AI for the Handwerk: How SHK, Elektro and Bau Trades Win Back the Office Day

It is 9pm. The van is unloaded, the crew has gone home, and the Chef of a six-person SHK Betrieb sits at the kitchen table with a laptop, writing the three Angebote that were promised this week. Each one means digging out the last similar job, retyping the positions, guessing the current material price, and hoping the margin is still right. Two of them will go out tomorrow. The third will not get written at all, and that customer will call a competitor.

This is the real bottleneck in the German Handwerk. It is not that the work is not there - it is that there is no capacity to price it, invoice it, and answer the phone about it. The Zentralverband des Deutschen Handwerks counts around 250,000 unfilled positions, and Betriebe are turning orders away not only for lack of a fitter, but for lack of anyone to run the office^{4, 12}. Meanwhile only 4 percent of Handwerksbetriebe use AI at all, even though 89 percent call digitalisation an opportunity^{1, 3}.

This guide is for the owner or Meister of an SHK, Elektro or Bau Betrieb who is tired of doing the Papierkram after a twelve-hour day. The honest frame first: AI will not lay your pipe or pull your cable. What a custom AI agent does is take over the back-office load - the Angebote, the Rechnungen, the Kundenkommunikation - so the Betrieb stops turning away work for lack of office time. Below is exactly how that works, what it costs, and how to run a 60-day pilot without betting the firm on it.

The Office Day Nobody Has

Every Handwerksbetrieb runs two businesses at once: the trade on site, and the office that turns that trade into money. The trade is where the pride is. The office is where the orders quietly die - the quote that never got written, the invoice that went out three weeks late, the customer who never got a reply. In a small Betrieb there is no Büro to absorb this; it lands on the Chef, after hours.

• Angebote are the first casualty - Writing a proper Angebot from scratch means retrieving a past job, retyping positions and re-pricing material; when there is no time, the quote does not go out and the order is lost⁷.
• Invoices slip - Rechnungen written late mean cash comes in late, which for a small Betrieb is the difference between paying suppliers on time and not.
• Customers wait in silence - Enquiries and "when are you coming" messages sit unanswered for days because the person who would reply is up a ladder, and silence sends customers to a competitor.
• The Chef becomes the bottleneck - Every quote, every awkward invoice, every price question funnels to one person who is already working the longest day in the firm.
• Orders get declined for office reasons - Betriebe turn away work not because they lack a fitter for the job, but because they lack the office capacity to price and manage it^{4, 12}.
• The fax still runs - In the Bitkom 2026 study, 26 percent of Handwerksbetriebe still use a fax machine regularly, against 4 percent using AI - a snapshot of where the office backlog comes from³.

The pattern is the same across SHK, Elektro and Bau: the constraint is not the skill on site, it is the hours in the office. That is precisely the shape of work an AI agent removes.

What an AI Agent Actually Does for a Handwerksbetrieb

There is a lot of noise labelled "KI fürs Handwerk", and most of it is a chat window that answers general questions. A real trades agent is different: it does the specific back-office jobs of your Betrieb, grounded in your own prices and your own customers, and it leaves the judgement to the Meister.

The agent loop for a Betrieb

1. Capture - Takes the input in the form you already produce it: a voice note from the van, a photo of the site, a rough email from a customer, a scribbled Aufmaß.
2. Draft the Angebot - Turns that input into a structured quote, priced from your own Kalkulation and material list, in your template.
3. Approve - The Meister checks the quote, adjusts if needed, and sends - the agent never fires a price into the world unseen.
4. Invoice - On job completion, generates the Rechnung as a compliant E-Rechnung (XRechnung or ZUGFeRD), ready to approve⁹.
5. Communicate - Answers routine customer messages, confirms Termine, and acknowledges enquiries within minutes instead of days.
6. Chase - Watches open invoices and sends polite, correctly-timed Zahlungserinnerungen, escalating only the ones that need the Chef.
7. Keep the trail - Files every document GoBD-conform, so the Steuerberater and any Prüfung find a clean record.

The difference from what you have today

The category matters. A generic chatbot writes fluent text with no idea what your material costs, so it produces a quote no Meister would ever send. A custom agent does the opposite: it grounds every quote and message in your real numbers and your real customers, and it never sends a price without you.

The Numbers Behind the Gap

The Handwerk knows digitalisation matters. It just has not started. The gap between recognition and action is the whole opportunity - and the data on it is unusually clear.

