AI for German Stadtwerke and Energieversorger: How Municipal Utilities Automate Marktkommunikation (MaKo), Zählerwesen and Netzanschluss With Custom AI Agents

On any given Wednesday in a German Stadtwerk, a MaKo-Sachbearbeiterin opens her queue at 7:30 a.m. There are 218 unbearbeitete UTILMD-Klärfälle from 14 Marktpartnern, 47 MSCONS-Lastgangsfehler from the Smart-Meter-Gateway-Administration, a Beschwerde from a Kunde whose Lieferantenwechsel has been pending for three weeks, a Netzanschluss-Antrag for a 22-kW-Wallbox that has been sitting unbearbeitet since the Osterferien, and an email from the Geschäftsführung asking why the MaKo-Klärfall-Quote is up 18 percent year-over-year. The phone rings. It is the Vertrieb asking how many neue Marktpartner the team can onboard before the MaKo-2026-Stichtag on 1 April.

The Energiewirtschaft has been talking about Digitalisierung for a decade. Inside the German Stadtwerk profession, the math has finally caught up with the talk. The VKU represents around 1,500 kommunalwirtschaftliche Unternehmen across Energy, Water and Waste; the 743 VKU-Mitglieder im Strom collectively manage roughly 31 million Stromzähler¹⁴⁵. Around 102,000 people work at kommunalen Energieversorgern (Strom, Gas, Wärme)⁵. And around 80 percent of surveyed Stadtwerke cite Zeitmangel, Fachkräftemangel and missing Digital-Expertise as central obstacles to transformation²⁴. Hiring has stopped being a strategy. Throughput per MaKo-Sachbearbeiter and Kundenservice-Agent has become the constraint.

This guide is for the IT-Leiterin, the Vertriebsleitung or the Geschäftsführerin who has watched the AI hype cycle for two years and now wants a concrete answer to one question: can a custom AI agent actually take over MaKo-Klärfälle, Customer Service triage, Netzanschluss-Bearbeitung and Smart-Meter-Plausibilität inside SAP IS-U, Wilken or Schleupen.CS - and if so, how?

The Stadtwerk Crunch

The reason Stadtwerk-Arbeit consumes so much time is rarely the energy delivery itself. It is everything around it: the 14th UTILMD-Klärfall before lunch, the 200th Rechnungsfrage from a Kunde this week, the Wallbox-Anschlussantrag that has been in the GIS-Queue for three weeks, the MSCONS-Lastgangsfehler from the SMGW because the Datenkonzentrator was on vacation, the Wechselablehnung from a Marktpartner that came back with the wrong PARTIN. Multiply that by four Sparten (Strom, Gas, Wärme, Wasser) and the math gets ugly fast.

• ~1,500 VKU-Mitgliedsunternehmen - The Verband kommunaler Unternehmen represents around 1,500 kommunalwirtschaftliche Firms across Energy, Water and Waste; the VKU-Energie covers 743 Mitglieder im Strom alone¹⁴⁶
• ~31 million Stromzähler under VKU-Mitgliedern - The VKU-Energie members collectively manage around 31 million Stromzähler, a meaningful share of the German Marktrolle landscape⁴
• ~102,000 employees in kommunaler Energieversorgung - Around 102,000 people work at VKU-Mitgliedsunternehmen in Strom, Gas and Wärme - the talent base under maximum demographic pressure⁵
• ~80% cite Zeitmangel and Fachkräftemangel - The zfk and BET-Consulting analysis confirms around 80 percent of surveyed Stadtwerke see Zeitmangel, Fachkräftemangel and missing digital expertise as central obstacles²⁴²⁵
• MaKo 2026 Stichtag on 1 April 2026 - The MaKo 2026 release (BNetzA Mitteilung Nr. 34) is a stabilization release, not a revolution - but the EDIFACT format updates across UTILMD, MSCONS, COMDIS, ORDERS, ORDRSP, REMADV, PARTIN and INVOIC require every Marktteilnehmer to be ready¹⁰⁸
• ~1.6 million iMSys installed (mid-2025) - The Bundesnetzagentur-Monitoring confirms about 1,607,202 Smart-Meter installations as of 30 June 2025, with around 760,000 mandatory installations counting towards the statutory target¹⁷
• The Kunde is changing too - Energy customers trained by mobile-app banking and same-day Lieferando expect app-level responsiveness. The Stadtwerk that answers in two weeks loses the Wechselkunden to one that answers same-day.

