AI in the HVAC & Plumbing Trade

Overview

What this whitepaper covers

The whitepaper is designed as a working document. The first two parts frame the topic, the third part describes eleven use cases along the order chain, and the fourth part provides tools for making and implementing a decision. Anyone who already knows where time is lost in their firm can jump straight to Part III.

• Part I · Framing
• 01Management summary3
• 02The five key messages4
• 03Market situation and figures5
• 04How widespread AI actually is6
• Part II · Foundations
• 05What AI means in an HVAC and plumbing firm7
• 06Six levels of automation8
• 07Where time is actually lost9
• 08Screening a use case for suitability10
• Part III · Use cases
• 09Call handling and fault reports11
• 10Limits and metrics for the phone assistant12
• 11Enquiry qualification and appointment prep13
• 12Quote preparation14
• 13Worked example: heat-pump quote15
• 14Scheduling and dispatch16
• 15Voice documentation in the field17
• 16Company knowledge, made findable18
• 17Fault diagnosis and maintenance support19
• 18Materials, purchasing and van stock20
• 19Site documentation and variations21
• 20Customer communication22
• 21Marketing, recruitment and visibility23
• Part IV · Decision and implementation
• 22What not to introduce prematurely24
• 23Calculating cost-effectiveness25
• 24Scoring matrix and firm size26
• 25Technical prerequisites27
• 26Legal framework: GDPR and the AI Act28
• 27AI governance and co-determination29
• 28The 90-day implementation plan30
• 29Practical examples A and B31
• 30Practical examples C and D32
• 31Final check: are you ready?33
• 32Closing remarks and initial review34
• 33Sources and studies35

Management summary

In an HVAC and plumbing firm, AI is first and foremost an organisational tool

Public debate about artificial intelligence often revolves around autonomous systems, robots or fully automated companies. For a plumbing, heating and air-conditioning firm, the short-term benefit lies somewhere else entirely.

AI is particularly good today at taking in information, structuring it, summarising it and preparing it for the next step in the process. That is precisely where many firms suffer avoidable friction. It is rarely spectacular. But it costs time every day, and it hits exactly the people who are hardest to replace.

Seven situations every firm knows

• A caller describes a boiler breakdown, but the crucial details are missing.
• The technician records a voice note that someone later has to transcribe by hand.
• To prepare a quote, information is gathered from emails, photos and measurement records.
• An employee searches for an installation manual or a comparable past job.
• A maintenance report is technically correct but hard for the customer to understand.
• After the job is finished, photos, readings or signatures are missing.
• Management cannot see where orders are stalling or why follow-up questions arise.

In all of these situations, AI can prepare work, organise information and flag missing details. It should not, however, issue technical approvals on its own, nor make binding designs, prices or safety-relevant decisions without oversight. That dividing line runs through the entire whitepaper.

The right question to start with

An HVAC and plumbing firm should not begin with a broad question such as "Where can we use AI?". That question almost always leads to a discussion about tools and rarely to a result. A far more productive starting point is:

"Where do our staff regularly lose time because information is missing, has to be entered more than once, or has to be searched for again and again?"

Anyone who can answer that question for their own firm has already done the hardest part of an AI project. Everything that follows – selection, testing, integration, oversight – is craft in the best sense: it can be planned, checked and, if necessary, corrected.

Market situation

A strong sector under economic and organisational pressure

Germany's plumbing, heating and air-conditioning trade closed 2025 at a high level, but with slowing momentum. Orders and utilisation declined, and the business is increasingly carried by repair and maintenance rather than genuine modernisation work. The service and maintenance business, by contrast, remained a reliable revenue pillar. The industry association points to political uncertainty around the building energy act and the subsidy landscape as the central brake; it expects no fundamental turnaround for 2026.1

At the same time, the technology in the installed base is shifting. With around 299,000 units sold, the heat pump was the best-selling heating technology of 2025 for the first time.2 For firms this means not only new products but, above all: more advisory effort, more coordination with electrical and building trades, more documentation and more follow-up questions.

Several developments coincide

• New-build is losing importance as an order driver.
• Modernisation decisions are being delayed by political and economic uncertainty.
• Repair, maintenance and servicing are gaining weight – and with them small-scale, documentation-heavy tasks.
• Customers expect fast responses and quotes they can follow.
• Heat generators, controls and building services are becoming technically more complex.
• Experienced staff are needed for diagnosis and planning, yet still spend time on administration.
• Knowledge is spread across people, folders, emails, messengers, service vans and individual software tools.

