Tool · Map
AI use-case map for energy utilities: 25 use cases, data needs and AI Act classification
Which AI application needs which data, and when the AI Act takes a closer look
- 25 use cases
- As of 17 October 2026
Book an implementation check →
30 minutes on the project that's pressing for you right now. No sales loop.
The AI use-case map for energy utilities sorts 25 AI use cases from grid, metering, sales, heat and water by market role, data needs, classification under the EU AI Act and publicly documented industry status.
A classification for typical market roles, not a review of your individual case and not legal advice. Whether a system is high-risk for you depends on its intended purpose, protective function and critical-entity status; your legal department or counsel will confirm that. The high-risk conditions rely on the European Commission's draft guidelines of 19 May 2026, which are not final yet.
- The AI Act explicitly lists 2 of the 25 use cases as high-risk, and none of them is in the grid: credit checks and applicant screening.
- 5 only become high-risk under conditions: when the AI directly protects an asset and the operator is designated a critical entity.
- Chatbots and voicebots have been subject to the Art. 50 transparency duty since 2 August 2026.
- Art. 4 applies to every case: deployers must take measures for their staff's AI literacy.
The map: 25 use cases
Each card gives the benefit, the data the use case needs, the classification under the EU AI Act as a condition, and the publicly documented industry status with a source. As of 17 October 2026.
What stands out is where high risk does not sit. Under the draft guidelines most grid applications drop out as long as a person decides or an independent protection layer remains. The clear-cut cases are in sales and HR.
Grid
DSO
Low-voltage grid state estimation
Estimates voltage and current in local grids where only some connections are metered. That is exactly the evidence German DSOs need for grid-oriented control under Section 14a EnWG.
- Data: Smart meter readings (TAF 10), substation measurements, grid topology from the GIS, asset master data
- EU AI Act, High-risk only under conditions: As an estimate for the control room, not a safety component under the draft guidelines. The open case is AI that triggers Section 14a control itself with no independent protection layer left. Then the protective function and the operator's critical-entity status decide.
- Rules: Section 14a EnWG; BK6-22-300; VDE FNN guidance on grid state determination; Section 66 MsbG
- Industry status, proven in pilots: The VDE FNN guidance explicitly allows neural networks but expects the same accuracy as from classic estimators. According to VDE ETG, field tests in the GridAnalysis project show high accuracy in low-voltage grids. Source: VDE ETG: AI in grid control technology
Article: AI in grid state determination: what measurement has to deliver →
DSO · TSO
Load and generation forecasting
Predicts consumption and generation per grid node for the coming hours and days. Redispatch 2.0 requires such forecasts with about 36 hours' lead time.
- Data: Grid and smart meter time series, weather data, asset master data from the market master data register, schedules
- EU AI Act, Usually not high-risk: Usually not high-risk. The draft guidelines list consumption forecasts and forecast-based grid optimisation as counterexamples. That can change if forecast, optimiser and switching command together take on a protective function; the draft assesses such chains as a whole.
- Rules: Sections 13, 13a, 14 EnWG (Redispatch 2.0); Section 66 MsbG; Art. 4 AI Act
- Industry status, proven in pilots: According to VDE ETG, load and feed-in forecasts based on weather and camera data are among the applications that already run reliably or have passed field tests. Source: VDE ETG: AI in grid control technology
DSO
Grid planning and congestion analysis
Runs thousands of scenarios for the grid expansion plan. A trained surrogate model approximates the load flow calculation; the critical cases still go through the exact calculation.
- Data: Grid model for high, medium and low voltage, regional scenario, connection data of controllable consumption devices, load profiles
- EU AI Act, Usually not high-risk: No high-risk purpose. Planning is not operation, and the AI protects no asset. For a formal plan the method still has to remain traceable.
- Rules: Section 14d EnWG; Art. 4 AI Act
- Industry status, research: The dena project Data4Grid studied data analytics and AI in distribution grids, with an expert report by Fraunhofer IEE. Evidence of routine use in expansion planning is still missing. Source: Fraunhofer IEE on the final report of the dena project Data4Grid
Article: AI Grid Simulation: Congestion Analysis for the Section 14d Plan →
DSO · TSO
Condition forecasting for cables and transformers
Combines diagnostic measurements, operating data and fault history into a running condition forecast. Assets get replaced before they fail, not by the calendar.