What the studies show

• Opportunity recognised - 89 percent of Handwerksbetriebe see digitalisation as a chance for their business^{1, 3}.
• Action not taken - Only 4 percent actually use AI, and 84 percent say AI plays no role for them at all right now³.
• Intent is building - Around 10 percent plan to introduce AI, so the 4 percent is a floor, not a ceiling³.
• The workforce gap is structural - Roughly 250,000 positions are unfilled, and the demographic curve makes it worse as the Babyboomer generation retires^{4, 12}.
• Digital laggards lose the hiring race - 54 percent of Handwerksbetriebe agree that without digital tools they lose the competition for skilled workers¹³.
• Skepticism is honest, not misplaced - Firms doubt AI can conjure a missing fitter, and they are right; the win is relieving office work, not replacing the trade³.

The takeaway is not "adopt AI because everyone else is". Almost nobody else is. It is "the office backlog is measurable, the tools to clear it exist, and the first mover in a local market gets the orders the slower Betrieb has to decline".

5 Use Cases That Work Today

Not every office task is a good first AI candidate. The five below are - they are repetitive, high-volume, and have a clear correct output. Start with Angebote, the biggest time sink and the clearest win. Add the next once the first runs cleanly.

1. Angebote from a Voice Note or Photo

The single biggest lever. The Chef records a 30-second voice note in the van or snaps a photo of the site, and the agent turns it into a structured Angebot priced from the Betrieb’s own Kalkulation, ready to check and send.

• Voice and photo capture - Input in the form you already produce; no sitting down to type at 9pm
• Priced from your Kalkulation - Positions and material priced from your real numbers, not a generic average
• Your template - The quote comes out in your layout with your terms, ready to send
• Documented time saving - A Stuttgart Elektro firm reports around 90 minutes a day saved; two-hour quotes drop to 15 to 20 minutes^{6, 7}
• Fast first value - The first automated Angebot lands within the first two weeks of a pilot

2. E-Rechnungen In and Out

No longer optional. Since 1 January 2025 every B2B Betrieb must be able to receive structured E-Rechnungen, and the duty to send them phases in from 2027. The agent handles both directions and keeps it GoBD-conform.

• Structured output - Outgoing invoices as XRechnung or ZUGFeRD, not a PDF that no longer counts^{9, 10}
• Incoming processing - Reads inbound E-Rechnungen so they do not sit unread in a mailbox¹¹
• Invoice on completion - The Rechnung is drafted the moment the job is marked done, not weeks later
• GoBD-conform filing - Every document archived in the form the Finanzamt expects
• Faster cash - Invoices that go out on time get paid sooner, which matters most to a small Betrieb

3. Kundenkommunikation

The quietest lost-order machine. Enquiries that sit unanswered for days go cold. The agent handles the routine layer so nobody waits, and flags anything that needs a person.

• Minutes, not days - Enquiries acknowledged fast, so the customer does not call the next Betrieb
• Routine answers - "When are you coming", "what does it cost roughly", appointment confirmations handled automatically
• Escalation to a person - Anything needing judgement, a site visit or a real decision goes to the Chef
• Consistent tone - Every reply sounds like your Betrieb, not a random late-night message
• Fewer angry calls - Silence is what makes customers angry; the agent removes the silence

4. Invoice Chasing (Zahlungserinnerungen)

Pure repetitive work nobody enjoys, and for a cash-tight Betrieb often the fastest financial win. The agent watches open Rechnungen and nudges at the right time.

• Open-invoice watch - The agent tracks which Rechnungen are unpaid and how overdue they are
• Timed reminders - Polite Zahlungserinnerungen sent on the right schedule, automatically
• Escalation only when needed - The awkward phone call stays human; the routine nudges do not
• Cash-flow visibility - The Chef sees what is outstanding without opening five folders
• Money in faster - Getting paid sooner beats almost any efficiency gain for a small firm

5. Terminplanung and Dispatch

The coordination layer that eats the Chef’s phone. The agent keeps the calendar straight and reduces the double-bookings and no-shows that waste a crew’s day.

• Appointment confirmation - Termine confirmed and reminded automatically to cut no-shows
• Rescheduling - Routine moves handled without a phone call, escalated only when tight
• Crew visibility - Who is where, when, kept current so nobody drives to the wrong site
• Buffer for travel - Realistic gaps between jobs so the day does not collapse by noon
• Fewer wasted trips - A confirmed appointment beats a locked door and a lost hour

How the Agent Fits Your Betrieb

A trades agent only works if it sits on top of what you already run and never becomes a second system you have to feed by hand. Here is what the working setup looks like for a small Betrieb.