A firm cannot replace missing service technicians with a language model. But it can try to relieve the skilled staff it has of tasks that do not require a master craftsman, a technician or an experienced service engineer. It is precisely in that gap – between what genuinely requires expertise and what is merely legwork – that the realistic contribution of AI lies.

Why the topic matters now

An early market stage – and a window of opportunity

In the German trades, only 4 per cent of the firms surveyed currently use AI, with a further 9 per cent planning to. At the same time, 35 per cent believe that early adoption can create a competitive advantage, and 33 per cent expect AI to fundamentally change business models in the trades. Only 29 per cent of trade firms say they have staff who can work with AI.3

The view beyond the trades

In 2025, 26 per cent of German companies with at least ten employees used AI. Among companies with 10 to 49 employees the figure was 23 per cent, among those with 50 to 249 employees 36 per cent, and among large companies with 250 or more employees 57 per cent.4 The most common reasons cited against adoption were a lack of knowledge (72 per cent), uncertainty about the legal consequences (62 per cent) and data-protection concerns (60 per cent).4

Across the wider German SME sector (the "Mittelstand"), adoption stands at 20 per cent according to a special analysis of the KfW SME Panel – equivalent to just under 780,000 companies, a fivefold increase within six years. What matters is less the size than the environment: companies with research and development activity and pronounced digitalisation activity use AI significantly above average. Companies with neither graduates on staff nor any innovation activity use AI with a probability of just 8 per cent.6

The figures are not directly comparable: population, survey period and the definition of AI differ considerably between studies. The gap between 4 per cent in the trades and 26 per cent in the economy as a whole is nonetheless clear enough to serve as a directional signal.

What this means for HVAC and plumbing firms

Success is decided not by firm size but by orderly processes, digitally available information, clear responsibilities, a concrete use case, sufficient data quality and acceptance in the office and on the job site alike. The advantage comes not from buying a tool but from better processes, well-maintained data and trained staff.

A practical classification

Six levels – and the question of where your firm stands today

The following classification orders AI applications not by technology but by depth of intervention. The further down the table, the greater the potential benefit – and the greater the effort required for integration, permissions and oversight.

The typical mistake at the start

Many firms try to enter directly at level 4 or 5 because a vendor's demonstration was impressive. The outcome is regularly the same: the project gets stuck on interfaces, permissions and data quality, ties up management time and ends without a result. Afterwards "AI doesn't work for us" becomes the verdict in the firm – and the simple, effective levels 1 to 3 are buried along with it.

Two questions before every level decision

1. What happens in the worst case? A poorly worded text draft costs two minutes of correction. A wrongly triggered order costs money. A wrong technical assurance costs a customer.
2. Who notices – and when? An error caught before it is sent is one work step. An error caught by the customer is a complaint.

Analysis

Where time is actually lost in the firm

A typical HVAC and plumbing job passes through ten stations. At each one, information is handed over – and it is precisely at these handovers that the errors arise which cost time later on.

The typical break points

Suitability screening

How to recognise a good use case

A process is especially well suited to AI support when several of the points below apply. Rule of thumb: the more ticks, the more a pilot is worthwhile. Fewer than five ticks is a sign that the process, not the tool, is the first order of business.

• Much of the information exists as free text, email, voice note, photo or PDF.
• Staff enter the same details more than once into different systems.
• The process broadly follows a recurring pattern.
• Missing information regularly leads to follow-up questions.
• Experienced staff carry out simple preparatory work their qualification is not needed for.
• The quality of the result depends above all on complete information.
• A final human check is organisationally possible.
• The process occurs often enough to generate an economic benefit.
• An error is noticed before it reaches the customer.
• There is a named person responsible for the professional quality of the result.

Four contraindications

Use case 1

Call handling and fault reports

From a missed call to a structured job.

In an HVAC and plumbing firm the telephone is the central customer interface – and at the same time the biggest source of interruption. Calls come in while staff are working on quotes, ordering materials, coordinating technicians, checking invoices, serving customers at the counter or out on a job themselves. On a Monday morning and after the first frost, this multiplies.

An AI-supported phone assistant can take calls outside business hours or when the line is busy. What matters is not that a synthetic voice sounds as human as possible. What matters is whether usable, complete information reaches the firm in the end – and whether the customer felt taken seriously.