The result is a Stadtwerk where the most experienced MaKo-Sachbearbeiter and Kundenservice-Agents spend the largest share of their day on the most repetitive work, while Marktanalyse, Tarif-Strategie, Netzplanung and Geschäftskundenbetreuung get squeezed into the last hour. AI agents do not fix this by being clever. They fix it by removing the repetitive 60 percent that nobody, on either side of the desk, wants to do.

What an AI Agent Actually Does in a Stadtwerk

The market is full of products labelled as “KI für Stadtwerke”. Most of them are chatbot widgets in the Kundenportal. A real AI agent goes much further: it owns the workflow, not just the front-end. Here is the difference, in plain Stadtwerk terms.

The agent loop in a Stadtwerk

1. Capture - Pulls inputs from the MaKo-Plattform (Seeburger, BCS, Robotron, Procilon, ponton-x), the ERP (SAP IS-U / S/4HANA Utilities, Wilken, Schleupen.CS, Robotron, Lima, kVASy), the CRM, the GIS / Netzdaten, the Smart-Meter-Gateway-Administration (GWA), the bank feed and the customer email inbox.
2. Classify - Reads the input - an UTILMD-Klärfall, a Rechnungsfrage, a Netzanschluss-Antrag, eine MSCONS-Anomalie, ein Lieferantenwechsel - identifies the Sparte, the Marktrolle, the Geschäftsprozess and the Zählpunkt, and routes to the right pipeline.
3. Extract - Pulls structured fields: MPID, Lokations-ID, Zählpunktbezeichnung, Marktpartner, Geschäftsprozess-Code, OBIS-Kennzahl, Bilanzkreis - with a confidence score for every field.
4. Enrich - Looks up the Vertragsakte, the Tarif-Logik, the Bilanzkreis-Zuordnung, the GIS-Lage, the Netzentgelt-Tabelle, the relevant Geschäftsprozess and the previously known Marktpartner-Pattern.
5. Propose - Generates the full output: EDIFACT-Antwort (UTILMD, MSCONS, COMDIS), draft Antwort an den Kunden, Netzanschluss-Bescheid mit Berechnung, Smart-Meter-Plausibilitäts-Bericht, Forderungs-Mahnstufe, Wechsel-Bearbeitung.
6. Decide - Above the confidence threshold the agent posts directly into the MaKo-Plattform or the ERP; below it the case lands in a Sachbearbeiter review queue with a one-click approve or correct.
7. Learn - Every correction feeds back. Patterns the agent sees three times on a specific Marktpartner or Kundengruppe become rules it applies automatically.
8. Audit - Every step is logged with timestamp, document hash, agent version and (on review) the human who signed off - the audit trail for a BNetzA-Anhörung and the Bilanzkreis-Plausibilität share the same event log.

The difference from what you have today

The category matters. Kundenportal-Chatbots help with FAQs. SAP IS-U and Schleupen.CS handle the system of record. Seeburger BIS and Robotron MaKo handle the EDIFACT-Pipe. But the Sachbearbeiter still has to assemble the answer, draft the EDIFACT response, run the Klärfall against the Geschäftsprozess and update the ERP - across the chatbot, the MaKo-Plattform, the ERP and the GIS. A custom agent goes one layer further: it removes the work entirely from the human path until something looks unusual or judgement-dependent.

6 Use Cases That Work Today

Not every Stadtwerk-Prozess is a good first AI candidate. The ones below are - they are high volume, repetitive, well documented and have a clear correct answer. Start with one Sparte in one use case. Add the next once the first runs cleanly.

1. Marktkommunikation (MaKo) Klärfall-Bearbeitung

The single highest-leverage use case. A 100,000-Zählpunkt-Stadtwerk receives 5,000 to 15,000 MaKo-Nachrichten per month across UTILMD, MSCONS, ORDERS, COMDIS, REMADV, INVOIC and PARTIN. The agent natively parses the MaKo-2026-formats, runs Klärfall-Triage and drafts the response.