- Data: Tan delta and partial discharge measurements, online dissolved gas analysis, load and temperature curves, fault and maintenance history
- EU AI Act, High-risk only under conditions: Mostly not high-risk: the draft guidelines do not classify predictive maintenance on a gas pipeline as high-risk because the safety systems keep running independently. If the AI itself detects hazardous states and that independent layer is missing, it can be a safety component at a critical entity.
- Rules: Annex III point 2 AI Act; KRITIS umbrella act; Art. 4 AI Act
- Industry status, research: In the AMika research project at the University of Wuppertal, 24 grid operators supply diagnostic data for ageing curves of medium-voltage cables. Source: University of Wuppertal: AMika project
TSO · DSO
Drone inspection of lines and towers
Drones fly along overhead lines and towers, image recognition marks damage to insulators and fittings or corrosion. A person assesses the findings.
- Data: Image and LiDAR captures, tower and line register, inspection history
- EU AI Act, Usually not high-risk: Usually not high-risk. The image recognition reports and does not intervene; the draft guidelines do not classify an inspection robot that only detects as high-risk.
- Rules: Implementing Regulation (EU) 2019/947 (drone operations); GDPR for images of people and property; Art. 4 AI Act
- Industry status, proven in pilots: Grid operators use drones, LiDAR and image recognition in line inspection, in some cases beyond the pilot's visual line of sight. Source: NETZinspect project page: drones, LiDAR and AI in line inspection
Article: AI-assisted drone inspection of power grids: BVLOS and automatic image recognition →
DSO · TSO
Control room assistance
Prepares alarms, measurements and operating instructions for shift staff, summarises faults and suggests analysis steps. People decide and switch.
- Data: Control system data (measurements, switching states, alarms), fault logs, operating instructions, grid model
- EU AI Act, Usually not high-risk: Early warnings and assistance where the control room decides are not high-risk under the draft guidelines. The line to anomaly detection as a basis for decisions is fluid; the provider's statement of intended purpose tips the balance.
- Rules: IT security catalogue under Section 11(1a) EnWG; BSIG (German NIS2 implementation); Art. 4 AI Act
- Industry status, research: An AI agent from Fraunhofer IOSB-AST that automates routine analyses has been in a test run with three German grid operators since February 2026. Source: Fraunhofer: towards secure energy (2026)
Article: Agentic AI in Energy Utilities: Autonomous Grid Management →
TSO · DSO
Anomaly detection in grid operation
Detects critical grid states in data patterns, warns the control room and suggests countermeasures, for example on load distribution, stability or disconnections.
- Data: High-resolution grid measurements, control system data, grid model, large sets of simulated faults for training
- EU AI Act, High-risk only under conditions: The draft guidelines name exactly this purpose as a high-risk example. It only applies if the operator is designated a critical entity. Then deployer obligations apply from 2 December 2027, among them human oversight (Art. 14) and logs kept for at least six months (Art. 26).
- Rules: Annex III point 2 AI Act; Art. 14, Art. 26 AI Act; KRITIS umbrella act
- Industry status, research: In the eKI4DS project, Fraunhofer IEE and a transmission system operator are researching AI that detects critical grid states. So far it has only been tested on smaller test grids. Source: Fraunhofer: towards secure energy (2026)
Article: EU AI Act Annex III: which AI at a utility is really high-risk →
TSO · DSO · Heat/water
Site protection at substations and plants
Analyses camera, radar and radio sensors at the fence, detects intruders and drones and alerts the control room.
- Data: Video, radar and radio data, site plans, alarm history
- EU AI Act, High-risk only under conditions: Perimeter protection is the first high-risk example for electricity supply in the draft guidelines, because it directly protects physical integrity. The operator's critical-entity status remains the condition. If the system identifies people biometrically at a distance, Annex III point 1 applies as well.
- Rules: Annex III point 2 AI Act; KRITIS umbrella act (physical protection); GDPR and Section 4 BDSG (video surveillance)
- Industry status, no public evidence: Layered detection chains are a topic of the KRITIS umbrella act. No public figure shows how often learning image analysis runs in them. Source: European Commission: draft guidelines on high-risk classification
DSO
Pre-checking grid connection requests
Checks incoming applications for completeness, extracts documents and gives a first capacity estimate before a planner takes over.
- Data: Application data in the VDE FNN data set, unit data sheets, grid model and GIS, capacity maps
- EU AI Act, Usually not high-risk: No high-risk purpose. If the system communicates directly with applicants, the transparency duty under Art. 50(1) applies.
