AI Is Already on Your Jobsite. Is It in Your Subcontract Yet?

This isn't speculative — construction firms are actively using AI tools for scheduling optimization, safety monitoring (flagging hazards from jobsite camera feeds), supply-chain coordination, and increasingly autonomous agents handling multi-step operational tasks. AGC's 2026 outlook survey found 61% of firms already using or investing in AI, rising to 78% among mid-size general contractors — the segment most subcontractors report to directly, day to day.

That adoption is happening at the operational level, often faster than the contractual framework around it is being updated. A GC's project team might be using an AI scheduling tool this month that wasn't in use when the current subcontract template was last revised.

This gap between operational adoption and contractual catch-up is a familiar pattern with fast-moving technology generally — the tools tend to show up on the jobsite well before legal and risk-management teams have finished updating the paper that's supposed to govern their use.

It's worth directly asking your GC contacts which specific AI tools, if any, they're using for scheduling or safety on your active projects — most project teams are happy to describe their own workflow, and the answer tells you concretely whether this risk is theoretical or immediate on a given job.

This isn't a hypothetical concern — it's being actively raised in current legal commentary. A Construction Seyt legal blog piece from April 2026 notes explicitly that AI tool use is outpacing the contract language governing it: subcontracts often don't specify whether AI-generated estimates, safety flags, or scheduling outputs create liability, and inconsistent AI-related permissions across contract tiers create real risk gaps between a prime contract and the subcontracts flowed down from it.

That last point matters specifically for subcontractors: even if a prime contract has been updated to address AI use and liability, there's no guarantee that language has been mirrored down into your subcontract — which means you could be operating without any contractual clarity on a risk that technically already exists on your project.

This inconsistency between contract tiers is worth flagging to your own counsel or review process even if you can't fully resolve it yourself — knowing that the gap exists, and where, is itself useful information heading into a negotiation.

Consider a concrete scenario: a GC's AI-powered scheduling tool generates a sequence that assumes your crew can complete a task faster than is realistically possible, and that schedule becomes the baseline against which your performance — and potential liquidated damages — gets measured. Or an AI safety-monitoring system flags (or fails to flag) a hazard that affects how your work is sequenced or halted. In both cases, the question of who bears responsibility for an AI-driven error, versus a human-driven one, is genuinely unresolved unless the contract says something specific about it.

This is a fundamentally new category of contractual gap — not a variation on a familiar risk like weather delay or a differing site condition, but a novel one created directly by a fast-adopted technology outpacing the paper that's supposed to govern the relationship.

A useful practical test: if you can't point to a specific clause that would answer "who's responsible if the AI tool got this wrong," that's a real gap, not a hypothetical one — and it's worth raising the question directly with the GC rather than assuming it's covered somewhere in the general liability language.

Most contract review — whether done by an attorney, an internal team, or even an AI-assisted tool — is built around a playbook of known, established risk categories: indemnity, payment terms, liens, delay, insurance. AI-attributed liability isn't yet a standard line item in most playbooks, simply because it's a genuinely new category that didn't need addressing five years ago. That means it's exactly the kind of risk that can slip through even a careful review, if the review isn't specifically looking for it.

As AI tool use on jobsites continues at the pace the AGC data shows, this gap is likely to close over the next year or two as standard contract language catches up — but in the meantime, it's a live, unaddressed risk on many active subcontracts.

This is also a good argument for periodically revisiting your own firm's standard review checklist, not just relying on whatever list of risk categories it was built around originally — new categories like this one emerge faster than most static checklists get updated.

Look specifically for whether your subcontract addresses AI-generated outputs — schedules, safety assessments, estimates — at all, and if it does, whether it clearly assigns responsibility for errors in those outputs. If it's silent, that's worth raising, particularly on any project where you know the GC is using AI-driven scheduling or safety tools directly.

This is precisely the kind of emerging, easy-to-miss risk that a thorough, currently-updated contract review should be built to catch. See how RCS keeps its review current with emerging construction-contract risk, including gaps like this one that didn't exist in older standard playbooks.

Expect standard contract language to catch up on this over the next couple of years, much as it eventually did for other technology-driven risk categories in the past — until then, the burden of catching the gap falls on whoever is actually reviewing the contract in front of them.

This is a good candidate to add explicitly to your firm's own standard review checklist right now, well ahead of when it becomes a standard industry line item — being early on a risk category like this one is a genuine, if modest, competitive advantage.