AI presents itself as an attractive means of research, analysis and document preparation. Seeking advice, analysing critical information and drafting documents appear amongst the most common uses for AI. Its attractiveness is drawn from AI’s ability to produce material that uses flowery language, appears to be logically written and is confident. In the world of construction contracts, this proposition presents the attractiveness of giving AI a contract, project information and necessary data so that it can help manage the contract faster than a human. At this point, enter the prompt: “Draft a contract claim in terms of the JBCC 6.2 Principal Building Agreement for an extension of time with costs based on an ongoing delay to access caused by a direct contractor.” Let’s remember this prompt for later reference.
The result is that AI may be able to draft contractual claims. It can summarise clauses and quote them. Therefore, the attraction is obvious. However, construction contracts are not writing exercises. They are legal and commercial instruments in which a wrong conclusion can cost considerably more than time saved by reaching that conclusion quickly.
Returning to the prompt suggested earlier, AI will usually dish out a fairly good structure for the claim, including suggestions on the contractual clause references with a host of other advisory suggestions. But did AI consider whether a notice or a claim is appropriate in terms of the contract considering the stage of the delay? Whether the delay impacts the project critical path? Whether contractual preconditions have been met? Whether the time periods for a notice or claim are being satisfied? Whether there is an underlying entitlement for additional time or costs? Whether the claims procedure is being complied with?
The consideration of “Can AI manage a construction contract?” is now amplified by “What happens when AI manages a construction contract incorrectly?” The point at this juncture is that generating the wording of the claim is often the easy part. Getting the process technically correct is the difficult part. The key issue is that AI generates a false sense of comfort by providing well-written responses which also appear to be logically structured but may be wrong.
In 2025, Donrich Thaldar conducted a study entitled “Can AI think like a lawyer? Evaluating Generative AI in South African Law”, in which it was found that AI hallucinates case law, misapplies legal principles or provides an incomplete legal analysis. AI was found to provide unreliable retrieval and application of case law (source: Thaldar, D - "Can AI Think Like a Lawyer? Evaluating Generative AI in South African Law" [2025] PER 41). Now, picture a scenario where a claim is submitted in reliance on incorrect or non-existent authority. Applying this more practically to the contract, what certainty is there that the correct clauses or interpretations are being referenced?
A further complication is that construction contracts are complex and include inter alia general terms, special terms, bills of quantities, standards of measurement, SANS codes, tender clarifications and factual considerations. Their management goes beyond clause identification, and overlooking these considerations may result in incorrect conclusions being drawn.
AI could be used to simplify the text of a contract document and extract the roles and obligations of the parties, providing plain English summarisation. It can also assist you to extract every notice requirement or obligation in a contract or even alternative wording to a clause. The assistance that AI will provide depends largely on the prompts inserted and the background information that is shared. It is recommended that pure legal research is avoided and that final enforceability, strategic insights and statutory compliance is done with human oversight.
In conclusion, AI works well as a supplementary tool, however the challenges of relying exclusively on AI must be understood.
Email to a Friend
AI presents itself as an attractive means of research, analysis and document preparation. Seeking advice, analysing critical information and drafting documents appear amongst the most common uses for AI. Its attractiveness is drawn from AI’s ability to produce material that uses flowery language, appears to be logically written and is confident. In the world of construction contracts, this proposition presents the attractiveness of giving AI a contract, project information and necessary data so that it can help manage the contract faster than a human. At this point, enter the prompt: “Draft a contract claim in terms of the JBCC 6.2 Principal Building Agreement for an extension of time with costs based on an ongoing delay to access caused by a direct contractor.” Let’s remember this prompt for later reference.