The use of AI in construction projects is growing fast. AI is assisting with tasks as varied as building and infrastructure design to worksite management, often making the work safer and more time and cost efficient. The technology is likely to be woven into the fabric of almost all aspects of project planning and execution in the coming years. However, like any tool, it can also have the opposite effect to that intended, and cause delay, increased costs and other issues. Most standard form contracts for large construction projects, including FIDIC and NEC, do not specifically address the use of AI. This can lead to liability gaps, claims or, at best, unnecessary and time-consuming inter-party correspondence when issues occur with the project AI. Therefore, when negotiating or amending such contracts, legal counsel need to take account of the technology’s deployment to allocate AI-related risk in a way that minimizes the likelihood of disputes.
What is clear at the outset is that AI is ideally suited to overcoming some of the most common causes of inefficiency on construction projects. In particular, its ability to synthetise and process vast amounts of real-time information alleviates interface difficulties caused by fragmented stakeholders – e.g., general and specialised contractors, engineers, architects and suppliers – exchanging information inadequately, inconsistently and/or asynchronously.
Common current use cases for AI in construction include:
- Project design – taking account of parameters like cost, load limits, energy efficiency etc., generative AI is used to create myriad design variations.
- Supply chain management – predictive analytical AI can study past lead times, prices and supplier performance, together with current and future supply needs, and suggest the best supplier or delivery routes for a good or service in terms of quality, price and time constraints.
- Project management and dynamic scheduling – AI can factor in real-time changes (e.g. weather, labour changes) into otherwise static schedules and suggest the most efficient reallocation of resources (such as optimizing workflows) to minimise delay.
- Quality control – AI ca improve and add to QC methods e.g., LiDAR (Light Detection and Ranging) scanners, whether attached to drones or handheld devices, generate precise 3D representations of construction sites. AI systems then compare these models to the original design plans, identifying inconsistencies that may not be noticeable to the human eye. Advanced computer vision algorithms can also process images to identify faulty installations or cracks before those defects cause bigger problems.
Nonetheless, AI can give rise to novel legal questions. Initial ways to respond to these questions are included below.
- Design – if the design is to be carried out (mainly) by AI, how much verification is it reasonable for the contractor to carry out? Ideally, the contract should set this out.
- Materials – in the case of AI procurement platforms, if the AI was trained on outdated data, it may not suggest materials that comply with ESG obligations of certain building codes or company/joint venture ESG commitments. The obligor may wish to seek warranties in this respect before delegating material procurement to another (or third) party’s AI.
- Extent of AI use, errors and exemptions – say that AI use is mandated by the employer and an (assumed) AI-enabled learning curve is built as an assumption into the project pricing. In practice, the contractor does not use AI as much as required and assumed efficiencies are lost. The contractor claims the AI was hallucinating or was otherwise wrong. This is easy to imagine where, for instance, a contractor ignores scheduling or supplier changes suggested by AI in response to an adverse weather event and the employer argues that the critical path was needlessly extended as a result. Complex evidential and causation questions may arise in such circumstances.
- In anticipation of such a scenario, the contract should provide for adequate prior training of staff in the use of AI; set out clearly what “checks” the user/operator is to carry out while using AI, and require that party to cease using AI where it is shown to be materially unreliable. For example, where the contractor is considering deviating from the envisaged/agreed use of AI, the contract may require him to carefully document both the episodes of the AI’s unreliability (typically straightforward via export) and the contractor’s reasoned alternative approach (an AI “exemption”), and to raise the matter as soon as possible with the employer. The contractor would be well advised to include relief in the event of AI error where, prior to the use of AI, the contractor would have otherwise been able to rely on requests for information, clarification or instructions from the engineer, with time and/or cost consequences where the engineer did not properly provide them. It will be for the party who contracted the AI from the AI developer/vendor to recover in turn from the developer/vendor where the AI itself (rather than how it was used) is at fault.
- Intellectual property – since the AI will learn from its deployment on the project, whether or not in a technology-specific way, the related IP rights should be clearly attributed as between the user, developer or third parties to avoid possible infringement issues. Data privacy feeds into this since the AI may be trained on data of project stakeholders whose consent to the same should be obtained. All necessary data privacy regulations should be carefully followed.
- Insurance – many insurance policies may not cover AI-caused delays and so the parties will need to determine whether to seek additional cover and, if so, how that cover is paid.
In conclusion, and as a basic matter, the contract should address, among other things, what AI can or must be used on the project; by whom and when; what standard of care is required when checking AI output; what the protocol is where the AI’s output is inaccurate; what rules govern the ownership of the output and what insurance policies are to apply and who is to pay for them.
Please get in touch if you have any questions concerning the use of AI on construction projects as it was not possible to cover many points in a short blog post.