Skip to main content
Vertical AnalysisScore Under 20

Energy and Utilities Agent Readiness: Why Power Companies Are Dark to AI Agents

The $1.5 trillion US energy sector is almost entirely invisible to AI agents. Billing portals require human logins. Usage data is locked in proprietary smart meters. Outage reporting starts with a phone call. The first utility that exposes structured rate plans and real-time usage via API will capture every AI-powered energy broker in its market.

AH
AgentHermes Research
April 15, 202614 min read

The $1.5 Trillion Sector AI Agents Cannot Reach

Energy and utilities are among the oldest industries in America. They built their customer infrastructure before the internet existed, and most of it has not fundamentally changed since. The result is an entire sector that AI agents cannot interact with in any meaningful way.

In our 500-business scan, energy and utility companies consistently scored under 20 on the Agent Readiness Score — firmly in ARL-0 Dark territory. The pattern is universal: every customer-facing function requires either a human login or a phone call. There is no structured data layer that an agent can discover, authenticate against, or call.

This matters because AI energy brokers are coming. When a user asks an AI assistant to “find me a cheaper electricity plan” or “switch my power company,” the agent needs to compare rate plans, check usage history, and initiate a transfer. Today, it cannot do any of those things.

$1.5T
US energy market
<20
avg readiness score
0
utilities with MCP servers
3,000+
US electric utilities

Four Ways Utilities Block AI Agents

Every utility we scanned fails agent readiness for the same four structural reasons. These are not bugs — they are the architecture itself.

Billing Portals

Problem: Customer-login-only access. No public API for rate plans or billing structures.

Agent impact: AI agents cannot compare energy rates across providers on behalf of users.

Usage Data

Problem: Trapped in proprietary smart meters and utility-specific dashboards. Green Button standard exists but almost nobody exposes it via API.

Agent impact: Agents cannot analyze energy consumption patterns or recommend efficiency changes.

Outage Reporting

Problem: Phone-first or SMS-only. Outage maps are interactive web widgets, not structured data.

Agent impact: Agents cannot check outage status or report outages programmatically.

Service Transfers

Problem: Start/stop/transfer service requires calling a phone number and waiting on hold.

Agent impact: AI moving assistants cannot automate utility transfers when users relocate.

Current State vs Agent-Ready: Side by Side

Every utility function has an agent-ready equivalent. The gap is not capability — it is exposure.

Function
Current (Human-Only)
Agent-Ready
Rate Plans
"View plans" behind login wall
get_rate_plans() returns structured pricing JSON
Usage Data
Proprietary meter dashboard
get_usage() returns Green Button XML or JSON
Outages
Interactive map widget, phone hotline
check_outage_status() returns structured incident data
Service Transfer
Call 1-800 number, wait 20 min
transfer_service() completes in one API call
Billing
PDF statements mailed monthly
get_bill() returns line-item JSON with usage breakdown
Solar/DER
Apply via web form, 6-week process
check_solar_eligibility() returns instant qualification

The Smart Grid Paradox: Digital Data, Zero Access

Smart grid infrastructure has been the largest utility investment of the past decade. Over 110 million smart meters are deployed across the US, generating granular 15-minute interval usage data for every customer. This data is fully digital, sitting in utility databases right now.

The paradox is that none of it is agent-accessible. Smart meters communicate with utility backend systems using proprietary protocols. The data flows into billing systems, MDMS (Meter Data Management Systems), and analytics platforms — all behind corporate firewalls with no public API surface.

The Green Button initiative was supposed to solve this. Launched by the US Department of Energy in 2012, it defines a standard XML schema for energy usage data. Over 60 utilities claim compliance. But most only offer Green Button Download My Data — a manual CSV export from a customer portal. The Connect version, which provides OAuth-based API access, is supported by fewer than 10 utilities nationwide.

For AI agents, the distinction between Download and Connect is the difference between a locked filing cabinet and an API endpoint. The data exists. The standard exists. The API surface does not.

The opportunity: A utility that exposes Green Button Connect as an MCP resource lets AI energy advisors analyze usage patterns, recommend rate plans, identify waste, and project savings — all without the customer logging into anything. The first utility to do this becomes the default data source for every AI energy assistant.

