The problem with a risk that has no history
Insurance has been here before. When a new kind of exposure builds up inside existing policies, nobody prices it, nobody excludes it, and the gap only shows up when claims arrive. Cyber went through this. AI is now showing the early signs of the same pattern.
The pricing question gets most of the attention. The staffing question gets less, and it comes first. A carrier, MGA or broker cannot price what its people cannot assess.
Where the exposure actually sits
Most mid-sized companies do not run one big AI project. They have dozens of small AI features that arrived through software updates:
- A recruiting platform that ranks applicants
- A customer service tool that answers questions with a language model
- A finance application that drafts or categorizes entries
- A sales tool that scores leads and writes outreach
Nobody ran a procurement review on most of these. Nobody wrote down what they do. So when a loss happens, the business often cannot say which tool was involved, who approved it, or what human check existed.
That matters well beyond insurance. If you are a controller, a plant manager or a sales leader, the same visibility gap sits in your own operation.
What this means for insurers and brokers
Coverage is still unsettled. Professional liability, general liability and media liability forms are being reworded, and a single AI-related loss can land between several policies, with each one pointing at another. Specialty products are starting to appear, but the market is early.
The practical response is to change what underwriters ask. Useful questions include:
- What does the AI do, and does it make or influence decisions about customers or employees?
- Where does a human review the output, and can the process be rolled back?
- How many vendors depend on the same underlying model?
- Does the insured keep an inventory of AI tools and require vendors to disclose new ones?
Each of those needs someone in the room who can judge the answer. That is the hiring problem.
The roles that are hard to find
Three profiles keep coming up when carriers and risk-heavy employers describe the gap.
Underwriters with technical fluency. They do not need to build models. They need to read a vendor stack, ask a sharp follow-up and know when an answer is evasive.
Actuarial and analytics people who can work without history. Scenario modelling, stress testing and judgment under uncertainty are different from fitting a curve to 20 years of claims. Candidates with both the credentials and the comfort with ambiguity are few.
AI governance and risk specialists. Someone has to own the inventory, the vendor language and the oversight documentation. In many mid-sized organizations that person does not exist yet, or the job has been added to someone's full-time role.
The same needs show up outside insurance. Finance teams want people who can audit AI-assisted reporting. Operations teams want someone accountable for automated decisions. Employers in all six areas we recruit for are starting to ask for this.
Why the talent is tight
These roles sit between disciplines. Data people rarely have underwriting or audit experience, and risk professionals rarely have hands-on AI knowledge. Immigration friction in both countries makes it harder to bring in specialists from abroad, so most hiring has to be done from the domestic pool.
That usually points to three options:
- Hire a technically strong person and teach the risk side.
- Hire a seasoned risk person and pair them with a data specialist.
- Bring in contract expertise to build the first framework, then decide what to make permanent.
The third works well when the need is real but the scope is not yet clear. A 6 to 12 month contract to build an AI inventory, draft vendor disclosure terms and set oversight standards is easier to define than a permanent role nobody has scoped.
How to avoid a bad hire
Titles are inconsistent in this area, so screen for what the person has done. Ask for a specific example of a risk they assessed without clean data, and how they decided. Ask how they would find AI tools nobody told them about. Strong candidates answer with a method, not a buzzword.
Be clear internally about reporting lines too. A governance hire who has no authority over procurement or underwriting guidelines will be frustrated within a year.
What to do next
Start with an inventory of where AI already touches your decisions, then list which skills you lack to assess it. That list becomes your hiring brief, whether the answer is a permanent hire, a contractor or a development plan for someone you already have.
A Traverse recruiter can help you define those roles and find candidates in Canada and the United States.