A possible call flow

1. What is your name, and which property is this about?
2. Is this a fault, a service, a modernisation or a general enquiry?
3. Does the problem concern heating, hot water, plumbing or ventilation?
4. How long has the fault existed?
5. Is a fault code displayed? If so, which one?
6. Is there any smell of water, gas or burning?
7. Are other flats or parts of the building affected?
8. On which number can we reach you?
9. When is access to the property possible?

What comes out of it

Use case 1 · continued

Limits, metrics and checklist

An AI phone assistant must not issue a remote diagnosis as a firm statement, must not quote prices and must not promise appointments that dispatch has not confirmed. In the event of a gas smell, water leak, fire risk or comparable hazard, a fixed safety message must apply – as a rigid rule, not as a language model's free interpretation. This flow must be tested before the system goes live.

Suitable metrics

Checklist before going live

• There are defined categories for call reasons.
• Emergencies and hazards are stored as a fixed rule, not as a model decision.
• The assistant promises no appointments and no prices without approval.
• Customers recognise at the start of the call that they are speaking with an automated system.
• Call data is stored only as long as necessary; deletion periods are defined.
• Transfer to an employee is possible at any time – without a magic word.
• Results are spot-checked during the first few weeks.
• Dialects, background noise, site noise and unclear statements have been tested.
• A failure of the assistant does not mean a failure of reachability.

Use case 2

Enquiry qualification and appointment preparation

Not every enquiry needs an on-site appointment straight away.

Many firms lose time because enquiries are treated as orders too early. Only later does it turn out that important prerequisites are missing – and by then with a site visit, a survey and an afternoon of costing already behind them.

Six typical cases

• For a bathroom renovation, the floor plan, budget range and desired timeframe are missing.
• For a heat-pump enquiry, it is unclear what heating surfaces are present.
• A prospect expects a non-binding full plan without placing an order.
• A repair enquiry concerns a make the firm does not service.
• The property lies outside the usual service area.
• A customer demands an appointment even though access to the plant room is not secured.

AI-supported enquiry qualification can gather information via a web form, phone or email, check it for completeness and prepare it for the office. It does not replace advice – it ensures that the advice starts with complete information.

Use case 3

Quote preparation

From scattered information to a checkable basis for a quote.

Preparing quotes is one of the most time-consuming commercial-technical tasks in an HVAC and plumbing firm. It becomes especially demanding when information has to be brought together from a survey, site photos, customer emails, plans, a schedule of works, a manufacturer quote, wholesale terms, technician notes, earlier comparable quotes and subsidy requirements. That this work happens in the evening in many firms is no accident: it requires uninterrupted concentration.

A possible flow

1. Documents and images are assigned to a job.
2. The AI extracts the relevant details.
3. Missing information is explicitly flagged as open.
4. Services are grouped by trade or construction phase.
5. A draft of the specification and cover letter is produced.
6. The estimator adds quantities, items, mark-ups and prices.
7. A responsible member of staff checks and approves the overall quote.

This rule is not distrust of the technology. It follows from the fact that a quote is a legally binding document. A language model does not bear liability. The firm does.

Worked example

Heat-pump quote: what the AI prepares – and what it does not

For a heat-pump enquiry, a firm receives the following documents: a customer email, the consumption bills from the last three years, photos of the existing system, a heat-load calculation from a planner, a floor plan and the manufacturer's product documents. Six sources, three formats, one job.

What happens next

The master craftsman checks the technical plausibility: does the proposed heat pump match the calculated heat load? Are the heating surfaces suitable for the required flow temperature? What about the electrical panel? The estimator produces the commercial assessment. The AI replaces none of these steps. It reduces the effort of reviewing, sorting and structuring – that is, the part that previously took almost an hour without being professionally demanding.

Where the research stands

That capturing the installed base is a worthwhile area for digitalisation in the HVAC and plumbing trade is also borne out by research. In the BMBF-funded DiBesAnSHK project (running from 09/2021 to 08/2024), the Fraunhofer Institute for Solar Energy Systems ISE, together with industry partners, developed a method that captures heating components using off-the-shelf smartphones and tablets, classifies them via image recognition and feeds them into a system configurator – with the aim of enabling on-site survey and rapid quotation.7 In the ongoing WESPE project, the approach is being transferred to heat-pump conversions.8

The potential is considerable. In practice, however, quality depends on the image, the installation situation, the equipment generation and the available data. A cramped rating plate in a dark plant room remains a cramped rating plate in a dark plant room.

