Aipedia: Structural and civil engineers

Generative design and structural optimisation with AI

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

Generative and optimisation tools can propose lighter, lower carbon structures, but the engineer who adopts the result owns the design decision and every assumption behind it. Optimised structures often have less reserve, so errors show sooner.

Part of: AI in structural and civil engineering

In short

Optimisation software searches for the most efficient member sizes, layouts or forms against the objectives and constraints it is given. The output is only as good as those inputs: missing load cases, idealised connections, or no allowance for robustness and buildability can produce a design that is efficient on paper and fragile in reality. Part 2A reg 11K and CDM 2015 reg 9 put the duty on the designer, not the tool. PI usually responds to claims that the adopted design was negligent, but cover depends on the wording and on keeping a clear record of the engineering decisions.

How engineers use generative and optimisation tools

Uses range from topology optimisation for transfer structures and nodes, to automated sizing of steel and timber members, to generative layout studies that test column grids against carbon, cost and floor depth. Many tools now include AI components that learn from previous runs or suggest options. They are attractive because clients want lower embodied carbon and leaner frames, which our guide to AI embodied carbon calculations covers.

The engineer's role changes rather than shrinks. Instead of sizing each member, the engineer defines the problem: loads, limits, materials, connection types and what counts as a good answer. That framing is itself design work, and mistakes in it carry through to every option the tool generates.

The duties that apply

Part 2A reg 11K (England): take reasonable steps so the design, if built, would comply, and provide sufficient design information. Part 2A.
CDM 2015 reg 9: when preparing or modifying a design, a designer must take into account the general principles of prevention and eliminate, reduce or control foreseeable risks to those who build, maintain, clean or use the structure. Regulation 9.
Statement of Ethical Principles: Principle 5 asks engineers to consider fully the risks of AI and apply precaution where harms are serious or irreversible. Statement of Ethical Principles.
IStructE: senior engineer review of AI-produced design and regular tool evaluation.

What can go wrong

Optimisers remove what they are not told to keep. Typical problems include: robustness and disproportionate collapse provisions not modelled; connection stiffness idealised so real joints behave differently; erection stability and temporary conditions ignored; vibration or deflection limits set too loosely; and forms that are hard to fabricate, inspect or maintain, which raises CDM issues. Because optimised members have little spare capacity, a small modelling error has less margin to absorb it.

Illustrative scenario: an engineer accepts an optimised floor layout that meets strength checks and the client's carbon target. The model omitted a partition load allowance and assumed composite action the construction sequence did not provide. Floors deflect beyond tenant tolerances, fit-out is delayed and the developer claims for remedial strengthening and lost rent.

How PI usually responds

PI commonly responds to a claim that the design adopted was negligent, and it will not usually matter that an optimiser proposed it. What matters is whether a reasonably competent engineer would have accepted that design with those assumptions. Claims may also turn on what the engineer was asked to do and promised, a point covered in our scope of duty guide. Where a carbon or efficiency outcome was warranted rather than advised on, the position can be harder, and fitness for purpose wording in appointments is often outside standard PI. See engineer negligence claims examples.

Practical controls

Write down the objectives, constraints and load cases given to the tool before running it.
Check the preferred option with conventional analysis, including robustness and construction stages.
Review buildability and maintenance with the contractor and principal designer under CDM.
Keep a design decision record showing the engineer, not the tool, chose the option and why.
Avoid promising carbon or weight outcomes as guarantees in appointments.
Run a sensitivity check on the chosen option: change key loads and stiffness assumptions and see whether it still works.

How Apex helps

Apex arranges PI for structural engineers. A broker reviews your cover, your appointment terms on performance outcomes, and how your policy treats optimised and AI-assisted design.

Sources

Frequently asked

Who owns a design produced by a generative design tool?

For liability, the engineer who adopts and issues it. Part 2A and CDM place designer duties on people, not software. Intellectual property in tool outputs is a separate question governed by the licence terms, which should be checked before relying on them.

Are optimised structures riskier from a PI point of view?

They can be, because there is less reserve capacity to absorb modelling errors, and robustness or construction stages may not be modelled. Good records of assumptions and independent checks reduce that risk and help defend a claim.

Do CDM duties apply to choices made by an optimisation tool?

Yes. CDM 2015 reg 9 applies to whoever prepares or modifies a design. If an optimised form is harder to erect, maintain or clean safely, the designer adopting it must consider those risks and eliminate, reduce or control them as far as reasonably practicable.

Does PI cover a missed carbon target from an optimised design?

It depends on whether you advised with reasonable care or guaranteed an outcome. PI usually responds to negligent advice, but contractual guarantees and fitness for purpose obligations are commonly limited or excluded. Check the appointment and the policy wording.

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.