AI in construction: the bad and the ethical

In our companion article, "AI in Construction: The Good", we explored the current state of artificial intelligence, its practical applications across the construction sector and the governance and legal frameworks that organisations need to understand. That article made the case that AI presents a genuine and significant opportunity for an industry that remains one of the least digitalised sectors of the economy — but that realising that opportunity depends on robust governance, clear accountability and informed decision-making. This article turns to the other side of the equation. Governance and ethics are distinct but connected: governance concerns frameworks, accountability and compliance, whereas ethics addresses the harder questions of judgment, fairness, trust and the kind of organisation one aspires to be. In the AI context, it is the ethical questions where much of the long-term risk resides. What follows examines the key risks that construction organisations cannot afford to ignore, the ethical considerations that must inform AI adoption and the practical steps that leaders should be taking now.

The Bad: Eight Risks You Cannot Ignore

1. Hallucination

AI systems can still produce highly convincing but completely incorrect outputs. Confident does not mean accurate — and in construction, where a wrong number or a missed clause can have significant consequences, that matters enormously.

2. Bias

Poor or biased training data can cause AI to reproduce — or even amplify — existing patterns of discrimination and unfairness.

3. Cyber and data exposure

Connecting AI to internal systems creates new vulnerabilities for cybersecurity, confidentiality, and data leakage.

4. Autonomous action

When AI agents move beyond answering questions and start taking actions, the risk profile changes fundamentally.

5. Accountability gaps

When an AI-assisted decision causes loss, responsibility may be genuinely unclear — whether it sits with the developer, the supplier, the employee, or the organisation. This is already a live legal question and it is not yet settled.

6. Deskilling

Over-reliance on AI can gradually erode professional judgment and institutional expertise — and that process can be invisible until it matters.

7. Physical consequences

This is specific to construction; in most sectors, an AI error produces a bad document. In construction, it can affect money, safety, buildings, and physical infrastructure. The stakes are different here.

8. The so-called "assurance vacuum"

The greatest risk may not be any of the individual risks described above — it may be the danger of automating processes without adequate assurance sitting around them.

The overall message is not that AI should be avoided. It is that adoption must be matched by oversight, accountability, security, human judgment and genuine assurance. The question is how organisations translate that principle into practice.

Seven Things Construction Leaders Should Do Now

1. Experiment

Organisations that wait for AI to fully mature will surrender both knowledge and competitive advantage. Start a live pilot — in a controlled way, with governance in place.

2. Pick high-value use cases first

Focus on clearly defined applications where AI can produce measurable improvements in cost, quality, or speed. In construction, tender analysis, variation management, and safety monitoring are strong starting points.

3. Build governance before you scale

Put a framework in place before widespread adoption makes it difficult to control. It is far easier to do this at pilot stage than after the technology is embedded across the business.

4. Find your shadow AI

Understand where AI is already being used informally, without governance or disclosure. Your project managers could be using consumer AI tools to draft programme updates, variation notices, and commercial correspondence. If you do not know about it, you cannot govern it.

5. Invest in AI literacy across the business

Investment into AI literacy should take place across the business, not just inside the silos of technology teams. Your commercial, legal, safety, and operations teams all need to understand what AI can and cannot do. That literacy is what makes governance real rather than theoretical.

6. Make human accountability explicit

Ensure that human judgment, assurance, and accountability are explicit parts of every important AI-enabled workflow. "The AI recommended it" must never be treated as an acceptable answer.

7. Review your supply-chain contracts

If a subcontractor uses AI and it causes a delay or a defect, does your contract address that?

The organisations that succeed in this space will not be those that adopt AI fastest. They will be those that combine innovation, discipline, and trust most effectively.

The Ethical Question: Trust, Accountability and Human Judgment

At the heart of the ethical question is this: should a particular decision be delegated to a machine at all? That is the question organisations need to be asking — not just "can AI do this?" but "should a machine be making this call?"

For some decisions, the answer is clearly yes — routine, low-stakes, data-processing tasks where AI is more reliable than humans. But for others — decisions affecting employment, safety, planning, access to services, or legal rights — the answer is that humans must stay meaningfully involved. Not simply notified after the fact, but genuinely in the loop.

Transparency matters enormously here. People should understand when AI has materially influenced an outcome that affects them. In construction, that could mean a subcontractor understanding that AI played a role in their exclusion from a tender. Or a site worker understanding that an AI system flagged their conduct. AI systems used in high-impact contexts must be tested for fairness, reliability, robustness, and unintended consequences before they are deployed — and monitored thereafter. Explainability is not optional when the stakes are high. The broader societal impacts deserve attention too: job displacement, inequality, and the concentration of technological power are not abstract concerns. In a sector that employs millions across the supply chain, they are very real.

It follows that trust is a competitive advantage. The organisations that demonstrate responsible, governed AI adoption are likely to be the ones that clients trust with major programmes, that talent chooses to join, and that regulators treat as partners rather than subjects. Done well, this is not a constraint — it is a differentiator.

Final thoughts

Artificial intelligence is likely to become one of the most consequential technologies the construction sector has encountered in decades. Its potential to improve productivity, enhance decision-making, strengthen risk management and optimise asset performance is considerable. Equally, however, its adoption introduces new legal, regulatory and ethical challenges that cannot be ignored.

The key issue is not whether AI will become part of the construction industry, but how organisations choose to deploy it. Those that combine innovation with effective governance, clear accountability and informed human oversight will be best placed to realise the benefits while managing the risks. In that sense, the future of AI in construction will not simply be determined by technological capability, but by the quality of the decisions made around its use.