The layers

1. Your prices and templates - The agent is grounded in your Kalkulation, material prices, Stundensätze and quote template, captured once at setup.
2. Capture - It takes input as you produce it: voice note, photo, email, Aufmaß - no new app the crew has to learn.
3. Drafting and approval - It drafts the Angebot, Rechnung or reply, and the Meister approves before anything leaves the Betrieb.
4. Your existing tools - It reads and writes through your Handwerkersoftware, email and accounting or DATEV export, not a parallel database.
5. Compliant trail - Every document filed GoBD-conform and every action logged, so the Steuerberater and any Prüfung find a clean record.

What sits where

The principle is simple: your systems stay the source of truth, and the agent is the office worker that turns your inputs into finished, compliant documents. Any setup that copies your customer and price data into a parallel store breaks both data governance and your peace of mind.

Where the data lives

• Customer and price data - In your systems, untouched; the agent works through approved connections
• Agent state - On EU-resident infrastructure under a DSGVO-compliant Auftragsverarbeitungsvertrag
• Model inference - Via EU-resident endpoints, with a no-training clause so your quotes and customers never train a vendor model
• Documents - Archived GoBD-conform for the statutory retention period, ready for any Prüfung
• Backups - In the EU, encrypted at rest

The 60-Day Pilot Playbook

The biggest mistake a Betrieb makes with AI is waiting for a perfect plan. The right path is the opposite: one workflow, sixty days, and a clear yes-or-no at the end.

The phases

1. Days 1-10: Capture your prices - Get your Kalkulation, material prices, Stundensätze and quote template into the agent. This one step decides whether every future quote is right.
2. Days 11-20: First Angebote - The agent drafts quotes from real voice notes and photos. The Meister checks each one against what he would have priced, and corrects the agent where it is off.
3. Days 21-35: Add E-Rechnung - Switch outgoing invoices to structured E-Rechnung and turn on inbound processing, checked against your accounting or DATEV export.
4. Days 36-45: Add Kundenkommunikation - Let the agent acknowledge enquiries and confirm Termine, with a person approving anything non-routine.
5. Days 46-55: Add invoice chasing - Turn on timed Zahlungserinnerungen for open invoices, escalating the hard cases to the Chef.
6. Days 56-60: Go or no-go - Compare office hours before and after, count the quotes that got written that would have been skipped, and decide whether to expand.

Checklist before you start

What to measure

• Quote turnaround - Time from enquiry to Angebot sent, before and after (often two hours down to under 20 minutes)⁷
• Quotes not skipped - How many Angebote got written that would previously have been dropped
• Invoice timing - Days from job completion to Rechnung sent
• Days sales outstanding - How much faster open invoices get paid with automated chasing
• Office hours recovered - Evenings and weekends handed back to the Chef

E-Rechnung, DSGVO and the EU AI Act

Compliance is where a careless setup gets a Betrieb into trouble. The good news: for the back office of a Handwerksbetrieb, the three rule sets that matter each have a concrete answer, and none is a blocker.

E-Rechnung and GoBD

The E-Rechnung is now the baseline, not a project. Since 1 January 2025 every domestic B2B Betrieb must be able to receive and process structured E-Rechnungen, and the duty to send them arrives from 2027 for larger firms and 2028 for all^{9, 10}.

• Structured, not PDF - Only XRechnung or ZUGFeRD count; a simple PDF is no longer a valid E-Rechnung⁹
• Receive from day one - The obligation to receive already applies to every B2B Betrieb, with no customer consent required¹⁰
• Send in stages - From 2027 for Betriebe above 800,000 EUR turnover, from 2028 for everyone⁹
• GoBD retention - Invoices archived in the correct form for the statutory period, unaltered
• The agent handles both directions - Generating valid outgoing E-Rechnungen and reading incoming ones so nothing is missed

DSGVO - the data side

A Betrieb’s customer list, addresses and prices are commercially sensitive and often personal data. Data protection and IT security are the barriers Handwerksbetriebe cite most, so the setup answers them first¹⁴.

