MEP & BUILDING PLANNING · INDUSTRY HUB

AI for building planning and MEP/HVAC: SIA standards search, schema generation, energy modelling and BIM

How Swiss MEP/HVAC and building planners use AI for SIA standard search, HVAC schema generation, energy modelling and BIM enrichment – within SIA standards, MuKEn 2014 and Minergie.

Researched & fact-checked by: DuneDive LLC · As of: 2026-05

MEP/HVAC and building planning in Switzerland: overview

Building services engineering (MEP/HVAC) in Switzerland in 2026 is a market with around 4,500 specialist planning offices for heating, ventilation, air-conditioning, sanitary (HVACS) plus electrical and building automation. Market leaders are large multi-disciplinary engineering offices with MEP departments (Amstein + Walthert, Gruner, BG Ingenieurs Conseils, IttenBrechbühl) plus mid-size boutique offices with 5-50 staff. Industry bodies are suissetec (Swiss-Liechtenstein Building Services Association), USIC (Swiss Association of Consulting Engineers) and SIA (Swiss Society of Engineers and Architects).

AI in MEP/HVAC in 2026 has four application areas. First as a research tool over the extensive SIA standard body (around 600 SIA standards relevant to MEP and building planning), the cantonal MuKEn requirements (cantonal model rules in the energy sector) and the Minergie standards. Second as a generator tool for schemes, principle circuit diagrams and calculation templates. Third as a modelling tool for energy balances and load profiles. Fourth as a BIM enricher for models from Revit, ArchiCAD or Allplan.

The sector is in productive motion in 2026. suissetec and USIC offer continuing education on AI use in MEP; some larger offices have built dedicated AI teams. The associations point to SIA 2024 (Building air-conditioning) and SIA 380/1 (Thermal energy in building construction) as central standard works to be respected in AI applications. A sector-specific AI guidance in the style of the SAV bar regulation does not exist as of May 2026 – the sector aligns with SIA standards, cantonal authorisation practice and the revised FADP for project-related personal data.

Why AI pays off in 2026 for MEP and building planning

Four pressure points hit the sector simultaneously.

First: standard-body complexity. The SIA standard series comprises around 600 individual standards, plus MuKEn per canton, Minergie standards (Minergie, Minergie-P, Minergie-A, ECO-Plus), KBOB recommendations and federal laws (Energy Act, CO2 Act). An MEP planner has to identify the relevant set for every project – a task that AI RAG significantly accelerates in 2026.

Second: energy and CO2 balance in focus. With the MuKEn 2026 revision (expiry of MuKEn 2014 transition periods in most cantons) and the CO2 Act, requirements on energy efficiency, renewable heating systems and life-cycle assessment rise. AI-supported energy modelling with load profiles from real data is no longer a nice-to-have in 2026 but a competitive advantage.

Third: skills shortage and apprentice bottleneck. suissetec reports structural skills shortages in 2025 and 2026 in MEP planning offices and on building sites. AI tools that relieve repetitive tasks (schema generation, calculation templates, quantity surveying) give senior engineers more time for complex design questions and juniors more room for learning tasks.

Fourth: BIM mandate in public tenders. Federal and cantonal authorities in 2026 increasingly require BIM models as delivery format for larger projects. Pure 2D plans no longer suffice for federal building projects and several cantonal ones. AI tools for BIM enrichment (materials, manufacturer data, energy ratings, life-cycle data) relieve manual entry work significantly.

The 2026 point: AI-supported MEP planning is not "a toy for innovation offices", it is the precondition to meet the rising standard depth and time demands of public tenders.

Where AI works productively in MEP and building planning in 2026

Five application clusters cover the bulk of realistically automatable work today.

SIA standard search and MuKEn research with RAG. SIA standards (especially SIA 380/1 thermal energy, SIA 2024 building air-conditioning, SIA 384/1 heating systems, SIA 385/1 sanitary, SIA 387 lighting) and cantonal MuKEn requirements are indexed in a proprietary knowledge base. Questions like "What does SIA 380/1 require for specific heating power in residential buildings" return grounded answers with citations. Important: SIA standards are copyrighted; access is permitted only for SIA members with a standard subscription. See Retrieval-Augmented-Generation.

