Photovoltaics, wallboxes and heat pumps have sharply increased connection requests, while meter changes and switching processes run on tight deadlines. Many utilities still handle these cases by hand between portal, billing system and grid planning. Agents take over the case work; grid control stays explicitly out of scope.
Industry 03/103 example agentsAutonomy level L3–L4
Example agents
Three agents that take over work here
Examples from typical workflows, not client references. Which agents pay off for you is shown by assessing your own processes.
01
The grid connection agent
Grid · connection management
Checks registrations for PV systems, wallboxes and heat pumps for completeness, reads data sheets and wiring diagrams, matches them to the connection point, requests missing evidence from the installer and prepares the technical pre-check for grid planning.
Read-only access to grid data, no connection approval, no switching operations, deadlines per request in the calendar.
Human in the loop
Grid planners decide on approval and conditions; the agent delivers the checked file.
What to measure
Time to a complete registration · queries to installers · missed deadlines
02
The meter reading agent
Metering · billing
Checks meter readings for plausibility before billing, resolves outliers with customers through photo requests, recognises meter replacements and estimates and prepares corrections in the billing system before faulty bills go out.
Clear plausibility cases at L4; corrections affecting bills already sent at L3.
Systems
SAP S/4HANA UtilitiesMeter data managementCustomer portalEmail · SMS
Guardrails
Corrections only within defined deviations, no disconnections and no dunning, customer contact labelled as AI.
Human in the loop
Billing confirms corrections above the threshold and every case with a complaint.
What to measure
Corrections after billing · open cases · billing-related calls
03
The market communication agent
Retail · switching processes
Handles rejections and errors in the data exchange with grid operators and suppliers: reads the error message, finds the cause in master data or contract, fixes standard cases and resends the message.
Known error patterns at L4; new ones stay at L3 until they are in the golden set.
Systems
Market communication platformSAP S/4HANA UtilitiesCRMTicketing
Guardrails
Changes only to approved master data fields, messages only in the prescribed formats, deadlines monitored.
Human in the loop
The switching team resolves cases with contract disputes or unclear ownership; the agent documents every attempt.
What to measure
Open error messages · switches on time · manual rework
Framework
What to factor in for this industry
Critical infrastructure: AI as a safety component in the supply of electricity, gas, heating or water is high-risk under the AI Act. The examples here therefore stay in case work and do not touch grid operation.
NIS2: Energy suppliers belong to the sectors with high cybersecurity requirements. For agents that means their own identity, minimal rights and complete logs.
Customer contact: Messages to customers are labelled as AI (Art. 50 AI Act); disconnections and dunning remain human decisions.
Note, not legal advice: whether a specific use falls under these rules is for you and your legal counsel to decide. We provide the technical facts for it, from the agent profile to the audit trail.
