AI Risk Assessment Workflows
Risk assessment is what happens after triage and before pricing. It is the underwriter's structured analysis of what could go wrong with a risk and how to price or modify the coverage to handle that exposure. AI accelerates risk assessment by absorbing background research, structuring analysis, and producing the documentation that supports the underwriter's judgment.
What You'll Learn
- How to build an AI-assisted risk assessment workflow for new submissions
- Industry-specific exposure prompts for common commercial classes
- How to use AI for COPE (Construction, Occupancy, Protection, Exposure) analysis on property risks
- Risk score prompts and the right way to interpret them
Where AI Fits in Risk Assessment
A typical risk assessment workflow has these steps:
- Read submission and supplements
- Research the prospect's industry and operations
- Identify exposures by line of coverage
- Identify mitigating factors
- Assess loss history relative to industry
- Recommend coverages, limits, and any modifications
- Document the analysis
AI helps most with steps 2, 3, 5, and 7. Steps 4 and 6 are still where the underwriter's experience matters.
Industry Exposure Prompts
A useful pattern is to keep a library of industry-specific prompts and load the relevant one when a new submission comes in.
You are an experienced commercial general liability
underwriter. The applicant is a [INDUSTRY] in [STATE].
Produce a structured exposure analysis with:
- Top 5 GL exposures specific to this industry
- Top 3 property exposures
- Typical professional or E&O exposures (if any)
- 3 industry-specific underwriting questions
- 3 typical loss control recommendations
- 3 industry certifications or standards that, if present,
are positive signals
Use general industry knowledge. Do not invent specific
statistics. Where you cite a number, label it "approximate".
Example industries to load: HVAC contractors, restaurants, manufacturers, technology consulting firms, trucking, daycares, fitness studios.
COPE Analysis for Property
For property risks, COPE (Construction, Occupancy, Protection, Exposure) is the standard framework. AI can produce a COPE worksheet from a property submission.
You are a commercial property underwriter. Below is a
property submission for a [BUILDING TYPE] in [STATE].
Produce a COPE analysis:
CONSTRUCTION:
- ISO construction class (best guess from description)
- Year built and rehab history (if known)
- Roof type and age
- Sprinklered? (Yes / No / Partial)
- Major positives and negatives
OCCUPANCY:
- Primary occupancy and any secondary
- Hazards in occupancy
- Hours of operation
PROTECTION:
- Public protection class (if known)
- Distance to fire station and hydrant
- Sprinkler / alarm / monitoring details
- Ratings (UL, FM)
EXPOSURE:
- Adjacent properties and their occupancies
- Wind, hail, flood, earthquake exposure (general for the
ZIP/region)
- Crime rate notes if known
End with 3 underwriting concerns and 3 follow-up questions
to the broker.
Submission: [paste]
Loss History Benchmarking
A 5-year loss run shows a pattern. Whether that pattern is normal or alarming depends on the industry.
You are a commercial casualty underwriter. The applicant is
a [INDUSTRY] with [REVENUE] annual revenue. The 5-year
loss summary is below.
Produce a benchmarking analysis:
- Frequency assessment (above / at / below typical for
this industry)
- Severity assessment (above / at / below typical for
this industry)
- 3 loss patterns worth flagging
- 3 follow-up questions about loss control
- A general qualitative score (Better / Average / Worse
than typical)
Use general industry knowledge. Where you cite a frequency
or severity benchmark, label it "approximate" and do not
invent specific carrier loss ratios.
Loss summary: [paste]
Risk Scoring Prompts
A risk score is a structured way to compare submissions. AI can apply a defined rubric consistently across many submissions.
You are a commercial lines underwriter. Score the
following submission against this rubric:
Each factor is scored 1 (poor) to 5 (excellent):
- Years in business and stability
- Industry inherent risk
- Loss frequency vs. industry norm
- Loss severity vs. industry norm
- Loss control program
- Financial strength
- Quality of submission information
Output:
- Score for each factor with a 1-sentence justification
- Total score out of 35
- Top 3 strengths
- Top 3 concerns
- Recommendation (Pursue / Conditional / Decline)
Submission and loss data: [paste]
This produces a consistent, defensible score across many submissions, which is useful for comparing accounts in a portfolio.
Combining Risk Assessment Outputs
For a complete account, you can stack the prompts:
- Run industry exposure prompt to load the standard exposure list
- Run COPE prompt for property risks
- Run loss benchmarking prompt against the 5-year loss run
- Run risk scoring prompt to produce a portfolio-comparable score
- Combine into a one-page risk assessment summary
A combination prompt:
You are a senior commercial underwriter producing a one-page
risk assessment summary. Combine the four analyses below
into a structured one-page summary with:
- Account snapshot (3 sentences)
- Top 5 exposures
- COPE highlights (only if property)
- Loss history takeaway
- Overall risk score and recommendation
- Top 3 actions before binding
Industry exposure analysis: [paste]
COPE analysis: [paste]
Loss benchmarking: [paste]
Risk score: [paste]
What AI Cannot Do
- Replace site visits. For larger or more complex risks, an actual loss control inspection is irreplaceable.
- Verify financial statements. AI can summarize them, but verification of audit quality, going-concern, and material misstatements is a CPA-level activity.
- Predict the next loss. AI is pattern-based, not predictive in any actuarially valid sense for individual risks.
- Make the bind decision. That is the underwriter's call, supported by the AI analysis.
Documentation That Holds Up
Good risk assessment documentation lets the next underwriter (or the regulator) understand why you did what you did. A few discipline rules:
- Save the prompts, the inputs, and the outputs together with the file
- Mark which AI tool and version was used
- Have the underwriter's review notes attached, including any disagreements with the AI output
- Date and sign the assessment
This audit trail is essential under NAIC Model Bulletin frameworks now adopted in many states.
Key Takeaways
- AI accelerates the research, exposure analysis, COPE, loss benchmarking, and documentation steps of risk assessment.
- Industry-specific exposure prompts and a portfolio-consistent risk scoring rubric are the highest-leverage AI patterns for underwriters.
- AI does not replace site inspections, financial verification, or the underwriter's bind decision.
- Save prompts, inputs, outputs, and underwriter review notes together for audit-grade documentation under emerging AI governance rules.