- Rules: Section 8(7) EEG (grid connection portal); VDE-AR-N 4105 and 4110; GDPR
- Industry status, no public evidence: Portals and standardised application data are spreading; the VDE FNN data set has been at version 3.0 since April 2025. There are no public figures on how much of the pre-check uses AI. Source: VDE FNN: digitalising the grid connection process
Article: Grid Connection Bottlenecks for Charging Parks: Digitalising Grid-Operator Processing →
Metering and market communication
MPO · DSO
Meter data validation and meter reading recognition
Finds implausible meter readings and load profiles, proposes substitute values and reads meter values from customer photos.
- Data: Meter readings and load profiles, consumption history per metering point, master data, photos of meters and installations
- EU AI Act, Usually not high-risk: No high-risk purpose. The draft guidelines explicitly name photo-based quality checks of meter installations as a counterexample.
- Rules: MsbG (meter data processing); VDE-AR-N 4400 (metering code); GDPR
- Industry status, proven in pilots: According to PwC, one grid operator resolves 90 per cent of exception cases in meter reading validation fully automatically. Another utility recognises around 50,000 meter readings a month by image recognition. Source: ZfK on the PwC study on AI in the energy industry
DSO · MPO · Supplier
Exception cases in market communication
Sorts rejected or incomplete messages, identifies the cause and prepares the correction. Since the 24-hour supplier switch, every hour counts.
- Data: EDIFACT messages and responses (APERAK, CONTRL), master data of market and metering locations, process history
- EU AI Act, Usually not high-risk: No high-risk purpose. Duties come from data protection and the deadlines of the market processes, not from Annex III.
- Rules: GPKE, WiM, MaBiS (BNetzA determinations); GDPR; Art. 4 AI Act
- Industry status, proven in pilots: According to PwC, 79 per cent of utilities have introduced first AI solutions, 61 per cent of them still experimentally. Automated exception handling is among the documented cases in live operation. Source: ZfK on the PwC study on AI in the energy industry
BRP · Supplier
Forecasting in balancing group management
Forecasts consumption and generation in the balancing group per quarter hour, so that less imbalance energy is settled at the reBAP price.
- Data: Load profiles and schedules, weather data, smart meter readings, reBAP and intraday prices
- EU AI Act, Usually not high-risk: Not high-risk. The draft guidelines explicitly do not classify imbalance forecasts in 15-minute resolution under human oversight as high-risk.
- Rules: Section 4 StromNZV; Standard balancing group contract (BK6-23-102); Section 69 MsbG
- Industry status, research: In the Kopernikus project on industrial energy flexibility, Fraunhofer IPA compares forecasting methods to lower imbalance costs. Source: Fraunhofer IPA: lowering imbalance costs through a forecast benchmark
Sales and customer service
Supplier · DSO
Chatbot and voicebot in customer service
Answers questions on instalments, tariffs, moving house and meter readings around the clock and hands over to a person when things get tricky.
- Data: FAQ and knowledge articles, tariff and contract data, customer data with authorisation checks for account queries
- EU AI Act, Transparency duty under Art. 50: Art. 50(1) has applied since 2 August 2026: anyone talking to the bot must be able to tell it is an AI, unless that is obvious. The duty falls on the provider, and a utility that puts a bot into service under its own name can be the provider itself. Generated speech in a voicebot must be marked machine-readably under Art. 50(2), for systems placed on the market before that date from 2 December 2026.
- Rules: Art. 50 AI Act; GDPR; Art. 4 AI Act
- Industry status, widespread: According to PwC, customer service is the most common AI area at utilities, at 29 per cent. One grid operator answers 95 per cent of enquiries there by chatbot on its own. Source: ZfK on the PwC study on AI in the energy industry
Article: AI Customer Service and Churn Prevention at Utilities →
Supplier
Churn prediction and customer retention
Calculates a cancellation probability per customer and directs win-back efforts to the contracts where they pay off.
- Data: Contract and tariff data, contact and complaint history, payment behaviour, consumption data within purpose limitation
- EU AI Act, Usually not high-risk: No Annex III purpose. The legal questions sit in data protection: profiling and the right to object, purpose limitation of meter data. Grid data in a multi-utility group touches unbundling rules.
- Rules: Art. 21, Art. 22 GDPR; Section 69 MsbG; Section 6a EnWG (unbundling)
- Industry status, proven in pilots: According to PwC, sales is an AI area at 12 per cent of utilities. Source: ZfK on the PwC study on AI in the energy industry
Article: AI Customer Service and Churn Prevention at Utilities →
Supplier
Credit check at contract signing
Assesses the ability of new household customers to pay and helps decide on the contract, prepayment or the instalment amount.
