The professional indemnity insurance market has been soft for two years. Rates fell at what Clyde and Co described as "unprecedented speeds" in 2024, driven by growing capacity from MGAs and increased competition. Construction firms have benefited. That softness may not last.
The variable that is hardening underwriter attention, even while rates stay low, is AI. More than 60% of PI market professionals now say that cyber and technology, including AI, will further drive demand for PI cover. And 72% believe AI will increase the risk of practice, according to the same Clyde and Co survey. The market has not yet repriced this. But it is watching.
AI Exclusions Are Already Appearing
In the US, several major carriers have begun filing AI exclusions with state regulators. Berkley introduced what legal firm Hunton Andrews Kurth describes as an "absolute" AI exclusion, covering D&O, E&O, and fiduciary liability lines broadly. Since January 2026, Verisk/ISO endorsements have allowed carriers to strip generative AI liability from commercial general liability policies.
The UK market is moving more cautiously. Marsh's July 2025 property and construction PI market update is explicit: "Insurers are paying closer attention to the technology processes within firms, especially concerning the use of AI and digital tools. This scrutiny includes an interrogation of governance and supervision, which will become part of an underwriting assessment. Also, there are one or two PI insurers starting to apply conditions/exclusions around AI which we are monitoring."
That is not yet a crisis. It is a signal. The firms that get ahead of it will be better positioned when the market turns.
What Underwriters Are Asking
Legal firm Browne Jacobson published a detailed breakdown of PI insurer questions for construction professionals in February 2026. The questions are specific and operational.
For construction practices, underwriters want to know: What checks are in place to ensure AI tools recognise material compatibility issues, such as materials that are safe in isolation but not in combination? What processes exist to validate AI recommendations against building codes and safety standards? How do you verify AI-generated scheduling and cost estimates against real-world constraints and historical project data?
These are not abstract questions about AI policy documents. They are asking whether you have robust human oversight processes. The insurer's concern is the potential claim: an AI tool neglects accessibility requirements in a design. An AI tool misses a material interaction. A professional relies on AI output without adequate verification. Each is a plausible PI scenario.
The PI and AI Picture
72% : of PI market professionals believe AI will increase risk of professional practice (Clyde and Co)
90% : expect the number of PI claims to increase over the next two years as AI use grows (Clyde and Co)
£25m : Lloyd's-backed limit offered by Armilla's purpose-built AI liability policy (2026)
Lloyd's Is Moving Too
The Lloyd's market is approaching this from both sides. On one hand, it is tightening scrutiny of AI governance in the firms it insures. On the other, it is beginning to underwrite AI risk directly.
Lloyd's insurers introduced a new AI error coverage product in 2025 through Armilla, a Y Combinator-backed startup. The policy covers legal claims when customers or third parties are harmed by underperforming AI systems. Coverage is contingent on AI performance declining significantly below initial expectations rather than any single error. By 2026, Armilla had expanded limits to £25 million in Lloyd's-backed cover. Munich Re has similarly introduced dedicated AI insurance products.
The emergence of purpose-built AI liability policies is significant. It signals that the insurance market now treats AI failure as a distinct, insurable risk category rather than something folded into existing professional liability.
"The insurer does not want your AI policy. They want to know what happens when the AI gets it wrong."
What Construction Firms Need to Document
The practical steps for maintaining cover, and securing the best terms when the market tightens, come down to documentation.
First, maintain a written register of AI tools used in professional service delivery, with a risk assessment for each. This aligns with both the RICS AI standard and what underwriters are beginning to ask for. Second, document your verification processes. When an AI output is used in a professional deliverable, record who checked it, how, and what professional judgement was applied. Third, review your AI tool supplier contracts. Most AI software providers significantly limit their liability. Standard PI wordings may exclude situations where recovery rights have been limited by contract. This is a gap your broker needs to know about.
The firms best positioned for the AI-era insurance market are not the ones that avoid compliance challenges in AI. They are the ones that can demonstrate they use it responsibly, with clear governance, documented oversight, and professional accountability at every step.
Governance Documentation That Protects Your Cover
Responsible with AI helps construction and surveying firms build the documentation trail that underwriters are beginning to require. Explore our practical governance tools at responsiblewithai.com.
