Aipedia: Building services engineers

AI heating and cooling load calculations

Aipedia, by the Apex broking team · Last reviewed 2026-10-27

AI can speed up heat loss, cooling load and plant sizing calculations, but the engineer remains responsible for every input and for the plant size that results. Undersized or oversized plant can lead to M&E claims, and AI makes input errors easier to miss.

Part of: AI in building services engineering

In short

AI tools now take a model or drawing set, infer fabric, occupancy and internal gains, and return heating and cooling loads and suggested plant sizes. The risk is not usually the arithmetic but the assumptions: wrong U-values, missing glazing, generic occupancy or the wrong design conditions. CIBSE Code clause 4.2 requires the engineer to be competent to check the result, and Part 2A regulation 11K requires a design that would comply if built. Comfort and capacity claims usually allege negligent design, and PI normally responds subject to the wording and any fitness for purpose obligations.

How AI is used for load calculations

Typical uses include extracting room areas and fabric data from BIM models, populating heat loss and cooling load software, suggesting internal gains and occupancy profiles, sizing emitters, chillers and boilers, and writing the calculation summary that goes to the client or contractor. Some tools learn from previous projects and propose loads by building type before any detailed model exists.

That speed is attractive at stage 2 and 3 when plant space and electrical capacity need fixing early. The problem is that early AI estimates often become the basis for plant rooms, risers and incoming supplies, and are never properly reworked once the design firms up.

The duties that apply

CIBSE Code clause 4.2: only undertake work you are competent for, or under competent supervision. An engineer signing off loads must be able to judge whether the inputs and outputs are reasonable. CIBSE Code.
Statement of Ethical Principles, Principle 3: work within competence, with the right expertise especially in novel or emerging areas. Statement of Ethical Principles.
Building Regulations Part 2A: regulation 11F (competence) and regulation 11K (designers must ensure the design would comply if built and give sufficient design information). Part 2A.
CIBSE Code clause 1.3: no statements known to be untrue or misleading, which covers calculation summaries presented as checked when they were not.

What can go wrong

Common failure points are fabric assumptions copied from a template, glazing ratios read wrongly from drawings, internal gains for the wrong use class, design conditions for the wrong location, and ventilation or infiltration rates that do not match the ventilation strategy. AI-generated summaries rarely flag these, so a reviewer sees a clean output with no obvious error.

Illustrative scenario: an AI tool populates cooling loads for an office fit-out using generic occupancy and equipment gains. The engineer checks totals against a rule of thumb, which looks reasonable, and issues the schedule. The tenant's actual use is far denser. After occupation, meeting rooms overheat, the tenant complains, and the landlord claims for replacement plant, enabling works and loss of use, alleging the design did not meet the brief.

Checking AI loads without redoing everything

A proportionate check does not mean recalculating every room by hand. It means confirming the inputs that drive the result: fabric and glazing against the drawings, occupancy and equipment gains against the brief, design conditions against the location and ventilation rates against the ventilation strategy. Then spot-check a sample of critical rooms independently, such as corner offices, south-facing meeting rooms and spaces with high equipment loads. If the AI result and the independent check disagree by more than the engineer would expect, the whole run should be treated as unreliable until the cause is found.

How PI usually responds

A claim that plant was undersized or did not meet a stated performance is usually framed as negligent design, and PI commonly responds subject to the policy terms. Difficulties arise where the appointment includes performance guarantees or fitness for purpose obligations, which policies may not cover in full; see fitness for purpose versus reasonable skill and care. The M&E specification claim scenario shows how specification and sizing claims tend to develop, and scope of duty in engineer PI claims explains how the scope of your appointment shapes the loss you can be liable for. Using AI does not usually change cover in itself, but insurers may ask how outputs are checked.

Practical controls

Record every input the tool used: fabric, glazing, gains, occupancy, design conditions and ventilation rates.
Have a competent engineer check inputs against the drawings and brief, not just totals against rules of thumb.
Mark early AI estimates as provisional and rework them when the design is fixed.
Confirm occupancy and use with the client in writing before plant is selected.
Keep the checked calculation, not only the AI summary, on the project file.

How Apex helps

Apex arranges PI for building services engineers. A broker reviews your cover and how your policy treats AI-assisted load calculations and plant sizing, including any performance obligations in your appointments.

Sources

Frequently asked

Can I rely on an AI tool for heat loss calculations?

You can use one, but you cannot hand it responsibility. CIBSE Code clause 4.2 and Part 2A regulation 11F expect a competent engineer to check inputs and results, and the designer remains responsible for the plant size that follows.

What usually goes wrong with AI load calculations?

Mostly inputs: fabric data, glazing, internal gains, occupancy and design conditions. The calculation engine may be fine while the assumptions are wrong, and a neat summary can hide that.

Will PI cover a claim for undersized plant?

PI usually responds to claims that the design was negligent, subject to the wording. Performance guarantees or fitness for purpose terms in an appointment can create liability a policy does not fully cover, so check contracts before signing.

Should I tell clients I use AI for load calculations?

There is no CIBSE rule requiring it, but being clear about your method and the assumptions you relied on, and confirming occupancy and use in writing, makes later disputes easier to resolve.

Check your PI covers how you use AI

Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start a PI quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.