FCA authorised · FRN 724952 0117 325 0027Goods in transit: buy onlineQuote & buy →
Apex Insurance Brokers
Speak to a brokerGet a quote →

Aipedia: Project managers

AI-generated project risk registers: complete-looking, not complete

AI can populate a risk register in minutes, but a register that lists generic risks and misses the one specific to the project gives the client false comfort. Risk registers are often the first document produced in a negligence claim to show what the PM knew and advised.

In short

Generative AI is good at producing long, tidy risk registers with owners, ratings and mitigations. The problem is that the content is drawn from typical projects, not this one. A register can be 80 lines long and still miss a ground condition, a third-party consent or a client dependency that a site visit or stakeholder conversation would have revealed. When things go wrong, the register becomes evidence of what the PM identified and how they advised. RICS-regulated PMs need a written reliability decision on material AI outputs. PI usually treats the resulting claim as negligence, subject to the policy terms.

Using AI in your work? Make sure your professional indemnity covers it. A specialist broker can check your wording and find cover that does.
Get a second opinion on your PIDoes PI cover AI mistakes?

How PMs use AI for risk

Common uses include generating a starter register from a project description, suggesting mitigations and owners, scoring probability and impact, and summarising the register for the board. Some tools pull risks from lessons-learned libraries across past projects. The output is fast and comprehensive-looking, which is exactly the problem: completeness is assumed because the list is long.

An AI tool does not visit the site, read the lease, sit in the stakeholder meeting or know the client's internal politics. The risks that cause the biggest claims are often the ones only those sources reveal: a neighbour's right of access, an occupier's operational constraint, a planning condition, a funding deadline. Generic ratings can also mislead, because a tool cannot know that a modest-sounding risk is critical on this project.

The duties that apply

Identifying and communicating risk is central to the PM role, and the standard is the reasonably competent PM. APM's ChPP standard assesses professional practice generally but sets no AI rule. For RICS-regulated PMs, the RICS AI standard requires a written decision on the reliability of material outputs (4.2), and the baseline knowledge duty (2) expects an understanding of erroneous outputs and bias. On construction projects, a PM acting for the client is helping deliver the client's duties under CDM 2015 regulation 4, and design-stage health and safety risk sits with the principal designer; see AI design risk registers for principal designers.

What can go wrong

Illustrative scenario. A PM issues an AI-generated register for an extension to an occupied building. It covers weather, supply chain and labour availability in detail but omits the need to keep a fire escape route open during works, a constraint raised by the occupier at an early meeting. The works block the route, the occupier's operations are disrupted and the programme is rewritten. The client says the PM failed to identify and advise on an obvious project-specific risk.

In the claim, the register cuts both ways. A detailed, reviewed register with evidence of project-specific input supports the PM. A generic register with no audit trail suggests the PM accepted a tool's output without thinking about the project.

Mitigations are another weak spot. AI tends to propose reasonable-sounding actions ("monitor closely", "engage early with stakeholders") with no owner, date or trigger. If the risk materialises, the client will ask what was actually done. A register whose mitigations were never real actions offers little defence, however neatly it is formatted.

How PI usually responds

A missed risk is usually framed as a failure to advise or warn, which PM PI policies are designed for, subject to the terms. Whether AI was involved rarely changes the analysis, but some wordings may ask about or address AI use, so check yours; see does PI cover AI mistakes?. If the missed risk causes injury or property damage, other policies such as public liability may also be relevant.

Practical controls

How Apex helps

Apex arranges PI for project managers and a broker can review how your policy treats AI-assisted risk advice. See also AI in change control, AI lessons-learned search and the AI in project management hub.

Sources

Frequently asked

Is an AI-generated risk register good enough?

Only once a competent person has checked it against the actual project. The register is judged on whether it identified the risks a reasonable PM would have spotted, not on its length.

Can a risk register be used against me in a claim?

Yes. It is commonly disclosed and shows what you identified, when and how you advised. A reviewed register with a clear audit trail usually helps; a generic one with no evidence of review can hurt.

Should I tell the client the register was AI-assisted?

RICS-regulated PMs must give written notice of AI use in advance. For others it depends on the appointment, but explaining how the register was built and what project-specific review was done manages expectations and helps if it is later questioned.

Who owns design risks, the PM or the principal designer?

On construction projects the principal designer plans, manages and monitors design-stage health and safety risk, while the PM manages project risk for the client. Appointments should make the split clear, especially if one firm holds both roles.

Check your PI covers how you use AI

Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start a PI quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.