Schema and principle circuit-diagram generation with vision-capable LLM. From requirements (building type, heating system, ventilation concept, sanitary distribution) a vision-capable LLM generates proposals for principle schemes, heating distribution schemes and ventilation duct schemes. The engineer checks the scheme against SIA standards, adds project-specific components and integrates it into CAD plans. Precondition: models trained with schema data (Anthropic Claude Vision, GPT-4 Vision, local multimodal models).

Energy consumption modelling and load profiles. From building data (geometry, insulation, window glazing, occupancy) and climate data (SIA 2028 climate data) a model generates the annual energy balance, peak load profiles and CO2 balance. Validation against SIA 380/1 calculations and against measured consumption data of similar buildings. Saves hours in the pre-project stage and quantifies comparisons between heating system variants.

BIM model enrichment with manufacturer data. From the architect BIM model (Revit, ArchiCAD, Allplan) MEP components are extracted and enriched with manufacturer specifications, energy ratings, life-cycle data and cost. The MEP planner reviews the proposals, corrects and releases. Important: BIM enrichment can be automated, the design decision stays with the engineer.

Tender and bill-of-quantities preparation. From released plans an agent extracts quantities, suggests SIA NPK positions (standard positions catalogue) and generates a bill-of-quantities draft. The cost engineer or MEP planner reviews positions, corrects and supplements. Saves significant preparation time on larger projects.

Across all applications: SIA standards and manufacturer data are copyrighted – access in a RAG system must follow the licence. Project-related personal data (building owner, tenant lists, measurement data from smart meters) must be protected under the revised FADP. EU/CH hosting with DPA is standard.

How a MEP office starts with AI – in 6 steps

01Inventory and licence check: which software is in use (Plancal nova, MagiCAD, BIM solutions)? Which SIA standard subscriptions does the office hold? Which tools are used privately? Licence clarification for SIA standards and manufacturer data in a RAG system.
02Draft an internal AI guideline: permitted models and hosting regions, clear separation between AI preparation and engineer review, licence-compliant use of SIA standards, ban on shadow AI.
03Start SIA RAG as pilot: indexing of SIA standards included in the office subscription plus internal office guidance and past project reports. Access for all staff. Four to eight weeks of implementation.
04Decide hosting architecture: EU/CH hosting with DPA and no-training guarantee for all SIA standard and project data. For especially sensitive projects (authorities, critical infrastructure) local hosting (Llama 3.x, Mistral) on own servers.
05Schema and energy-model workflows as a second use case: vision-LLM schemes for typical design cases (heating distribution multi-family house, ventilation duct school), energy model for the pre-project stage. Mandatory review of every output by an engineer.
06BIM integration and tender automation as a third use case: BIM enrichment with manufacturer data, tender preparation with NPK proposals. Four-eye control on every tender. Quarterly review: time saved, error rate, staff acceptance.

Where a MEP office should start in 2026

Three stages, in this order.

Stage 0 – Inventory, AI guideline, licence check. Which software is in use (Plancal nova, MagiCAD, IGEA, BIM solutions)? Which SIA standard subscriptions does the office hold? Which tools are used privately by staff (ChatGPT, Claude, Microsoft Copilot)? An inventory and an internal AI guideline are the precondition. Licence check for SIA standards and manufacturer data is mandatory.

Stage 1 – SIA RAG as pilot. Realistic is the indexing of SIA standards covered by the office's standard subscription, plus internal office guidance and past project reports. Low risk (no personal data), high benefit (all staff gain research access). Four to eight weeks of implementation.

Stage 2 – Schema and energy-model workflows. After a successful pilot: vision-LLM schemes for typical design cases (heating distribution multi-family house, ventilation duct school), energy model for pre-project stage. Eight to twelve weeks.

Stage 3 – BIM integration and tender automation. Only after two successful use cases: BIM enrichment with manufacturer data, tender preparation with NPK proposals. Mandatory: licence compliance for manufacturer data, validation of every BIM output by an engineer, four-eye control on tenders.

For smaller MEP offices (3-10 staff) a collaborative approach with the trade association (suissetec, USIC) or specialist software providers (Plancal, MagiCAD extensions) is often more sensible than an in-house setup. Important: every licence question on SIA standards and manufacturer data must be clarified before productive use.