• MaKo-2026-konform out of the box - UTILMD (Strom and Gas), MSCONS, COMDIS, ORDERS, ORDRSP, REMADV, PARTIN, INVOIC against the BNetzA Mitteilung 34 dictionary⁸¹⁰
• Klärfall-Klassifikation - Lieferantenwechsel-Ablehnung, Marktpartner-Stammdaten-Konflikt, Lastgangsfehler, Zählpunkt-Konflikt, Rechnungsdiskrepanz - jeder Fall mit Geschäftsprozess-Bezug
• GPKE / GeLi Gas Logik - The agent applies the Geschäftsprozesse zur Kundenbelieferung mit Elektrizität and Geschäftsprozesse Lieferantenwechsel Gas per Marktrolle¹²
• Drafts the outbound EDIFACT message - Generated and queued in the MaKo-Plattform (Seeburger, BCS, Robotron) for sending after Sachbearbeiter approval
• Bilanzkreis-Plausibilität - Cross-check against the Bilanzkreis-Zuordnung and the Lieferantenwechsel-Historie
• Time saved - Klärfall-Bearbeitung that previously took 8 to 25 minutes per case drops to 1 to 3 minutes of review

2. Customer Service Triage and Rechnungsfragen

The biggest queue in the Kundenservice. A 100,000-Zählpunkt-Stadtwerk receives 500 to 1,500 Kundenanfragen per day, the vast majority routine: “wie hoch ist meine Abschlagszahlung jetzt”, “wo bleibt meine Jahresrechnung”, “ich ziehe um”, “wie hoch ist mein Verbrauch im Vergleich zum Nachbar”. The agent classifies the intent, pulls the relevant Vertrags- and Verbrauchsdaten and drafts the reply.

• Intent classification - Rechnungsfrage, Abschlagsänderung, Umzug, Tarifwechsel, Beschwerde, Netzanschluss-Anfrage, Stoerungsmeldung - jede mit eigenem Routing
• Grounded answers - The agent looks up the Vertragsakte, the laufenden Abschläge, the Jahresverbrauch and the Bilanzkreis-Lage before drafting
• Sparten-übergreifend - The agent handles Strom, Gas, Wärme and Wasser in one pipeline; the Kunde with three Sparten gets one consistent answer
• Tone matching - The draft uses the Stadtwerk’s standard tone, not a generic AI voice
• Human-send guardrail - Nothing leaves the Stadtwerk without a Sachbearbeiter pressing send; the agent prepares, the Mensch reviews
• Time saved - 40 to 60 percent of Kundenservice-Bearbeitungszeit disappears within the first month - the result the m3 / Fraunhofer IPK study highlights as the fastest-payback use case²⁰²¹

3. Netzanschluss and Hausanschluss Bearbeitung

The Netzanschluss-Antrag is where Stadtwerke lose 2 to 5 weeks per case. PV-Anlagen, Wallboxen, Wärmepumpen and Neubauten all require GIS-Lage-Prüfung, Netzentgelt-Berechnung, technische Plausibilität and ein Bescheid. The agent prepares all of it.

• Antrag-Eingang - The agent reads applications from email, the Kundenportal, the Installateur-Portal and the Bundesnetzagentur-Schnittstelle
• GIS-Lage-Prüfung - The agent queries the GIS (Smallworld, ArcFM, neuralabs) for Netzkapazität, Trafostation-Auslastung and Anschlussart
• § 14a EnWG-Klassifikation - Wärmepumpe, Wallbox or Hausspeicher classified against § 14a EnWG, with the entsprechenden Steuerbarkeit-Anforderungen flagged¹⁶
• Berechnung Netzentgelt + Anschlusskostenbeitrag - Generated against the local Netzentgelte-Tabelle and the firm-specific Berechnungsregel
• Bescheid-Entwurf - Generated in the firm template with all relevant Anlagen and Hinweisen, ready for Bauleitung approval
• Time saved - Netzanschluss-Bearbeitungszeit drops from 14 to 28 days down to 2 to 5 days end-to-end

4. Zählerwesen and Smart-Meter (iMSys)

The iMSys-Rollout under MsbG produces a continuous flood of MSCONS-Lastgangdaten in 15-minute resolution per Zählpunkt. As of mid-2025 about 1.6 million iMSys were installed nationwide; the rollout is behind targets but accelerating¹⁷. An agent ingests the MSCONS, runs plausibility, generates monthly Verbrauchsinformationen and reconciles dynamic Tarife under § 41a EnWG.

• MSCONS-Lastgang-Plausibilität - 15-minute MSCONS values per Zählpunkt scored against expected Lastprofile (H0, G0 etc.) and flagged for outliers
• SMGW-Administration-Reconciliation - The agent reads downstream of the BSI-zertifizierten SMGW-Administration; the GWA-Pfad stays intact and BSI-konform¹⁹
• Monatliche Verbrauchsinformation (§ 41a EnWG) - The agent generates the monthly UVI for every Kunde with iMSys with the comparison to the historical Verbrauch
• Dynamische Tarife - The agent reconciles MSCONS values against the dynamic Tarif-Struktur and generates the monthly Abrechnung
• § 14a EnWG-Steuerung - The agent tracks Steuerbarkeit-Anforderungen for Wärmepumpen, Wallboxen and Hausspeicher and reconciles with the Netzentgelt-Reduktion
• Time saved - Zählerwesen-Bearbeitung that previously took 3 to 5 minutes per Zählpunkt per month drops to under 30 seconds of review

5. Wechselprozesse (Lieferantenwechsel) and Tarif-Triage

The Lieferantenwechsel under GPKE is the regulated process at the heart of the Energiemarkt. Every wechsel-related UTILMD message must follow strict timelines (Fristen) under the BNetzA-Festlegung. The agent handles the full Wechsel-Choreography: An- und Abmeldung, Bilanzkreis-Zuordnung, Antwort-Tracking, Eskalation.

• An- und Abmeldungs-Bearbeitung - UTILMD-Nachrichten zu An- und Abmeldungen automatisch klassifiziert und beantwortet
• Fristen-Tracking - The agent tracks the GPKE-Fristen per case and escalates before the deadline
• Bilanzkreis-Zuordnung - The agent assigns the new Lieferant to the correct Bilanzkreis and writes back to the ERP
• Wechsler-Rückgewinnung - The agent identifies high-margin Kunden in the Abmeldungs-Pipeline and triggers Vertriebs-Aktion before the Frist expires
• Tarif-Triage - For Kunden-Anfragen the agent runs Tarif-Vergleich against the firm-eigenen Tarife (Grundversorgung, Sondertarife, dynamische Tarife)
• Time saved - Wechselprozess-Bearbeitung per Kunde drops by 50 to 70 percent

6. Forderungsmanagement and Inkasso

The Mahnwesen is high-volume, rule-bound and a known Sachbearbeiter time-sink. An agent reads the OPOS per Kunde and Sparte, applies the firm-specific Mahnstufen-Logik, drafts the Mahnschreiben and triggers Sperr-Verfahren only where the law and firm policy allow.

• OPOS-Reconciliation - Reads the OPOS per Kunde across Strom, Gas, Wärme, Wasser and consolidates the Forderung
• Mahnstufen-Automatik - Stufe 1 friendly reminder, Stufe 2 Mahnung, Stufe 3 Mahn-Eskalation, Stufe 4 Sperr-Ankündigung - per firm-specific rules
• Härtefall-Erkennung - The agent flags Härtefälle (Sozialhilfeempfänger, BG-Kunden) before the Sperr-Verfahren and routes them to Sozial-Sachbearbeitung
• Ratenzahlungs-Vorschlag - For overdue accounts the agent proposes a Ratenzahlungsplan within firm policy
• Sperr- and Entsperrungs-Workflow - The agent prepares the Sperr-Anweisung and the Entsperrungs-Bedingungen with full audit trail
• Time saved - Mahnwesen-Sachbearbeitung drops by 60 to 75 percent

The SAP IS-U / Wilken / Schleupen Architecture

In Germany the Stadtwerk-ERP landscape is split: SAP IS-U / S/4HANA Utilities dominates the larger Stadtwerke, Wilken ENER:GY and Schleupen.CS dominate the Mittelstand, plus Robotron robotron*ecount, Lima Energiewirtschaftssoftware, kVASy and a long tail of branch-specific tools. On top sits the MaKo-Plattform (Seeburger BIS, Robotron MaKo, BCS, Procilon, ponton-x) and the GIS (Smallworld, ArcFM, neuralabs). Any AI agent that wants to be useful in a Stadtwerk has to live on top of these systems without trying to replace them.

The five integration layers

1. Marktpartner-Eingang - MaKo-Plattform (Seeburger, BCS, Robotron, Procilon, ponton-x), customer email, customer portal, Installateur-Portal, BNetzA-Schnittstelle, Smart-Meter-Gateway-Administration - everything routed into one normalised inbox.
2. Klassifikation und Extraktion - The agent identifies the Marktpartner, the Sparte, the Zählpunkt, the Geschäftsprozess and extracts structured fields with confidence scores.
3. Vorschlag - The agent applies the firm-specific Tarif-Logik, Bilanzkreis-Zuordnung, Netzentgelt-Tabelle and Geschäftsprozess-Vorlagen and proposes EDIFACT-Antwort, Kunden-Antwort, Bescheid or Mahnung.
4. ERP- und MaKo-Schnittstelle - Approved messages flow into SAP IS-U / Wilken / Schleupen.CS via API, OData or import templates; EDIFACT-Antworten flow into the MaKo-Plattform for sending; nothing bypasses the system of record.
5. Audit und Reporting - Every action logged with timestamp, document hash, agent version and reviewer name - the audit trail for a BNetzA-Anhörung and the Stadtwerk-Steuerungs-Dashboard share the same event log.

What sits where

The principle is simple: the ERP, the MaKo-Plattform and the BSI-zertifizierte SMGW-Administration stay the system of record and the regulated path. The agent is the operator. Any architecture that tries to replicate the ERP or to bypass the BSI-Pfad breaks compliance on day one.

Where the data lives

• Marktpartner- und Kundendaten - In SAP IS-U / Wilken / Schleupen and the MaKo-Plattform, untouched. The agent works on a controlled copy via approved interfaces.
• Smart-Meter-Daten - In the BSI-zertifizierten GWA, untouched. The agent reads downstream of the GWA where MSCONS-Daten arrive in the EMT-Sphäre.
• Agent state - In an EU-resident environment (Frankfurt, Berlin or similar) operated under a DSGVO-konformen AVV that explicitly covers Energieverbrauchsdaten.
• LLM inference - Either via EU-resident endpoints (Azure OpenAI EU, AWS Bedrock EU, Anthropic Claude on AWS EU, or Mistral) or on a Stadtwerk-private deployment for sensitive Konzern- und Kundendaten.
• Prompt/output retention - Logged for audit. Not used for vendor model training - the AVV must say so explicitly.
• Backups - In EU, encrypted at rest, with documented retention aligned to BNetzA-Festlegungen, handelsrechtliche Aufbewahrungsfristen and project-specific contract requirements.

The 60-Day Pilot Playbook

The biggest mistake Stadtwerke make with AI is starting with a strategy. The right path is the opposite: start with one Sparte and one use case, 60 days.

The phases

1. Days 1-10: Scope - Pick the Sparte and the use case. The right starting point is Strom + Customer Service Triage or Strom + MaKo-Klärfall. Map the current Klärfall-Bearbeitung or Kundenservice-Prozess step by step. Document the typische Klärfall-Klassen and the 30 most frequent Customer-Service intents.
2. Days 11-20: Datenfundament - Audit Marktpartner-Stammdaten, the MaKo-Plattform-Konfiguration, the Tarif-Logik and the Geschäftsprozess-Vorlagen. Clean the obvious mess (alte MPIDs, doppelte Marktpartner). The agent will be only as good as this input.
3. Days 21-35: Build - Connect the agent to the ERP (SAP IS-U, Wilken, Schleupen.CS) and to the MaKo-Plattform (Seeburger, BCS, Robotron). Train it on the last 90 days of Klärfälle and Kundenanfragen. Run a shadow mode where every Antwort is generated but not sent.
4. Days 36-50: Parallel Pilot - Go live with Sachbearbeiter approval required on every outgoing EDIFACT message and customer answer. The agent posts; the human reviews. Track accuracy, time-per-Klärfall and exception rate. Tune the confidence threshold.
5. Days 51-60: Confidence Raise - For Marktpartner where the agent has been right 20+ times in a row, auto-approve routine UTILMD-Antworten. For Beschwerden and Eskalationen, always keep human review. Document the Verfahrensdokumentation for BNetzA-Anhörung.
6. Day 60+: Expand - Once Strom + first Use Case runs cleanly, add the second use case on the same Sparte, then expand to Gas, Wärme and Wasser one by one.

Checklist before you start

What to measure

• Time per MaKo-Klärfall - Baseline manual time (typically 8 to 25 minutes), target with agent (typically 1 to 3 minutes of review)
• Time per Customer Service inquiry - Baseline (3 to 10 minutes), target with agent (under 1 minute of review)
• Netzanschluss-Bearbeitungszeit - Baseline (14 to 28 days), target with agent (2 to 5 days end-to-end)
• Accuracy - Antworten and Klärfälle the human accepts unchanged - target 80%+ after week 4, 95%+ after week 8
• Exception rate - Items the agent kicks back as low-confidence - track the trend, not the absolute
• Fristen-Einhaltung - GPKE/GeLi-Gas-Fristen einhalten - target 100% within Fristen
• Kunden-Zufriedenheit - Spot-check the pilot Kunden after 60 days - faster responses, fewer dropped Wechsel, faster Bescheide

MaKo 2026, MsbG and the EU AI Act

Compliance is the make-or-break in a Stadtwerk. Every other consideration sits below it. The Versorger profession has four overlapping rule sets that all touch AI: BNetzA-Festlegungen (MaKo, GPKE, GeLi Gas, WiM), MsbG (Smart Meter), DSGVO and the EU AI Act. Each has a concrete answer; none of them is a blocker.

MaKo 2026 - the regulated data exchange

The MaKo 2026 release (BNetzA Mitteilung Nr. 34) takes effect on 1 April 2026. BET Consulting frames it as a Stabilisierungs-Release, not a Revolution - but the EDIFACT format updates require every Marktteilnehmer to be ready¹⁰⁸. The agent natively parses the new formats and stays in sync with each subsequent BNetzA-Mitteilung.

• UTILMD (Strom and Gas) - Stammdaten-Austausch für Marktpartner, Verträge, Zählpunkte¹³
• MSCONS - Messwert-Übermittlung, 15-Minuten-Lastgänge, Smart-Meter-Daten¹³
• COMDIS - Klärfall-Kommunikation zwischen Marktpartnern
• ORDERS / ORDRSP - Geschäftsprozesse: Wechsel, An- und Abmeldungen, Stammdatenänderungen
• REMADV / INVOIC - Zahlungsavise und Rechnungen zwischen Marktpartnern
• PARTIN - Marktpartner-Stammdaten-Aktualisierung
• BNetzA-Festlegung GPKE - Geschäftsprozesse zur Kundenbelieferung mit Elektrizität - the regulated workflow for every Lieferantenwechsel¹²
• BNetzA-Festlegung GeLi Gas - Same for Gas, with Gas-specific timelines and message types

MsbG - Smart-Meter and iMSys

The Messstellenbetriebsgesetz mandates the iMSys rollout: end consumers with annual consumption between 6,000 and 100,000 kWh and § 14a EnWG-controllable devices (Wärmepumpen, Wallboxen, Hausspeicher) must receive an intelligentes Messsystem¹⁴¹⁵. The agent reads downstream of the BSI-zertifizierten SMGW-Administration and never bypasses the regulated path.

• iMSys = mME + SMGW - The intelligente Messsystem consists of moderne Messeinrichtung plus Smart-Meter-Gateway under BSI-Schutzprofil BSI-CC-PP-0073/0077¹⁹
• § 29 MsbG Pflichtkundengruppen - Endkunden 6,000-100,000 kWh AND § 14a EnWG-Geräte must receive iMSys¹⁴
• § 41a EnWG dynamische Tarife - From 2025 Lieferanten must offer dynamische Tarife to iMSys-Kunden; the agent computes the monthly Abrechnung against 15-min MSCONS
• BSI-Pfad bleibt intakt - The agent never connects to the SMGW directly; it reads MSCONS downstream where the data is already in the EMT-Sphäre
• Monatliche Verbrauchsinformation - § 41a EnWG verlangt monatliche UVI für iMSys-Kunden - the agent generates and dispatches

DSGVO and BSI-Compliance

Energieverbrauchsdaten are personenbezogene Daten under DSGVO. The Schutzbedürfnis is high - a 15-minute Lastgang reveals when the Kunde is at home, when she cooks, when he charges his Wallbox. The agent must respect both DSGVO and BSI-Vorgaben.

• EU hosting only - Energieverbrauchsdaten processed by the agent stay in EU data centres
• AVV with explicit Energieverbrauchsdaten clause - Cover personenbezogene Verbrauchsdaten under DSGVO Art. 28 and reference the BSI-Schutzprofil for the SMGW
• No training on Kundendaten - Vendor agreement must include an explicit no-training clause covering prompts and outputs
• BSI-Pfad nicht umgehen - The agent never sits inside the GWA-Pfad; it operates in the EMT-Sphäre where the data is already cleared
• Audit logs available to BNetzA on request - Every agent action available for the BNetzA-Anhörung cycle

EU AI Act - what applies, what does not

The EU AI Act becomes fully applicable on 2 August 2026. Energy supply is a critical infrastructure but the standard Versorger-AI-Anwendungsfälle (MaKo, Service, Netzanschluss, Smart-Meter, Forderungen, Wechsel) are mostly limited-risk: transparency, AI literacy under Article 4, general-purpose model obligations on the vendor side²⁷²⁸. The exception is automated Bonitäts-Entscheidung on Kunden or automated Sperrung ohne menschliche Prüfung.

• Article 4 (AI Literacy) - Every Sachbearbeiter using the agent receives basic AI literacy training, documented²⁸
• Transparency obligations - Kunden are informed that AI is used in the preparation of communication and Bescheide; the Sachbearbeiter remains the responsible person
• Provider obligations - The LLM vendor (OpenAI, Anthropic, Mistral, etc.) carries the general-purpose model obligations; the Stadtwerk is the deployer, not the provider
• No high-risk classification (standard use) - MaKo, Service Triage, Netzanschluss, Smart-Meter and Forderungsmanagement are not Annex III high-risk
• High-risk exception - Automated Bonitäts-Entscheidung or automatische Sperr-Entscheidung ohne menschliche Prüfung ARE Annex III - do not deploy without Konformitätsbewertung

More Customers Without More Sachbearbeiter

The Fachkräftemangel is not going away. Hiring two more MaKo-Sachbearbeiter or Kundenservice-Agents is harder, slower and more expensive every year - and the Energie-Specialisten-Pool is narrower than the general Sachbearbeiter-Markt. The question that matters for Stadtwerk-Geschäftsführung over the next three years is not “wie spare ich Zeit” - it is “wie wachse ich ohne neue Köpfe”.

What an agent does to the capacity math

The economic logic

• The Kunde pays for outcomes - The Kunde pays the same Arbeitspreis whether the MaKo-Bearbeitung was manual or automated. The agent shifts cost without shifting price.
• Throughput becomes the moat - In a Fachkräftemangel market the Stadtwerk that can absorb more Marktpartner and Zählpunkte without proportional hiring wins consolidation opportunities
• Wechsler-Rückgewinnung is the margin - Stadtwerke routinely lose 30 to 50 percent of justified Wechsler-Rückgewinnungs-Chancen because the Sachbearbeiter cannot react inside the Wechsel-Frist. AI agents catch every signal.
• Retention compounds - Kunden who get same-day responses and 5-day Netzanschluss do not switch Anbieter. The retention curve flattens.
• Consolidation pressure is real - Stadtwerke Bielefeld reduced 300 Stellen in 2025; the next wave of consolidation favours the Stadtwerk that runs on modern infrastructure

How Superkind Fits

Superkind builds custom AI agents that sit on top of the systems Stadtwerke already use - SAP IS-U / S/4HANA Utilities, Wilken ENER:GY, Schleupen.CS, Robotron, Lima, kVASy - plus the MaKo-Plattform (Seeburger BIS, BCS, Robotron MaKo, Procilon, ponton-x), the GIS (Smallworld, ArcFM) and the Smart-Meter-Gateway-Administration - without forcing a switch. The deployment model is process-first: we map your MaKo-Klärfall-Logik, your Customer Service intents and your Netzanschluss-Workflow before we touch a line of code. The agent is built around your Stadtwerk’s reality, not a generic template.

What sits in the Superkind agent for a Stadtwerk

What Superkind brings to a Stadtwerk

• Process-first deployment - We map MaKo-Klärfall-Logik, Customer Service intents and Netzanschluss-Workflow before we build, so the agent fits your Stadtwerk’s reality
• ERP- und MaKo-native architecture - The agent works through SAP IS-U OData, Wilken / Schleupen APIs and the standard MaKo-Plattform-Schnittstellen (Seeburger, BCS, Robotron, Procilon, ponton-x)
• MaKo-2026-ready compliance - UTILMD, MSCONS, COMDIS, ORDERS, ORDRSP, REMADV, PARTIN, INVOIC against the BNetzA Mitteilung 34 dictionary
• Marktpartner-specific learning - Each Marktpartner’s Klärfall-Patterns, Stammdaten-Quirks and Antwort-Konventionen become rules the agent applies automatically
• Sparten-übergreifend - Strom, Gas, Wärme and Wasser in one pipeline - the Kunde with three Sparten gets one consistent answer
• BSI-Pfad respektiert - The agent reads downstream of the BSI-zertifizierten SMGW-Administration; the regulated path stays intact
• EU-hosted, DSGVO-ready - Frankfurt or Berlin hosting, AVV with explicit Energieverbrauchsdaten-Klausel, no-training, full audit logs
• Human-in-the-loop by design - Confidence thresholds are configurable; nothing auto-sends an EDIFACT to a Marktpartner below the line you set
• 60-day pilot scope - One Sparte, one use case, written success criteria - go or no-go after 60 days
• Long-term partnership model - We stay involved beyond the pilot; the agent evolves with each BNetzA-Mitteilung

Decision Framework

Not every Stadtwerk is ready, and not every Versorger needs a custom agent. The framework below helps locate which one you are.

Start with ERP-AI and Kundenportal-Chatbot if

• Under 10,000 Zählpunkte - Throughput is not yet the constraint; a horizontal Chatbot is enough
• Under 50 employees in Service und MaKo - The math does not yet justify a custom build
• No growth or consolidation pressure - Status quo is acceptable for the next 12 months
• You want to test AI broadly before committing - The SAP IS-U AI or the Schleupen.CS-Assistant is the lowest-friction starting point

Move to a custom agent if

• 50,000+ Zählpunkte across multiple Sparten - Volume and Sparten-Komplexität produce the ROI mathematics
• 80+ employees with MaKo and Service load - Coordination overhead and Klärfall-Volumen justify the build
• Hiring is blocked - You have tried to hire MaKo-Sachbearbeiter and Energie-Specialisten and the candidates are not arriving
• MaKo-Klärfall-Quote is rising - You are bleeding margin on Wechsler-Rückgewinnung you cannot react to in time
• iMSys-Rollout is accelerating - You see the MSCONS-Lawine coming and need the agent before the Lastgang-Flut
• Sanierungs- or Wachstumswelle in Netzanschluss - PV-Boom, Wärmepumpen-Welle, Wallbox-Pflicht create application volume your team cannot handle manually
• Consolidation or succession on the 3-year horizon - Modernised Stadtwerke are stronger Konsolidierungs-Partner

Wait if

• Marktpartner-Stammdaten are a mess - Clean MPIDs and PARTIN first; the agent will not save you
• No MaKo-Plattform-API access - Sort the technical access with Seeburger, BCS or Robotron before the project
• Resistance from the IT-Leitung - The IT-Leitung has to sponsor the project; without that, it stalls
• No budget for AVV review - The legal foundation is non-negotiable; budget the lawyer hours and the BSI-konforme Architektur

Frequently Asked Questions

Related Articles

• AI for Direct Marketing of Renewables and PPA Optimization
• AI for Property Management Firms: How German Hausverwalter Tackle WEG-Beschluesse and Nebenkostenabrechnung
• AI Agents on SAP: How the Mittelstand Builds Agents That Work Inside SAP
• AI for the Email Inbox: How the Mittelstand Triages 200 Daily Mails Without Hiring
• Intelligent Document Processing: How the Mittelstand Replaces OCR With AI Agents
• AI in Customer Service: How the Mittelstand Builds Knowledge-Grounded Agents
• AI Agent ROI: The KPI Framework That Convinces CFOs in 90 Days

Henri Jung

Co-founder of Superkind, where he helps SMEs and enterprises deploy custom AI agents that actually fit how their teams work. Henri is passionate about closing the gap between what AI can do and the value it creates in real companies. He believes the Mittelstand has everything it needs to lead in AI - it just needs the right approach.