- Data: Credit agency scores, payment and dunning history, contract and address data
- EU AI Act, Purpose listed in Annex III: Annex III point 5(b) lists evaluating the creditworthiness of natural persons or establishing their credit score. Recital 58 gives access to electricity, which such assessments determine, as an explicit reason. AI used to detect financial fraud is excluded. Obligations apply from 2 December 2027, and for every deployer they include a fundamental rights impact assessment before first use (Art. 27(1)).
- Rules: Annex III point 5(b) AI Act; Art. 26, Art. 27 AI Act; Art. 22 GDPR
- Industry status, no public evidence: There are no public figures on AI in credit checks in energy retail. In 2023 the CJEU ruled that even a score supplied by a third party can be an automated decision if the company relies on it decisively. Source: CJEU, judgment of 7 December 2023, C-634/21
Supplier · DSO · MPO
Checking bills before dispatch
Stops implausible bills before they go out: jumps in consumption, wrong price versions, duplicate items.
- Data: Bill items, consumption history, tariff and price master data, meter readings
- EU AI Act, Usually not high-risk: No high-risk purpose. If the system alone decides on disconnection or dunning, it becomes a question under Art. 22 GDPR.
- Rules: Sections 40, 40b, 40c EnWG; Art. 22 GDPR
- Industry status, no public evidence: No figures on AI in bill checking have been published. The deadlines that a late bill misses are set in Section 40c EnWG. Source: Bundesnetzagentur: bills and deadlines under Section 40c EnWG
Supplier · DSO
Inbound mail and documents in customer service
Reads emails, letters and forms, matches them to the contract account and routes them to the right team. It drafts replies; a person sends them.
- Data: Emails, letters and scans, forms, contract accounts, reply templates
- EU AI Act, Usually not high-risk: No high-risk purpose. If the system sends replies to customers itself, the transparency duty under Art. 50(1) applies.
- Rules: GDPR; German DPA guidance on AI and data protection (05/2024); Art. 4 AI Act
- Industry status, proven in pilots: According to PwC, customer service is the most common AI area at utilities, at 29 per cent. Source: ZfK on the PwC study on AI in the energy industry
Article: AI Customer Service and Churn Prevention at Utilities →
Heat and water
Heat/water
District heating: temperature control and heat load forecasting
Forecasts the heat load and lowers supply and return temperatures as far as customers allow. Only low temperatures make large heat pumps and waste heat economical.
- Data: Supply and return temperatures, heat meters, weather forecasts, network model, data from substations in buildings
- EU AI Act, Usually not high-risk: Optimisation and forecasting are not safety components under the draft guidelines. Heat supply is listed in Annex III point 2, though: if the AI takes on protective functions such as pressure or temperature limits without an independent protection layer, it needs a fresh assessment.
- Rules: Section 32 WPG (decarbonisation roadmap); GDPR for consumption data; Art. 4 AI Act
- Industry status, proven in pilots: The dena guide groups the network-side use cases into load forecasting, return temperature optimisation, predictive maintenance and leak detection, drawing on its own pilot projects. Source: dena: guide to AI in district heating (2024)
Article: AI in District Heating: Network Optimization and the Transformation Plan under WPG →
Heat/water
Leak detection in water and heat networks
Learns the normal behaviour of pressure, flow and consumption and reports deviations that point to a leak.
- Data: Pressure and flow measurements, pump data, remotely read meters, network plan
- EU AI Act, High-risk only under conditions: Detecting hazardous states is one of the protective functions in the draft guidelines, and both water and heat supply are listed in Annex III point 2. If the AI only reports to people and the protection systems run independently, it is not a safety component. If it shuts off sections itself, it can be one at a critical entity.
- Rules: Annex III point 2 AI Act; KRITIS umbrella act; GDPR for consumption data
- Industry status, research: Research projects detect leaks through virtual sensors. In 2023 the researchers did not yet see a practical, explainable solution. Source: ZfK: AI-based early detection of damage in drinking water networks (2023)
Article: AI in District Heating: Network Optimization and the Transformation Plan under WPG →
Cross-functional
all roles
Knowledge assistant for rules and determinations
Answers questions on determinations and internal instructions with sources, based on the company's own documents (RAG).