Use case 4

Scheduling and dispatch

Better preparation rather than fully automated route planning.

AI can support dispatch when sufficient data is available: staff qualifications, job location, expected duration, required spare parts, van stock, urgency, promised time windows, dependencies on other trades, access conditions, maintenance contracts and emergency-service rules.

Dispatch AI should therefore first produce suggestions and flag anomalies. Binding planning stays with the dispatcher. Their knowledge of customers, buildings and staff is the part that appears nowhere in the system – and it decides the quality of the day.

Use case 5

Voice documentation on site and in the field

Documenting without writing up reports after hours.

Voice-based documentation is one of the most obvious AI applications in the HVAC and plumbing trade – because it starts exactly where the work already happens and creates no additional screen time.

"At the Müller property, the gas condensing boiler was checked. Fault code 227. Ionisation electrode dirty, cleaned and burner inspected. System back in operation. Flue-gas readings within range. Condensate drain was partly blocked and has been cleared. Customer advised that the service was overdue. Recommendation: offer an annual maintenance contract."

From this note the system produces work carried out, cause identified, materials used, readings, customer advice, follow-up recommendation, billable items and an internal note – each in separate fields rather than in a flowing text that someone later has to take apart.

Use case 6

Company knowledge, made findable

When the right information exists but cannot be found.

Most HVAC and plumbing firms hold extensive knowledge: installation manuals, manufacturer information, maintenance checklists, price lists, work instructions, references to standards, safety briefings, templates, technical data sheets, earlier quotes, system histories and experience. The problem is rarely a shortage of documents. The problem is their distribution – across drives, mailboxes, folders, messenger groups and people's heads.

A knowledge-based assistant searches only approved documents and answers questions with a source reference. That source reference is precisely the difference between a useful tool and a well-worded risk.

Three answer levels a good assistant distinguishes

Use case 7

Fault diagnosis and maintenance support

Assistance for the expert, not a substitute for expertise.

AI can support fault-finding when it can access suitable information: fault code, equipment type, system history, readings, manufacturer documents, checks already carried out, known fault patterns and maintenance status. From this it can suggest possible causes, a sensible testing sequence or the right document.

What the AI cannot know in this case

It does not know the actual state of the system. It does not know installation errors, the local hydraulics, the electrical readings, changes made by other trades, and certainly not undocumented modifications. It can therefore supply a hypothesis but never a finding. In the field, the difference is not academic: turning up with the wrong spare part means a day lost.

Use case 8

Materials, purchasing and van stock

Less searching, fewer unnecessary trips.

Material problems cause considerable knock-on costs in the field: missing spare parts, wrong items, undocumented withdrawals, overloaded vans, obsolete stock, unresolved returns and spontaneous trips to the wholesaler. Every one of those trips costs not only diesel but an appointment.

Use case 9

Site documentation, handover and variations

Complete records are part of the value created – not its afterthought.

A technically flawless job can become a commercial problem if the documentation is incomplete. Extra work goes unbilled, the client disputes the scope, photos are missing, deviations from the order are undocumented, concerns were raised only verbally, commissioning documents are incomplete, the invoice cannot be issued – and in the event of a complaint the job history is missing.

AI-supported completeness check before closing a job

• Are before-and-after photos available?
• Have readings been recorded?
• Is the serial number documented?
• Was the customer instructed?
• Are there any open remaining works?
• Have extra services been described?
• Is there a signature?
• Are the materials used entered?
• Does a follow-up quote need to be created?
• Was a maintenance offer discussed?

Of all eleven use cases, this is the one with the most direct link to the result: a completeness check that finds one forgotten extra service per week has, in many firms, already paid for itself on paper before anyone has even mentioned time savings.

Use case 10

Customer communication

Explaining technical matters clearly.

HVAC and plumbing firms communicate daily with very different counterparts: private owners, tenants, property managers, architects, energy consultants, developers, commercial clients, municipalities and care facilities. Not every recipient understands terms such as spread, heating curve, hydraulic balancing, return temperature or flow rate. And not every technician has the patience to translate them at 6 p.m.

Three typical situations

Further functions

Use case 11

Marketing, recruitment and visibility

Good work is not enough if it cannot be found online.

Customers increasingly inform themselves online before contacting an HVAC and plumbing firm. Personal recommendations remain important, but they are supplemented by search engines, map services, reviews and, increasingly, AI-supported answer systems. Anyone who does not appear in these systems, or appears with contradictory details, is simply not considered.

