AI in Construction: The Second Wave Is Here

In the past week , I’ve had more conversations about the power of AI with people with different background in the construction industry than I’ve had in the past year combined. Project managers, contract administrators, engineers, commercial directors, dispute resolution specialists, you name it. They all wanted to talk about one thing: what next.

This wasn’t the usual “What is AI?” curiosity I saw in the first wave. The tone has changed. It’s sharper, more urgent, almost tinged with competitive anxiety. People aren’t wondering if AI will change construction anymore, they’re wondering how fast and how deeply it will happen. They want to know what it means for tenders, for live contracts, for disputes, for project delivery. They’re looking past the hype and asking: What will I be doing differently six months from now if I take this seriously? And what will I miss if I don’t?

This is why I call it the second wave. The first was awareness. This one is action time.

From the First Wave to the Second

Let’s rewind for a moment. The “first wave” of AI in construction started when generative AI hit mainstream awareness with ChatGPT 3.5 and 4. People dabbled. Some asked it to write tender letters. Others used it to summarise meeting notes or look for alternative contract wordings. The results were often surprising, sometimes accurate, sometimes laughably off, but it was enough to start the conversation.

In those early days, the technology was impressive but clunky. There was a clear gap between what was technically possible and what was safe, reliable, and ready for daily project work. Adoption was hesitant, especially come to construction industry, as it is a conservative industry by nature; it takes time to trust new tools, especially when they deal with sensitive contracts, high-stakes negotiations, or multimillion-dollar disputes.

Fast-forward to today, and the capability gap has narrowed dramatically. ChatGPT-5.0 is not just a text generator, it’s multimodal, meaning it can process and reason over text, images, code, and even structured data in ways that were out of reach just 18 months ago. It can ingest hundreds of pages of a contract, cross-reference them against your company’s playbook, and flag deviations in seconds. It can take a construction programme, identify the most likely critical path risks, and suggest mitigation measures before they happen. And it can do all of this in language that’s understandable to both lawyers and site managers.

This is why so many people I’ve spoken to this week are using words like “accelerating,” “compressing,” and “overtaking.” The time between a technological leap and its practical application is collapsing. What used to take years to filter into daily construction practice could now take months, or even weeks, if the right people decide to move.

Why Contracts and Disputes Are Prime for AI

In all the conversations I’ve had, one theme keeps surfacing: if AI is going to make a meaningful difference anywhere in construction in the short term, it’s in contracts and disputes.

Think about it. Contracts are vast, detail-heavy documents packed with obligations, timelines, and dependencies. They’re written in dense language that’s often interpreted differently by different parties. They’re accompanied by a mountain of related documentation, tender clarifications, meeting minutes, correspondence, RFIs, site diaries, programmes, risk registers. Together, these documents contain the complete history of a project’s risk allocation and delivery. Yet in most projects, we don’t analyse this data until something has gone wrong.

AI flips that pattern.

It can read, cross-reference, and summarise massive volumes of information in minutes. It can detect inconsistencies between a technical spec and the contract’s wording. It can highlight obligations with looming deadlines before they’re missed. And because it can be trained on your own past disputes, it can spot clauses and situations that historically have led to problems.

Imagine you’re in the early stages of tendering for a major civil project. The client’s draft contract is 140 pages long. Your in-house team might take several days to review and mark up the high-risk clauses. An AI assistant could run through the same document in under an hour, compare each clause to your standard terms, identify non-negotiable red flags, and even suggest alternate wording based on previous successful negotiations. You’d still have your legal and commercial experts review the results, but the heavy lifting of detection and first-pass analysis is done.

Now picture you’re midway through a complex build. Progress meetings are tense, and the programme is starting to slip. AI could scan the last month’s RFIs, variation requests, and site diaries, then match them to clauses on time extension and notice requirements. If it sees a pattern, say, multiple late approvals from the client’s team, it could flag this as a potential EoT entitlement before you’ve missed the notice window.

In disputes, this is equally powerful. Instead of a claims team sifting manually through thousands of documents for evidence, AI can group correspondence by theme, identify missing records, and create timelines that show when each party knew (or should have known) about key events. This doesn’t replace the role of experienced claims consultants or legal counsel, it amplifies them.

The Speed Problem

While the potential is clear, the pace is both the opportunity and the danger.