- Data: Rulebooks, determinations, operating instructions and process descriptions, with a clean permission model
- EU AI Act, Usually not high-risk: No high-risk purpose. Art. 50(1) requires a notice that an AI is involved unless that is obvious; with an internal assistant it usually is. The obligations for the underlying language model sit with its provider.
- Rules: German DPA guidance on generative AI with RAG (10/2025); Section 87(1) no. 6 BetrVG; Art. 4 AI Act
- Industry status, proven in pilots: According to PwC, support functions are an AI area at 21 per cent of utilities. Source: ZfK on the PwC study on AI in the energy industry
Article: The 3 Methods of AI Assistant Knowledge Integration: What You Need to Know →
all roles
AI assistance in everyday office work
Drafts, summaries, minutes, translations: generative AI in the workplace tools, for every department.
- Data: Emails, documents, calendars and file shares, governed by the existing access rights
- EU AI Act, Usually not high-risk: No high-risk purpose. Art. 4 applies to everyone: deployers must take measures for their staff's AI literacy. If the tool lends itself to monitoring performance or behaviour, the works council or staff council has a say.
- Rules: Art. 4 AI Act; Art. 28 GDPR (processing on behalf); Section 87(1) no. 6, Section 90(1) no. 3 BetrVG
- Industry status, proven in pilots: According to PwC, 79 per cent of utilities have introduced first AI solutions. Only about one in four has an AI roadmap. Source: ZfK on the PwC study on AI in the energy industry
Article: Shadow AI: The Underestimated Risk in European Companies →
all roles
Coding assistance in IT and business units
Writes code and tests, from controlling reports to the interface with the billing system.
- Data: Source code, configurations, tickets and documentation; no production data in prompts
- EU AI Act, Usually not high-risk: No high-risk purpose. The risks are information security and licensing: where the code lives, who runs the model, which licence covers the output.
- Rules: BSIG (German NIS2 implementation); IT security catalogue under Section 11(1a) EnWG; Art. 4 AI Act
- Industry status, proven in pilots: According to PwC, an AI assistant cut programming effort in one municipal utility's controlling team by 70 per cent. Source: ZfK on the PwC study on AI in the energy industry
all roles
Procurement and tenders
Compares bids with the specification, finds deviations in contract clauses and summarises bidder questions.
- Data: Specifications, bids, contract templates, supplier master data
- EU AI Act, Usually not high-risk: No high-risk purpose. The other direction matters more: anyone buying AI systems can require high-risk conformity in the tender and contract without disclosing their own critical-entity status (draft guidelines, paragraph 191). Evaluating the bids stays with the contracting authority.
- Rules: GWB and SektVO (public procurement law); GDPR; Art. 4 AI Act
- Industry status, no public evidence: There are no public figures on AI in utility procurement. Source: European Commission: draft guidelines on high-risk classification
Article: EU AI Act Annex III: which AI at a utility is really high-risk →
all roles
Pre-selection of job applications
Sorts and scores applications and suggests candidates for interviews.
- Data: CVs and application documents, job profiles, past selection decisions
- EU AI Act, Purpose listed in Annex III: Annex III point 4(a) lists the selection and assessment of applicants. Obligations apply from 2 December 2027; before use in the workplace, deployers inform workers' representatives (Art. 26(7)). Deployers governed by public law, and private ones providing public services, also need a fundamental rights impact assessment (Art. 27).
- Rules: Annex III point 4 AI Act; Art. 26(7), Art. 27 AI Act; Art. 22 GDPR; Section 95(2a) BetrVG, for municipal companies the state staff representation acts
- Industry status, no public evidence: There are no public figures for the energy industry. Source: Regulation (EU) 2024/1689, Annex III point 4
DSO · TSO · Heat/water
Attack detection in IT and OT networks
Learns the normal traffic in control and telecontrol networks and reports deviations that point to an attack.
- Data: Network traffic from SPAN ports or taps, log data, asset and protocol inventory
- EU AI Act, Usually not high-risk: Not a safety component: recital 55 excludes AI used solely for cybersecurity, and the draft guidelines list cyber monitoring of grid communication as a counterexample.
- Rules: BSIG (German NIS2 implementation); IT security catalogue; BSI guidance on attack detection systems
- Industry status, widespread: Operators of energy supply networks and energy plants have had to evidence attack detection systems to the BSI since 2023. Detection is one of their three functional areas. Source: BSI: FAQ on attack detection systems
How to read the classification
The EU AI Act column describes conditions, not a verdict. It has four levels:
- Purpose listed in Annex III: The AI Act lists the purpose explicitly as high-risk. Deployer obligations apply from 2 December 2027. The exemption in Art. 6(3) only covers narrowly defined auxiliary tasks.