The Agent-Ready Utility: Five MCP Tools That Change Everything

An energy utility that ships these five MCP tools goes from ARL-0 Dark to ARL-3 Functional — and captures every AI-mediated energy decision in its service area.

1
get_rate_plans()

Returns all available rate plans with pricing tiers, peak/off-peak rates, and eligibility criteria as structured JSON.

Maps to: D4 Pricing (5%)+4-6 pts
2
get_usage(meter_id)

Returns real-time or daily usage data from smart meters. Supports Green Button format. Enables AI energy advisors.

Maps to: D2 API Quality (15%)+8-12 pts
3
check_outage_status(location)

Returns current outage information for a geographic area with estimated restoration time.

Maps to: D8 Reliability (13%)+5-8 pts
4
transfer_service()

Initiates a service start, stop, or transfer. Accepts address, date, and account information.

Maps to: D3 Onboarding (8%)+6-10 pts
5
subscribe_outage_webhook()

Registers a webhook endpoint for real-time outage notifications. Agents get push updates instead of polling.

Maps to: D9 Agent Experience (10%)+3-5 pts

Combined, these five tools could move a utility from a score of 15 to 55-65 (Bronze to Silver). Add an OpenAPI spec, an agent-card.json, and a status page, and the score pushes past 70. No utility in our scan comes close to this today.

Deregulated Markets: Where the First Mover Wins

In 17 US states plus DC, electricity markets are deregulated. Customers can choose their retail energy provider. Texas alone has 100+ retail electricity companies competing for customers. This is where agent readiness becomes a competitive weapon.

Today, energy comparison happens on websites like PowerToChoose.org (Texas) or PAGasSwitch.com (Pennsylvania). Customers manually enter their zip code, compare rate tables, and fill out enrollment forms. An AI agent trying to do this for a user hits a wall: no API for rate comparison, no structured enrollment endpoint, no machine-readable plan details.

The first retail energy provider in a deregulated market to publish an MCP server with get_rate_plans() and enroll_customer() tools wins every AI-powered energy broker. When someone asks Claude or ChatGPT to “find me a cheaper electricity plan in Houston,” the only provider with an MCP server is the only one the agent can recommend and enroll the customer with directly.

Texas
100+ providers
PowerToChoose.org — no API
Pennsylvania
50+ providers
PAGasSwitch — form-only
Ohio
40+ providers
EnergizeOhio — PDF rate sheets

Frequently Asked Questions

Why do energy companies score so low on agent readiness?

Three structural reasons. First, utility billing systems are legacy mainframe applications (often COBOL-based) that predate APIs entirely. Second, regulatory frameworks assume human-to-human interactions for service changes. Third, utilities are natural monopolies in most markets, which reduces competitive pressure to innovate on customer-facing technology. The result is that the entire customer interface is built around login portals, phone trees, and PDF bills.

What is the Green Button standard and why does it matter for agents?

Green Button is a US Department of Energy initiative that standardizes how utilities share energy usage data. It defines XML schemas for interval usage data, making consumption data machine-readable. The problem: while 60+ utilities claim Green Button compliance, most only offer manual CSV downloads behind authenticated portals. True agent readiness requires a Green Button Connect API that agents can call programmatically with OAuth credentials.

Can a deregulated energy market become agent-ready faster?

Yes. In deregulated states like Texas, Pennsylvania, and Ohio, retail energy providers compete for customers. This competition creates incentive to expose rate plans, usage data, and enrollment workflows via API. The first retail energy provider in a deregulated market to publish an MCP server with rate comparison tools will capture every AI-driven energy broker shopping on behalf of customers.

What about smart grid data — isn not that already digital?

Smart grid data is digital but not agent-accessible. Smart meters transmit usage data to utility backend systems using proprietary protocols like DLMS/COSEM or Zigbee. The data exists in databases, but there is no public API surface. It is like having a warehouse full of goods with no front door — the inventory is digitized, but customers and agents cannot reach it without going through legacy portal authentication.


How agent-ready is your utility?

Run a free Agent Readiness Scan and see exactly where your energy company stands across all 9 dimensions. 60 seconds. No signup required.


Share this article: