AI Agents for Construction Companies: How the German Bau-Mittelstand Wins Back Aufmaß, Nachträge and Bauleitung

It is a Thursday evening in May. The Bauleiter of a mid-size Hochbau company in Nordrhein-Westfalen finishes the last Baustellenbegehung of the day, climbs into the Crafter, and pulls his phone out of his Warnweste. There are 28 unread WhatsApp messages from three Polieren, four photos of a Mauerwerk that “sieht komisch aus”, an email from the Auftraggeber asking why Position 03.45 from the LV has not been measured yet, and a calendar reminder for the Bautagebuch he has not started. The Frau at home has stopped asking when he will be back.

The Bau-Mittelstand has been talking about Digitalisierung for a decade. The math has finally caught up with the talk. The Hauptverband der Deutschen Bauindustrie reports that 60 percent of construction firms still cite a Fachkräftemangel as the dominant constraint, around 120,000 skilled workers will be missing by 2030, and Insolvenzen in the Bauhauptgewerbe rose 5.5 percent in the first ten months of 2025¹². Hiring has stopped being a strategy. Throughput on the Baustelle has become the constraint.

This guide is for the Inhaber, the Bauleiter and the Kalkulator 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 Aufmaß from PDF plans, draft VOB-konforme Nachträge, compile the Bautagebuch from voice notes and site photos, and feed the Bau-ERP - and if so, how?

The Bau-Mittelstand Squeeze

The reason building work consumes so much time is rarely the building work itself. It is everything around it: chasing prices from suppliers, decoding LV positions written in 1997, translating a Polier voice note into a Bautagebuch entry, looking up the right VOB-Paragraph for a Nachtrag, and re-confirming a Mengenermittlung that the Bauleiter already confirmed two weeks ago. Multiply that by 8 parallel Baustellen and the math gets ugly fast.

• 120,000 missing skilled workers by 2030 - The Bauindustrie projects a structural shortfall of roughly 120,000 Facharbeiter and Bauleiter across the trade by the end of the decade².
• 60 percent of firms still report acute shortage - The HDB Fachkräftesituation 2025/2026 shows 60 percent of construction companies with reported difficulties; in Tiefbau the share is 62 percent and slightly rising².
• Insolvenzwelle continues into 2026 - Insolvencies in the Bauhauptgewerbe rose 5.5 percent in January-October 2025, with the major Wohnungsbau wave expected to continue into the second half of 2026³.
• Productivity gap is global - McKinsey reports construction productivity growth averaged just 0.4 percent per year between 2000 and 2022, against 2.8 percent in the wider economy; the gap is the largest improvement opportunity of any sector¹⁶.
• 41 percent AI threshold reached - Bitkom’s 2026 Pressemitteilung “Durchbruch bei Künstlicher Intelligenz” finds 41 percent of German firms (20+ employees) now use AI actively, up from 17 percent in 2024 - construction lags but is catching up fast¹⁷.
• BIM-Stufenplan deadlines bite - Level II BIM is mandatory for Bundesbauten over EUR 0.5 million from 2025; Level III for projects over EUR 50 million from mid-2025; the wave reaches Mittelstand contractors through their Auftraggeber⁶.
• The Auftraggeber is changing too - Industrial and public clients have been trained by Procore and Dalux to expect digital deliveries: BIM models, structured Aufmaß-Blätter, time-stamped Bautagebücher. A paper-and-Excel contractor loses tenders.

The result is a Bauunternehmen where the most senior Bauleiter spend the largest share of their time on the most repetitive work, while Verhandlungsführung, Subunternehmer-Steuerung and Bauablaufoptimierung get squeezed into the last hour of the day. 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 Bauzaun, wants to do.

What an AI Agent Actually Does on a Baustelle

The market is full of products labelled as “KI für die Baustelle”. Most of them are PDF-OCR with a polish, or a Telegram-Bot that transcribes voice notes. A real AI agent goes much further: it owns the workflow, not just the data extraction. Here is the difference, in plain Bauleiter terms.

The agent loop on a Baustelle

1. Capture - Pulls inputs from Pläne (PDF, BIM, DWG), Leistungsverzeichnisse (GAEB-XML, Excel, PDF), supplier emails, Polier voice notes, site photos, Bautagebuch-App, Bedenkenanzeigen and the Bau-ERP.
2. Classify - Reads the input - a drawing change, a voice note from the Polier, an email from the Auftraggeber, a Lieferschein photo - identifies the Baustelle, the Gewerk and the Position, and routes it to the right pipeline.
3. Extract - Pulls structured fields: Mengenermittlung from drawings (m², m³, lfm, Stück), Anordnung details from emails, Wetterdaten and Beleg-Stunden from voice notes, Mängel from photos - with a confidence score on every value.
4. Enrich - Looks up the Position in the Auftrags-LV, retrieves the Urkalkulation, checks the VOB clause, references the Bauzeitenplan, applies the firm’s Nachtrags-Vorlagen.
5. Propose - Generates the full Aufmaß-Blatt, the Nachtragsangebot with § 2 Abs. 5 or § 2 Abs. 6 reference, the Bautagebuch entry, or the supplier order - ready for Bauleiter review.
6. Decide - Above the confidence threshold the agent posts directly into the Bau-ERP; below it the case lands in the Bauleiter review queue with a one-click approve or correct.
7. Learn - Every correction feeds back into the agent. Patterns the agent sees three times on a specific Baustelle become rules it applies automatically.
8. Audit - Every step is logged with timestamp, document hash, agent version and (on review) the human who approved it - the BGB and VOB audit trail is built by the system, not by the Bauleiter after the fact.

The difference from what you have today

The category matters. PlanRadar dominates Mängelverwaltung, Capmo the Kanban-style Projektmanagement, Dalux the BIM-bereich, 123erfasst the German Mittelstand with ERP-Anbindung¹¹. All four are excellent at their specific job. But in 7 out of 8 cases the Bauleiter still has to assemble the Bautagebuch manually in the evening - the tool collects photos and tasks, the human writes the report¹¹. A custom agent goes one layer further: it removes the work entirely from the human review path until something looks unusual or judgement-dependent.

6 Use Cases That Work Today

Not every Baustellen-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 Baustelle in one use case. Add the next once the first runs cleanly.

1. Aufmaß aus Plänen (Quantity Takeoff from Drawings)

The biggest single time sink in Kalkulation and Abrechnung. An agent reads PDF plans, DWG files or IFC/BIM models, identifies the Gewerk and the LV-Position, and generates the Aufmaß-Blatt with line items, quantities and confidence scores - ready for the Kalkulator to review.

• Volume is high - A typical Hochbau project has 200 to 1,200 LV-Positionen and 50 to 300 separate measurements per Gewerk
• Patterns are stable - Once the agent has seen a Bauweise three times (e.g., Mauerwerk KS 24, Estrich ZE 20, Dämmung WLG 035), it Aufmaß-es the fourth instance automatically
• ROI is fast - University of Kansas case studies show up to 76 percent time savings against manual takeoff; Coastal Construction reported takeoff dropping from 50 percent to 10 percent of the workweek - 14.5 hours saved per Plansatz²¹
• VOB/C compliant - DIN 18299 Abschnitt 5 explicitly permits Mengenermittlung from existing drawings if sufficient; the agent produces auditable evidence per line⁹
• BIM-ready - For BIM Stufe II and III deliveries, the agent maps quantities to IFC structures the Auftraggeber expects
• Sideways win - The same Aufmaß used in Kalkulation gets reused for Schlussrechnung - one source of truth across the project life cycle

2. Nachtragsmanagement (VOB-konforme Change Orders)

The single most underestimated source of margin in the Mittelstand. A 2026 IWU analysis points out that recent jurisprudence demands gap-free documentation - digitales Bautagebuch, photos and Bauzeitenpläne - to defend Nachträge against unjustified rejection¹⁰. An agent extracts the Anordnung, references the matching VOB-Paragraph, pulls the price from the Urkalkulation, drafts the Nachtragsbegründung and assembles the Nachtragsangebot.

• Anordnung capture - The agent reads emails, Bauberatungsprotokolle, voice notes and Plan-Anmerkungen for trigger events
• Automatic VOB-Bezug - § 2 Abs. 5 (Anordnung des Auftraggebers), § 2 Abs. 6 (zusätzliche Leistung), § 6 Abs. 6 (Bauzeitverlängerung) classified per case²³
• Urkalkulation-Bindung - Pricing pulled from the original Kalkulation, not invented; the audit trail proves it
• Begründungstext - The agent drafts a Nachtragsbegründung in the firm’s standard tone and structure - the Bauleiter reviews, edits, signs
• Dispute-ready documentation - Every Nachtrag carries the Anordnung, the date, the matching plan revision, the price basis and the Bauzeit-Impact
• Time saved - Nachträge that previously took 90 to 180 minutes per case drop to 15 to 30 minutes of Bauleiter review

3. Bautagebuch (Daily Logs from Voice and Photo)

The single most universal pain in Bauleitung. The Bauleiter speaks 5 to 15 voice notes per day into the phone, takes 30 to 80 photos, and writes the actual Bautagebuch in the evening. The agent flips this: the Bauleiter speaks and snaps; the agent writes.

• Voice-to-Bautagebuch - Polier voice note “Heute Beton C30/37 bei Achse 4 ab 7:30, Wetter 12 Grad, Polier Schmidt” becomes a structured Tageseintrag
• Photo classification - Photos automatically tagged to Gewerk, Bauteil, Achse and date; Mängel surfaced as separate items
• Wetterdaten and Beleg-Stunden - Wetter pulled from the nearest weather station; Beleg-Stunden pulled from the Zeiterfassungs-App
• Subunternehmer-Anwesenheit - The agent reconciles which firms were on site against the Bauablaufplan
• Auftraggeber-ready format - Output formatted in the Bautagebuch structure the customer requires (BMG, BMVI standard, or firm-specific)
• Time saved - 30 to 60 minutes of Büro-Zeit per Baustelle per day disappears; for a Bauleiter on 4 Baustellen, that is up to 4 hours back¹¹

4. Bedenkenanzeige and Mängelmanagement

Bedenken nach § 4 Abs. 3 VOB/B and Mängel under § 13 are not paperwork; they are the gap between a fair settlement and a Schiedsgutachten. An agent monitors site photos, supplier emails and Polier reports for signals that warrant a written Bedenkenanzeige or a Mängelrüge.

• Signal detection - Photos showing Untergrund-Schäden, Plan-Widersprüche, missing Vorleistungen trigger a draft Bedenkenanzeige
• Auto-paragraph reference - § 4 Abs. 3 (Bedenken gegen vorgesehene Art der Ausführung) or § 13 (Mängel) applied per case
• Time-stamped delivery - The agent enforces written delivery before the Gewerk continues - critical for liability transfer
• Mängel-Listen aktualisierung - PlanRadar, Capmo, Dalux or 123erfasst Mängel-Listen kept up to date automatically
• Reminder logic - Open Mängel get Fristen-Erinnerungen pushed to the Polier and the Subunternehmer

5. Lieferantenkoordination und Materialbeschaffung

The Materialkalkulation broke in 2022 when Stahl, Holz and Energie prices moved by triple digits in months. It has not fully settled since. An agent watches supplier prices, parses incoming Angebote, matches them against the Bedarf and the Bauzeitenplan, and proposes Bestellungen.

• Multi-Lieferanten-Vergleich - Angebote from Sonepar, Würth, Frankenhof, regional Baustoffhändler parsed into one comparison table
• Bedarfs-Matching - The agent maps each Position to the LV and the Bauzeitenplan, flags Liefertermine vs. Baufortschritt
• Preisbindung-Checks - Rahmenvertragspreise checked against ad-hoc Angebote; deviations flagged
• Bestellvorschlag - The agent drafts the Bestellung in the firm’s format with Liefertermin, Lieferadresse, Ansprechpartner, Konditionen
• Lieferschein-Abgleich - On delivery the Polier photographs the Lieferschein; the agent reconciles against the Bestellung and the LV
• Time saved - 60 to 70 percent of supplier-coordination time recovered for the Bauleiter and the Einkauf

6. Bauleitungs-KPIs (Project Steering Dashboard)

Most Bauunternehmen fly blind on their own Baustellen-Margen until the Schlussrechnung. An agent watches Aufmaß, Nachträge, Materialeinsatz, Stunden and Subunternehmer-Leistungen across all Baustellen, then produces a weekly Bauleitungs-Dashboard - which Baustellen are profitable, which are slipping, which need a Geschäftsführer-Eskalation.

• Baustellen-Profitabilität - Soll-Ist auf Position-Level, surfaced weekly not monthly
• Nachtrags-Quote - Nachträge je Baustelle in EUR and percent of Auftragssumme; trending up or down
• Stunden-Burn - Beleg-Stunden vs Kalkulation per Gewerk; early warning on Polier-Überlast
• Material-Drift - Materialkosten Soll vs Ist with reason codes (Preis, Menge, Nachtrag)
• Subunternehmer-Performance - On-time, in-quality and price-discipline per Subunternehmer over the last 12 months
• Kapazitäts-Forecast - Next-quarter capacity projection based on Auftragsbestand and Bauablaufpläne

The Bau-ERP-First Architecture

In Germany, the Mittelstand Bau-ERP landscape is dominated by a handful of vendors: RIB iTWO (RIB Software), Nevaris (Nemetschek), BRZ, Bau Master, California.pro (G&W), MOS'aik and a long tail of branch-specific tools. Any AI agent that wants to be useful in a Bauunternehmen has to live on top of the Bau-ERP without trying to replace it. Here is what the working architecture looks like.

The five integration layers

1. Baustellen-Eingang - Pläne (PDF, DWG, IFC), GAEB-XML LVs, emails, supplier portals, Bautagebuch-App, voice notes, site photos - everything routed into one normalised inbox.
2. Klassifikation und Extraktion - The agent identifies the Baustelle, the Gewerk and the Position, and extracts the structured fields with confidence scores.
3. Vorschlag - The agent applies the firm-specific Kalkulationsregeln, Kostengruppen DIN 276 and Nachtrags-Vorlagen, and proposes Aufmaß, Nachtrag, Bautagebuch entry or Bestellung.
4. Bau-ERP-Schnittstelle - Approved items flow into the Bau-ERP via GAEB-XML, REST API or import templates; 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 Schiedsgutachten and the Bauleitungs-Dashboard share the same event log.

What sits where

The principle is simple: the Bau-ERP stays the system of record. The agent is the operator. Any architecture that tries to replicate the Bau-ERP in a side-system breaks the audit trail on day one and breaks the Bauleiter’s sign-off on day two.

Where the data lives

• Baustellendaten - In the Bau-ERP and the DMS, untouched. The agent works on a controlled copy via approved interfaces.
• Agent state - In an EU-resident environment (Frankfurt, Berlin or similar) operated under a DSGVO-compliant Auftragsverarbeitungsvertrag.
• LLM inference - Either via EU-resident endpoints (Azure OpenAI EU, AWS Bedrock EU, Anthropic Claude on AWS EU, or Mistral) or on a firm-private deployment for sensitive Bautagebuch and Mandantendaten.
• 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 handelsrechtliche Aufbewahrungsfristen and project-specific contract requirements.

The 60-Day Pilot Playbook

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

The phases

1. Days 1-10: Scope - Pick the Baustelle. The right one is mid-size (EUR 2 to 10 million Auftragswert), recurring Gewerke (Hochbau or Tiefbau, not a one-off Sonderbau), an Auftraggeber willing to be referenced, and a Bauleiter who wants to be the pilot lead. Map the current Aufmaß and Nachtrags-Prozess step by step.
2. Days 11-20: Datenfundament - Audit Stammdaten (Lieferanten, Materialnummern, Kostenstellen), the GAEB-XML structure of the Auftrags-LV, the Urkalkulation, recurring Bauweisen. Clean the obvious mess. The agent will be only as good as this input.
3. Days 21-35: Build - Connect the agent to the Bau-ERP via GAEB-XML, REST API or import templates. Train it on the last 90 days of Aufmaß and Nachträge from the firm. Run a shadow mode where every Aufmaß-Blatt and Nachtragsangebot is generated but not submitted.
4. Days 36-50: Parallel Pilot - Go live with Bauleiter approval required on every output. The agent drafts; the human reviews. Track accuracy, time saved and exception rate. Tune the confidence threshold.
5. Days 51-60: Confidence Raise - For Gewerke where the agent has been right 20+ times in a row, auto-approve the Aufmaß-Blatt. For Nachträge above EUR 5,000, always keep human review. Document the Verfahrensdokumentation.
6. Day 60+: Expand - Once one Baustelle runs cleanly, add the next two. Once five Baustellen run cleanly, add the second use case (typically Bautagebuch and Lieferantenkoordination).

Checklist before you start

What to measure

• Time per Aufmaß-Blatt - Baseline manual time (typically 30 to 120 minutes per Gewerk), target with agent (typically 5 to 15 minutes of review)
• Time per Nachtrag - Baseline (90 to 180 minutes), target with agent (15 to 30 minutes of review)
• Bautagebuch time per Baustelle per day - Baseline (30 to 60 min), target (under 10 min of review)
• Accuracy - Aufmaß lines that the Kalkulator 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
• Cycle time - Plan-Eingang to Aufmaß-Blatt-Fertig - target 1-3 days
• Auftraggeber-Zufriedenheit - Spot-check the pilot Auftraggeber after 60 days - faster Nachträge, fewer disputes, time-stamped Bautagebuch

VOB, GAEB and the EU AI Act

Compliance is the make-or-break in a Bauunternehmen. Every other consideration sits below it. The trade has three overlapping rule sets that all touch AI: VOB (A, B, C), GAEB-XML standards and the EU AI Act. Each has a concrete answer; none of them is a blocker.

VOB Teil B - the contract spine

The VOB/B is the operating manual for German Bau-Verträge. The agent has to respect three core paragraphs: § 2 (Vergütung and Nachträge), § 4 (Bedenken and Mitwirkung) and § 6 (Behinderung und Unterbrechung)²³. The 2026 IWU analysis stresses that recent jurisprudence requires gap-free documentation; an AI agent generates that documentation as a side effect of doing the work¹⁰.

• § 2 Abs. 5 - Änderung der Leistung - The agent links every change to a written Anordnung and to the Urkalkulation for pricing
• § 2 Abs. 6 - Zusätzliche Leistungen - The agent identifies which items are within the original LV and which require a separate Nachtragsangebot
• § 4 Abs. 3 - Bedenken - The agent surfaces Plan-Widersprüche and Vorleistungs-Mängel as draft Bedenkenanzeigen for Bauleiter sign-off
• § 6 Abs. 6 - Bauzeit-Anspruch - The agent links Bauzeit-Verlängerungen to documented Behinderungen and to the Bauzeitenplan-Impact
• § 13 - Mängelhaftung - The agent tracks Mängelrügen, Fristen and Nacherfüllungsangebote with timestamps

VOB Teil C and Aufmaß-Disziplin

VOB/C (DIN 18299 ff.) governs the technical Vertragsbedingungen for individual Gewerke. The DIN 18299 Abschnitt 5 update explicitly permits Mengenermittlung from existing drawings if sufficient - which is the legal basis for AI-generated Aufmaß⁹. The agent records the source plan, scale, page and revision for every line item, which is exactly what a Schiedsgutachten or a court will want to see.

• Mengenermittlung aus Plänen - Explicitly permitted under DIN 18299 if drawings are sufficient⁹
• Trade-specific DIN - DIN 18331 Beton, DIN 18330 Maurer, DIN 18350 Putz, DIN 18353 Estrich - each has its own Mengenermittlungs-Regeln that the agent applies per Gewerk
• Aufmaß-Blatt audit trail - Every quantity carries plan page, scale, agent version, timestamp and reviewer
• Reuse for Schlussrechnung - The same Aufmaß-Blatt feeds Kalkulation, Zwischenrechnung and Schlussrechnung - one source of truth
• Auftraggeber-readable format - Output in GAEB DA 89 (Aufmaß) and the firm-specific PDF template the Auftraggeber expects

GAEB-XML - the data backbone

GAEB-XML is the de facto standard for digital exchange of LV, Aufmaß and Abrechnung between Auftraggeber, Planer and Bauunternehmen²⁷. Every serious AI agent for the Bau-Mittelstand must read and write GAEB DA 81, 82, 83, 84, 86, 89 and 90.

• GAEB DA 81/82/83 - Leistungsverzeichnis-Übergabe - read for Kalkulation, Aufmaß-Vorbereitung
• GAEB DA 84 - Angebotsübergabe - generated for Auftraggeber after Kalkulation
• GAEB DA 86 - Nebenangebot - generated for variant offers
• GAEB DA 89 - Aufmaß-Übergabe - generated from AI Aufmaß for Auftraggeber-Prüfung
• GAEB DA 90 - Rechnungsübergabe - generated for Zwischen- and Schlussrechnung
• Round-trip safety - The agent must produce GAEB-XML that passes Auftraggeber-side Validatoren without manual fix-up

BIM-Stufenplan and Bundesbauten

The BIM-Stufenplan of BMDV and BMWSB makes Level II BIM mandatory for federal Bundesbauten over EUR 0.5 million from 2025, and Level III for projects over EUR 50 million from mid-2025; smaller projects follow by 2027⁶⁷. Mittelstand contractors bidding on Bundesbauten or large industrial Auftraggeber are increasingly required to deliver BIM-konforme Daten.

• Level I - 2D-Plan-basierte Zusammenarbeit, AIA (Auftraggeber-Informations-Anforderungen) light
• Level II - Föderierte BIM-Modelle, AIA voll, BAP (BIM-Abwicklungsplan), IFC-basierter Austausch
• Level III - Common Data Environment, gemeinsame Modellführung, voll digitalisierter Lebenszyklus
• Agent bridges the gap - The agent reads 2D PDF plans and generates BIM-konforme Mengenermittlung, mapping quantities to the IFC structure of the Auftraggeber
• Bid eligibility - Mittelstand firms without BIM-Fähigkeit lose Bundesbau-Aufträge - the agent is the bridge while the firm builds in-house competence

EU AI Act - what applies, what does not

The EU AI Act becomes fully applicable on 2 August 2026. For a Bauunternehmen deploying AI agents, the relevant articles are mostly limited-risk: transparency, AI literacy under Article 4, and the general-purpose model obligations on the vendor side²⁴²⁵. The agent itself is typically not a high-risk system because Aufmaß, Nachtrag and Bautagebuch use cases are not Annex III categories.

• Article 4 (AI Literacy) - Every Bauleiter, Kalkulator and Polier using the agent receives basic AI literacy training, documented²⁵
• Transparency obligations - Auftraggeber are informed that AI is used in Aufmaß and Nachträge preparation; the Bauleiter remains the responsible person
• Provider obligations - The LLM vendor (OpenAI, Anthropic, Mistral, etc.) carries the general-purpose model obligations; the Bauunternehmen is the deployer, not the provider
• No high-risk classification - Aufmaß, Nachtrag, Bautagebuch and Lieferantenkoordination are not Annex III high-risk use cases
• DSGVO for site photos - Bautagebuch photos that show identifiable people require a DSGVO concept; the agent supports automatic Gesichts-Unkenntlichmachung where needed

The Growth Question: More Baustellen Without More Bauleiter

The Fachkräftemangel is not going away. Hiring two more Bauleiter is harder, slower and more expensive every year. The question that matters for Inhaber 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 Auftraggeber pays for outcomes - The Auftraggeber pays the same EUR per Position whether the Aufmaß was manual or AI-assisted. The agent shifts cost without shifting price.
• Throughput becomes the moat - In a Fachkräftemangel market, the firm that can run more Baustellen without proportional hiring wins the marginal Auftrag
• Nachträge are the margin - Mittelstand Bauunternehmen routinely lose 30 to 50 percent of justified Nachträge because they cannot document them in time. AI agents capture them all.
• Retention compounds - Auftraggeber who get next-day Nachtragsangebote and live Aufmaß-Verfügbarkeit do not switch firms. The retention curve flattens.
• Succession planning gets easier - A 50-person firm running 12 Baustellen on modern infrastructure is salable. A 50-person firm running 6 Baustellen on paper and Excel is not.