- High-risk only under conditions: Under the draft guidelines of 19 May 2026, Annex III point 2 only applies if both hold: the AI directly protects the physical integrity of the asset (safety component), and the operator is designated a critical entity under the CER Directive, in Germany under the KRITIS umbrella act.
- Transparency duty under Art. 50: Not high-risk, but a disclosure duty since 2 August 2026: people must be able to tell that they are talking to an AI.
- Usually not high-risk: No Annex III purpose and no safety component. Art. 4 (AI literacy), data protection and sector law still apply.
The industry status shows what is publicly documented, not what is technically possible:
- widespread: in use at many utilities, publicly documented
- proven in pilots: pilots or live operation at individual utilities, publicly documented
- research: research projects and field tests
- no public evidence: no public figure on use in the industry
Two things apply to every entry. Art. 4 requires measures for AI literacy, however harmless the system. And anyone who uses a purchased system for a new purpose can become its provider under Art. 25. The code stays the same; the classification does not.
Since 29 July 2026, Germany's Bundesnetzagentur has been the market surveillance authority for AI under the KI-MIG. Its AI service desk offers an online tool for a first risk assessment.
Where municipal utilities start
Data first
Almost every card above starts with data that is scattered across the company: readings in meter data management, topology in the GIS, master data in billing. Anyone who wants grid state estimation first needs a low-voltage grid model that is correct.
That is not an AI task. It decides whether the AI will be any good later, though.
From pilot to operation
According to PwC, 61 per cent of utilities using AI are still experimenting. A pilot shows that a model works. Operation means someone is responsible, the model is monitored, and there is a way back when it gets things wrong. That belongs in the plan from day one, with budget and support beyond the end of the project.
Governance before the regulator asks
An AI register with purpose, role as provider or deployer, classification and rationale is the shortest route to your own obligations. The classification on this map is a starting point, nothing more. It is only settled once the statement of purpose, the protection layer and the critical-entity status are clear.
How we work on this: AI for municipal utilities (from pilot to operation) and AI governance for energy utilities (register, roles, approval paths).
Sources
Each card carries the source for its industry status. For the classification overall:
Frequently asked questions about the AI use-case map
An overview of 25 AI use cases in grid, metering, sales, heat and water. For each case it gives market roles, data needs, the classification under the EU AI Act as a condition and the publicly documented industry status with a source. It classifies; it does not review individual cases.
No. Under the European Commission's draft guidelines of 19 May 2026, Annex III point 2 only applies if the AI directly protects the physical integrity of an asset and the operator is designated a critical entity. The draft lists load and consumption forecasts, early warnings to the control room and pure cybersecurity AI as counterexamples, and perimeter protection and anomaly detection as a basis for load distribution or disconnection decisions as high-risk examples. The draft is not final yet.
On this map, two: credit checks on household customers (Annex III point 5(b); recital 58 mentions access to electricity) and pre-selection of job applications (Annex III point 4). Deployer obligations apply from 2 December 2027.
Since 2 August 2026, Art. 50(1) applies: people must be able to tell they are talking to an AI unless that is obvious. Generated speech must be marked machine-readably under Art. 50(2); for systems placed on the market before that date, this applies from 2 December 2026. A customer service bot is not high-risk.
No. Whether a system is high-risk for you depends on the provider's statement of purpose, on whether an independent protection layer remains, and on your company's critical-entity status. Your legal department or counsel will confirm that. The Bundesnetzagentur offers an online tool for a first assessment through its AI service desk.
More in this topic
The topic AI in the energy industry collects articles and tools. What the AI Act requires in detail is in EU AI Act for energy utilities. How we take AI from pilot to operation is described in AI for municipal utilities, and register, roles and approval paths in AI governance for energy utilities.
Let's collaborate
Which use case is next for you?
Thirty minutes on what you're working on, whether grid, metering, market processes, IT architecture or AI. Afterwards you'll know how we'd approach it and who would sit at the table.
Energy and measurement technology since 2010 · innobu GmbH since 2014 · Based in Schleswig-Holstein, Germany
Whoever pitches also delivers: whoever leads the first call stays your contact right through to the steering committee, with established partners in the project. Direct: info@innobu.com