Where AI does not belong in MEP and building planning in 2026

Three areas where reservation in 2026 is advisable.

Design decisions without human review. A heating design, a ventilation airflow calculation, a fire-protection concept decision have consequences for safety, comfort and energy efficiency over decades. An AI-generated design may serve as a proposal, the check against SIA standards and the assumption of responsibility by the signing engineer stay human. The engineer signs – the AI does not.

Fire-protection concept decisions without a documented fire-protection officer. Fire-protection concepts are subject to cantonal fire-protection ordinances and the fire-protection standard of the Association of Cantonal Fire Insurances (VKF/AEAI). Responsibility lies with the fire-protection officer (BR-V), not with the model. AI can take over preparation, risk identification and research; the concept itself and the official submission belong to the BR-V.

Use of SIA standards or manufacturer data in unlicensed models. When a RAG system indexes SIA standards and generates answers, this is permitted only for persons holding a valid standard subscription. A cloud LLM learning on protected content breaches copyright and the SIA licensing model. EU/CH hosting with a no-training guarantee is not "best practice" here but a licence obligation.

Particularly delicate and not finally settled in 2026: fully automated generation of working plans without engineer review. Even though technically increasingly possible, the liability question for undetected errors (e.g. collisions, faulty design) is unresolved. 2026 practice: AI working plans are reviewed position by position by the engineer.

Trade-offs

STRENGTHS

SIA standard research takes minutes instead of hours – juniors benefit too
Schema generation with vision LLM significantly accelerates typical design cases
Energy modelling quantifies variant comparison early in the pre-project stage
BIM enrichment substantially reduces manual entry work – senior time freed for design
Tender and bill-of-quantities preparation more systematic with NPK proposals

WEAKNESSES

Licence duties for SIA standards and manufacturer data force controlled hosting
Responsibility for design stays with the engineer – AI does not replace SIA conformity review
BIM model validation requires position-by-position review – no full automation
Initial effort 8-15 days per use case plus 3 months accompaniment
Energy models must be validated against real-measured data – model drift possible

FAQ

May we index SIA standards in an AI tool?

Only with a valid standard subscription and only for subscription-entitled persons. SIA standards are copyrighted; open indexing in a cloud LLM system with model training or access by unauthorised persons breaches the licence. 2026 practice: private hosting of the RAG knowledge base (EU/CH hosting or local), access only for staff with a standard subscription, no model training on the protected content, documented subscription compliance.

How do we integrate AI in our BIM workflow?

Three steps. First: fix the BIM data model (IFC schema, manufacturer-specific extensions). Second: AI-supported enrichment in the pre-coordination phase – materials, energy ratings, life-cycle data are proposed, the MEP planner reviews and releases. Third: a validation step before every BIM delivery – collision check, SIA conformity check, quantity plausibility. Tools 2026: Autodesk Revit with AI add-ins, Allplan AI extensions, BIMcollab for issue tracking, in-house scripts for enrichment.

Are the MuKEn 2014 still current or is there a revision?

The MuKEn 2014 (cantonal model rules in the energy sector) remain the basis in 2026; a MuKEn 2025 revision is in consultation in several cantons, implementation follows cantonal practice. The transition periods of MuKEn 2014 expire in many cantons between 2025 and 2027 – new heating systems in 2026 typically must include renewable generation or reach an efficiency standard. The Cantonal Conference of Energy Directors (EnDK) publishes the current list. Practice recommendation: check the applicable cantonal MuKEn rules for every project.

What does suissetec say about AI in MEP offices?

suissetec sees AI in 2026 as an important efficiency lever in the context of the skills shortage. Events in 2025/2026 address AI in planning, execution and operation. A binding AI guidance at the depth of the SAV bar regulation does not exist – the sector aligns with SIA standards, cantonal MuKEn and the revised FADP. suissetec stresses that responsibility for design and execution stays with the qualified engineer or sanitary/heating technician – AI is a tool, not a replacement.

Sources

FITS YOUR STACK?

What this looks like in your business – a 30-minute intro call.

Book a call