Artificial intelligence is quietly exploding onto construction sites in Ontario, large and small. Not just in the form of drones or BIM, but in the scheduling software that governs how projects are planned, sequenced, managed and ultimately executed. Tools that use machine learning to generate and optimize project schedules are being marketed to owners, general contractors, and project managers as a faster, smarter alternative to traditional CPM scheduling. And from what little I have already seen, are genuinely impressive.
But what happens when the AI gets it wrong? When a schedule generated or optimized by an algorithm contains errors that cascade into delays, additional costs, or even a failed project. Who eats the cost?
Although I did have a client ask me about this in the abstract, the issue is for now, to my knowledge, hypothetical. However, it is unlikely to remain that way. It is one that construction parties and their lawyers will be answering in arbitration, court or adjudication over the coming decade.
The Problem with Trusting the Black Box
Needless to say, AI scheduling tools are only as good as the data fed into them. These systems learn from historical project data, productivity assumptions, and sequencing logic. However, they cannot account for site specific conditions, novel scopes of work, or the judgment of an experienced scheduler who has executed similar projects before.
When an AI-generated schedule underestimates the duration of a critical path activity, or fails to flag a logical sequencing conflict, the downstream effects can be severe: subcontractors are mobilized too early, procurement falls out of sync, and the project falls behind before the first concrete is poured.
The more dangerous scenario is when no one questions the schedule at all. Anyone who has experience using AI knows that when it comes from good software, it looks authoritative.
Where Will Liability Land?
While there do not appear to be any reported cases, or adjudication decisions involving AI-generated scheduling (or similar technology) this question will likely turn on a familiar framework applied to unfamiliar facts.
The software vendor. If a construction scheduling tool makes representations about its accuracy or reliability, and the schedule it produces contains fundamental errors, there may be a product liability or negligent misrepresentation claims. However, most vendors disclaim liability aggressively in their licensing agreements (if they do not, they will learn to!), and courts have quite rightly generally been reluctant to hold software developers liable for how professionals use their tools.
The scheduler or consultant. This is where exposure is most likely to concentrate. A licensed project controls professional who adopts an AI-generated schedule without critically reviewing it, validating its assumptions, or exercising independent judgment has very arguably fallen below the standard of care. No such tool changes the professional’s obligation to deliver a competent, accurate schedule. Delegating judgment to an algorithm is unlikely to constitute a viable defence. It certainly has not been so in the legal context, where lawyers have been found using hallucinated cases.
The contractor. If a GC submits an AI-generated baseline schedule to an owner without disclosure, and that schedule forms the basis for the project’s progress monitoring and delay analysis, the GC may find itself unable to rely on it when disputes arise — particularly if the schedule’s deficiencies contributed to the project’s problems.
Practical Steps to Protect Yourself Now
None of this post is meant to discourage clients and parties from using AI to optimize the product they deliver to market. In fact, if you are not exploring how AI can improve your business, you may well be falling behind. That said, whether you are an owner, GC, or consultant, the following risk management steps apply:
- Disclose AI use in scheduling. If your schedule was generated or materially optimized by an AI tool, say so in your contract documents and project communications.
- Require human sign-off. Any AI-generated schedule should be reviewed and certified by a qualified scheduler before submission or reliance.
- Audit your software contracts. Understand what liability your scheduling software vendor has disclaimed (and whether your own professional liability insurance covers AI-assisted work product).
- Document your review process. If a dispute arises, you will want to demonstrate that the AI output was critically evaluated, not simply accepted.
About the Author: Mark Gallagher is a partner at Soloway Wright LLP whose practice focuses on construction litigation and dispute resolution. He represents owners, contractors, subcontractors, suppliers, and design professionals in matters involving construction liens, delay claims, contract disputes, deficiencies, procurement issues, and commercial litigation. Mark regularly appears before Ontario courts and adjudicative tribunals, providing strategic, practical advice to clients navigating complex construction projects and disputes.
