before bills go out, ranked for a billing analyst to review.

  5. 05

    Regulatory report assembly

    Drafted schedules and narrative for FERC Form 1 and state commission reports, built from queried source data with lineage on every figure, for staff to verify and sign.

Where it goes wrong

The failure modes we design against from the first week, and the control for each.

  • Automating the hardest conversations first

    Hardship, payment arrangement and disconnection contacts carry the highest stakes, and roughly one in ten households reported a disconnection notice in 2020.3 Route these to people with AI assisting, and review outcomes by customer group so automation does not treat similar customers differently.

  • Using usage data beyond its permitted purpose

    Feeding interval data into marketing or third-party models can breach rules that require customer authorization for secondary purposes and limit data to what each purpose needs.5 Tag every data use with its purpose, keep a consent registry, and block secondary uses without a recorded authorization.

  • AI voice calls without consent

    An AI voice agent calling customers about payments or outages is an artificial voice call under the TCPA.6 Check consent before every automated voice call, tell customers they are speaking with an automated system, and offer a transfer to a person.

  • Generated numbers in a filing

    A language model that writes a figure into a report can produce a plausible wrong number, and Form 1 must be properly completed and verified.7 Let AI draft narrative only around values pulled by query from source systems, and require a named reviewer to sign off every schedule.

Rules that apply

  • Telephone Consumer Protection Act, as applied to AI voices (FCC)

    Primary text on federalregister.gov(opens in a new tab)

    The FCC has confirmed TCPA limits on artificial or prerecorded voice calls cover AI-generated voices, so automated AI voice calls to residential or wireless numbers need prior express consent or an exemption.

  • CPUC Decision 11-07-056 (privacy of electricity usage data)

    Primary text on docs.cpuc.ca.gov(opens in a new tab)

    For PG&E, SCE and SDG&E, bars using or disclosing customer usage data for any secondary purpose without prior, express, written authorization, and limits collection, retention and disclosure to what each purpose needs.

  • 18 CFR 141.1 (FERC Form No. 1)

    Primary text on ecfr.gov(opens in a new tab)

    Requires major electric utilities to file FERC Form 1 electronically each year by April 18, properly completed and verified, so AI-assisted preparation needs traceable, reviewed figures.

How to measure it

  • First-contact resolution for billing and outage contacts
  • Repeat contacts on the same issue within 30 days
  • Bill error and rebill rate
  • Accuracy of AI answers in sampled quality reviews against tariffs and source data
  • Report preparation cycle time and post-filing corrections

Questions buyers ask

Can a utility use a chatbot for billing and payment questions?

Yes, for questions that have a factual answer in your systems: balance, due date, why a bill changed, how a rate works. Ground every answer in the customer's account data and your tariffs, and show the source. Route payment hardship, disputes and disconnection questions to a person, with the assistant supporting the agent rather than making the decision.

Do we need customer consent to use smart meter data in AI?

It depends on the purpose and your state. Using interval data to bill and serve the customer is usually a primary purpose. Using it for marketing, product offers or sharing with third parties often counts as a secondary purpose; California, for example, requires the customer's prior written authorization. Map each AI use to its purpose and check your commission's rules before you build.

Can AI prepare FERC Form 1 or state commission reports?

It can draft them, but the filing must still be complete, accurate and verified by your staff. The safe pattern is to pull every figure by query from the general ledger and operational systems, keep the lineage, and let AI write narrative and cross-check consistency between schedules. A person signs off each schedule before anything is filed.

Sources

Every figure on this page comes from the sources below. We checked each link on .

  1. 1.2025 U.S. Electric Utility Residential Customer Satisfaction Study(opens in a new tab)J.D. Power, December 17, 2025 · jdpower.com
  2. 2.Short-Term Energy Outlook: Electricity, Coal, and Renewables(opens in a new tab)U.S. Energy Information Administration, November 10, 2026 · eia.gov
  3. 3.In 2020, 27% of U.S. households had difficulty meeting their energy needs(opens in a new tab)U.S. Energy Information Administration, April 11, 2022 · eia.gov
  4. 4.Electric Power Annual, Table 10.05: Advanced Metering Count by Technology Type, 2016 through 2025(opens in a new tab)U.S. Energy Information Administration, October 21, 2026 · eia.gov
  5. 5.Decision 11-07-056: Decision Adopting Rules to Protect the Privacy and Security of the Electricity Usage Data of the Customers of Pacific Gas and Electric Company, Southern California Edison Company, and San Diego Gas & Electric Company(opens in a new tab)California Public Utilities Commission, July 28, 2011 · docs.cpuc.ca.gov
  6. 6.Implications of Artificial Intelligence Technologies on Protecting Consumers From Unwanted Robocalls and Robotexts(opens in a new tab)Federal Communications Commission, Federal Register, September 10, 2024 · federalregister.gov
  7. 7.18 CFR 141.1: FERC Form No. 1, Annual report of Major electric utilities, licensees and others(opens in a new tab)Electronic Code of Federal Regulations · ecfr.gov