• EU-resident processing - Customer and price data stays in EU data centres, with no transfer to non-EU models without a valid legal basis
• Auftragsverarbeitungsvertrag - An AVV governs the agent provider and the model endpoint, covering purpose and deletion
• No training on your data - The vendor agreement includes an explicit no-training clause for your quotes, invoices and customer messages
• Data minimisation - The agent handles only the data a given task needs, not the whole customer database at once
• Access logging - Who did what and when is logged, useful for both DSGVO and your own records

EU AI Act - lighter than you think

For a Handwerksbetrieb using an agent to write quotes and answer emails, the EU AI Act is a light-touch, transparency-level use - not the high-risk category that makes headlines¹⁷.

• Low-risk use - Drafting Angebote, Rechnungen and replies is a transparency-level use as long as a human stays responsible
• Human in the loop - The Meister approves; the agent does not make binding decisions on its own
• Article 4 (AI literacy) - Staff using the agent get basic AI literacy, which for a small Betrieb is a short briefing, documented¹⁸
• Provider carries model duties - The general-purpose model obligations sit with the vendor; your Betrieb is the deployer, not the provider
• Transparency to customers - Where a customer interacts with an automated reply, it is made clear they can reach a person

Build vs Buy vs Bürokraft

There are three ways to fix the office bottleneck: hire a Bürokraft, buy a generic Angebotssoftware, or build a custom agent. Most Betriebe end up with a mix, and the right one depends on volume and how specific your pricing is.

The three options compared

The honest combination

• Agent for volume and routine - Angebote, E-Rechnungen, chasing and routine replies where a person would be expensive or impossible to hire
• A person for judgement - The tricky customer, the disputed invoice, the negotiation stay human
• Agent makes a hire go further - If you do have a Bürokraft, the agent removes the drudgery so that role does the work that needs a person
• Never the generic tool alone for pricing - A tool with no access to your real Kalkulation produces quotes you cannot send
• The maths flips with volume - A handful of quotes a month: a person or simple software is fine. Several a week and orders being declined: build the agent

How Superkind Fits

Superkind builds custom AI agents that sit on top of the systems a Betrieb already uses - Handwerkersoftware, email, accounting - without forcing a switch. The deployment model is process-first: we map how you currently quote, invoice and communicate before we touch a line of code. The agent is built around how your Betrieb actually works, not a generic trades template.

What sits in the Superkind trades agent

What Superkind brings to a Betrieb

• Process-first deployment - We map how you quote, invoice and communicate before we build, so the agent fits your reality
• Grounded in your prices - The agent quotes from your Kalkulation and material list, so a quote matches your margins, not a generic average
• Works from your inputs - Voice note, photo, email or Aufmaß, in the form your crew already produces
• End-to-end back office - Angebote, E-Rechnungen, Kundenkommunikation, chasing and Terminplanung in one flow
• E-Rechnung and GoBD ready - Compliant structured invoices in and out, archived correctly
• DSGVO-ready compliance - EU hosting, AVV, no-training guarantee, and access logging
• Human-in-the-loop by design - Nothing priced or sent without the Meister approving
• 60-day pilot scope - One workflow, written success criteria, go or no-go after 60 days

Decision Framework

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

Stay manual or hire if

• You quote rarely - A handful of Angebote a month does not justify a build
• Your office is already covered - A well-staffed Büro that is not the bottleneck needs no agent
• Every job is fully bespoke - If no two quotes share a structure, grounding is harder
• You have no digital prices - Build a price list first; the agent needs numbers to quote from

Build a custom agent if

• You write several Angebote a week - The volume is where the recovered hours come from
• You turn away work for office reasons - The lost orders are the clearest ROI
• You cannot hire the office help - The 250,000-person shortage is not going to solve itself⁴
• Your evenings are gone to Papierkram - The office day is landing on the Chef, after hours
• You want compliance handled - E-Rechnung and GoBD are now table stakes, not optional

Wait if

• Your price list is not written down - Capture your Kalkulation first, or every quote is a guess
• No one will approve drafts - Human sign-off is the safeguard; without an owner it stalls
• No budget for the AVV and setup - The compliance foundation is non-negotiable
• The office is genuinely not your constraint - Fix the actual bottleneck first

Frequently Asked Questions

Related Articles

• AI in Quoting and Pricing: How the Mittelstand Turns Enquiries Into Fast, Accurate Offers
• E-Invoicing in Germany: What the E-Rechnung Obligation Means for the Mittelstand
• AI in Field Service: How the Mittelstand Keeps Technicians Productive on Site
• AI Agents and the Labour Shortage: How the Mittelstand Does More With Fewer People
• AI in Customer Service: How the Mittelstand Answers Faster Without Losing the Human Touch
• AI in Construction: How Bau Firms Cut Admin and Keep Projects on Track

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.