In the first wave, AI’s progress was exciting but manageable. You could take your time experimenting with low-stakes tasks and deciding whether to invest further. In this second wave, the technology is improving faster than most companies’ ability to adopt it.

Here’s the uncomfortable truth: if you wait until AI in construction is “fully proven,” you’ll be years behind. Early adopters aren’t just learning to use the tools, they’re building proprietary knowledge bases that will become competitive moats. A contractor that’s been feeding its AI system structured data from every project since 2024 will, by 2027, have a living library of risk patterns, negotiation strategies, and dispute outcomes. That’s not something you can buy off the shelf.

This creates a widening gap between the “AI haves” and “have-nots.” And because AI thrives on data, the advantage compounds. The more you use it, the better it gets, faster than any human team could improve through traditional learning.

What the Second Wave Looks Like in Practice

To understand what this shift really means, imagine a day in the life of a contract manager on a large infrastructure project in 2026.

In the morning, before the site meeting, your AI assistant briefs you: the client issued two instructions last week that affect the programme. Based on the current progress data, there’s a 60% probability this will delay a critical activity. It recommends issuing an early warning notice today, and provides a draft with clause references and supporting evidence from the last three weeks’ correspondence.

At lunchtime, you get an alert. The subcontractor’s invoice just landed in your inbox. The AI system has already matched each line item against the contract rates and flagged three that exceed agreed limits. It suggests wording for a polite but firm clarification email.

In the afternoon, you sit down with the project team to review the risk register. The AI tool has re-ranked the top 10 risks based on new site data, flagged that the insurance coverage for one risk is insufficient, and suggested actions based on similar situations from your last five projects.

By the end of the day, you’ve dealt with more risks, more contract compliance, and more proactive dispute avoidance than most managers in the first wave could handle in a week.

And you’ve done it without working late.

Risks, Misuses, and Ethical Boundaries

It’s tempting to see this as a silver bullet. It isn’t.

AI makes mistakes. It can hallucinate facts or misinterpret ambiguous clauses. If you treat its output as final rather than advisory, you risk acting on faulty information. It can also replicate biases from the data it’s trained on, if your historic dispute outcomes are skewed toward one party’s advantage, your AI’s recommendations may be too.

Confidentiality is another major concern. Many AI systems rely on cloud processing. If you upload sensitive contract data to a public tool without proper safeguards, you may be breaching confidentiality or even privilege. Any serious AI adoption in construction needs strict data governance, clear boundaries on what can leave your network, and a way to audit how decisions are made.

There’s also the ethical dimension. If AI is capable of predicting which subcontractors are most likely to default, or which variations are most likely to be approved, how do you ensure that information isn’t used to manipulate rather than manage? Where’s the line between smart risk management and strategic exploitation? These are questions the industry must face before the technology becomes fully embedded.

What to Do in This Second Wave

1. Experiment on one live project. Don’t try to overhaul your whole operation at once. Pick a project and use AI for one or two specific tasks, clause comparison, correspondence summarisation, programme risk scanning. Measure the results and refine your approach.

   
   

2. Design for machine-readability. AI works best with clean, structured inputs. Start building contracts with annexes in tables, milestones with explicit dates, and consistent terminology. The easier it is for a machine to read, the more accurate the outputs.

   
   

3. Build guardrails. Decide early what AI can recommend, what needs human sign-off, and how outputs will be verified. Treat AI as a junior colleague, capable, fast, but always under supervision.

   
   

4. Focus on people, not just process. Use AI to free up time for the things only humans can do well: negotiation, relationship building, proactive communication. Most disputes are prevented in conversations, not in clauses.

   
   

5. Capture the learning. Every AI-assisted decision should feed into an internal knowledge base. Over time, this becomes your company’s unique competitive asset, one that competitors can’t easily copy.

   
   

A year ago, AI in construction felt like a novelty. Today, it feels like a race. The people I’ve spoken to this week aren’t asking whether the wave is coming, they can already feel the pull of the current. The second wave is here, and it’s bigger, faster, and closer than most of us expected.

The choice is simple. You can start now, experiment, adapt, and build the capabilities that will carry you forward, or you can watch from the shore as others learn to ride it. The technology won’t wait for you. Neither will your competitors.

The wave is here. And it’s your move now.