In the Bitkom survey of the trades, 85 per cent of firms stated that they offer at least one digital service – such as online appointment booking, online advice or electronic invoicing. Messenger services are used by 62 per cent and online meetings by 36 per cent.3 The basic equipment is therefore in place. The difference arises in consistency.

Attracting the next generation

AI can prepare job adverts and content. But the real advantage only emerges when the firm credibly shows what technology it uses, how training is organised, what development opportunities exist, how site and office work together, and what values apply. According to the Bitkom study, more than half of the training firms in the trades let their own apprentices help them with digitalisation.3 That is a remarkable finding: the next generation often brings the competence with them – if the firm lets them.

Cost-effectiveness

Not "How much can AI do?" but "What effort is removed?"

The three examples below are model calculations based on freely chosen assumptions. They show the calculation method, not a guaranteed result. Use your own case numbers and cost rates – a firm with 200 calls a month reaches a different decision from one with 600.

The underestimated effects

In practice, the greater lever often lies not in time but in what is currently simply lost: the extra service that is not billed, the missed call on a Monday morning, the follow-up quote that was never written, the invoice that goes out three weeks late. These effects are harder to quantify – but they are the ones that make the difference.

Technical prerequisites

AI is no substitute for a solid digital foundation

An HVAC and plumbing firm needs no IT department of its own. But it does need a traceable architecture – and the knowledge of where each piece of information arises. The typical system landscape comprises trade software or ERP, order and customer data, document storage, email, calendar, telephone system, a mobile field solution, accounting, time recording and, where applicable, stores and purchasing.

Three implementation stages

Data classes a solution must be able to distinguish

Legal framework

GDPR and the EU AI Act: what matters for HVAC and plumbing firms

The EU AI Act entered into force on 1 August 2024 and takes effect in stages. With the so-called Digital Omnibus, the timetable was adjusted in 2026: the European Parliament approved the text on 16 June 2026, and the Council gave its final approval on 29 June 2026.510

The legal situation continues to evolve. The information reflects the position as at July 2026 and does not replace legal advice. Check the current position before an implementation, and if in doubt with legal support.

AI governance

A small firm needs no rulebook. It needs minimum rules.

Checklist before production use

• Data-processing agreements have been reviewed; storage location and sub-processors are known.
• The use of company data for training is contractually governed – excluded if in doubt.
• Roles and access rights are documented; staff have been trained.
• Critical actions require approval; logs and deletion periods are defined.
• A procedure for errors and security incidents exists; the application can be switched off.
• A manual fallback process is described and operable in an emergency.

The German Federal Office for Information Security (BSI) notes that generative AI brings new IT-security risks and can amplify known threats – among them manipulation of inputs, leakage of confidential information and faulty outputs.11

Practical examples

Four typical starting situations

The following examples are generalised constellations of the kind that regularly arise in HVAC and plumbing firms. They describe possible approaches and expected directions of effect – not measured results from individual reference customers.

Practical examples

Documentation and onboarding

Self-assessment

Is your HVAC and plumbing firm ready for an AI pilot?

Rate each statement with 0 points (not present), 1 point (partly present) or 2 points (sufficiently present). Do not answer the questions alone – the assessment by management and by the office typically differ.

Interpretation

Closing remarks

AI will not replace good HVAC and plumbing work. It can organise it better.

The economic value of an HVAC and plumbing firm continues to arise from proper planning, skilled workmanship, reliable service, responsible decisions and the trust of the customer.

Artificial intelligence is worthwhile when it supports these tasks. It can take in information, prepare documents, make knowledge findable and speed up documentation. But it cannot assume responsibility for safety, technology, price or workmanship – and it should not try to.

The best entry point is therefore not a large transformation programme. It is a concrete, frequently occurring task whose improvement becomes measurable in day-to-day work. Anyone who starts there gains two things: a result and experience. Anyone who starts with the platform question gains a presentation.

Which AI use case is worthwhile in your HVAC and plumbing firm?

A structured initial review examines, among other things:

KrambergAI GmbH

AI consultancy and digital products for small and medium-sized enterprises (SMEs) in the German-speaking region. Data protection in accordance with the EU GDPR.
krambergai.com

A possible next step: assess the use of AI for your HVAC and plumbing firm · identify a suitable AI use case · request an AI potential report · arrange an initial conversation about introducing AI.