How Superkind Fits

Superkind builds custom AI agents that sit on top of the systems Bauunternehmen already use - RIB iTWO, Nevaris, BRZ, Bau Master, California.pro, MOS'aik, plus Capmo, PlanRadar, Dalux and 123erfasst on the Baustelle - without forcing a switch. The deployment model is process-first: we map your Aufmaß, your Nachtrags-Workflow and your Bautagebuch process before we touch a line of code. The agent is built around your firm’s reality, not a generic template.

What sits in the Superkind agent for a Bauunternehmen

What Superkind brings to a Bauunternehmen

• Process-first deployment - We map Aufmaß, Nachtrags-Workflow and Bautagebuch before we build, so the agent fits your firm’s reality
• Bau-ERP-native architecture - The agent works through GAEB-XML, REST APIs and the import templates of RIB iTWO, Nevaris, BRZ, Bau Master and the others. No replacement, no parallel system.
• VOB-ready compliance - § 2 Abs. 5, § 2 Abs. 6, § 6 Abs. 6 templates with full audit trail for Schiedsgutachten and court
• Firm-specific learning - Your Bauweisen, Lieferanten, Subunternehmer and Nachtrags-Vorlagen become rules the agent applies automatically
• Voice-first for the Baustelle - Polier speaks into the phone; agent produces the Bautagebuch entry in the firm’s format
• Multi-Lieferanten-Logik - Sonepar, Würth, Frankenhof, regional Baustoffhändler reconciled in one comparison
• EU-hosted, DSGVO-ready - Frankfurt or Berlin hosting, AVV with no-training, full audit logs
• Human-in-the-loop by design - Confidence thresholds are configurable; nothing auto-submits to the Auftraggeber below the line you set
• 60-day pilot scope - One Baustelle, 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 as the firm evolves

Decision Framework

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

Start with a Bautagebuch app if

• Under 10 employees, 1-2 Baustellen - Throughput is not yet the constraint; a Capmo or PlanRadar subscription is enough
• No Bau-ERP yet - Get the ERP in first; the agent sits on top of it
• No growth or succession pressure - Status quo is acceptable for the next 12 months
• You want to test the digital baseline first - PlanRadar, Capmo or 123erfasst is the lowest-friction starting point

Move to a custom agent if

• 25+ employees, 4+ parallel Baustellen - Coordination overhead and project count justify the build
• EUR 10+ million annual revenue - Volume produces the ROI mathematics
• Hiring is blocked - You have tried to hire Bauleiter and Kalkulator and the candidates are not arriving
• Aufmaß is the bottleneck - Schlussrechnungen drag, Nachträge get rejected on documentation, BWA delivery is late
• Bauleiter are doing the Bautagebuch at 21:00 - The senior team is spending evenings on data entry that the agent should be doing
• Auftraggeber demand BIM-konforme Daten - You need the bridge while you build BIM competence in-house
• Succession or sale on the 3-year horizon - Modernised firms sell for higher multiples

Wait if

• Stammdaten are a mess - Clean them first; the agent will not save you
• No GAEB-XML discipline - Sort the data exchange before the project
• Resistance from the Bauleiter - The senior Bauleiter has to sponsor the project; without that, it stalls
• No budget for AVV review - The legal foundation is non-negotiable; budget the lawyer hours

Frequently Asked Questions

Related Articles

• AI in Manufacturing: How the Mittelstand Automates the Shop Floor With Custom AI Agents
• AI for Quote Calculation and Pricing in the Mittelstand
• AI in Procurement: How the Mittelstand Automates Supplier Selection and Contract Negotiation
• Predictive Maintenance for the Mittelstand: How AI Agents Predict Equipment Failure
• Intelligent Document Processing: How the Mittelstand Replaces OCR With AI Agents
• AI Agent ROI: The KPI Framework That Convinces CFOs in 90 Days
• The EU AI Act for the Mittelstand: A Practical Compliance Playbook

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.