AI Playbook for Insurance Claim Fabrication — Auto Body Ring
A regional auto insurer's SIU received a referral from a claims adjuster: 34 claims in 8 months all involve the same body shop, all report similar damage descriptions, and 22 of the claimants share a zip code. Total paid: $612,000. The body shop was licensed 11 months ago.
When to use this playbook
- Use this playbook when the decision looks like the situation above: A regional auto insurer's SIU received a referral from a claims adjuster: 34 claims in 8 months all involve the same body shop, all report similar damage descriptions, and 22 of the claimants share a zip code.
- It is a fit when you have source files in hand and need a structured, reviewable analysis — not a generic chat answer about "Insurance Claim Fabrication — Auto Body Ring".
- Do not use it as a substitute for licensed, legal, clinical, or authorized official judgment in the domain.
What you'll need
- 34 claim files (adjuster notes, estimates, photos, payment records)
- Body shop licensure and ownership records
- Claimant address and phone roster
- Vehicle VIN and registration records
- Social network data (publicly available connections between claimants)
Attachments: Images (Images)
The Prompt
You are an insurance fraud investigator reviewing a suspected auto body fraud ring. I am attaching: Work only from the attached source files. If a conclusion is not supported, say so. Produce: 1. Map all relationship connections between the 34 claimants: shared addresses, phone numbers, social media connections, or employer. 2. Identify repair estimates that are implausible for the reported damage type—flag estimates where labor hours exceed industry standard by more than 20%. 3. Determine whether any vehicles were claimed as damaged in the same incident across multiple policies (staged accident indicators). 4. Assess whether the body shop's revenue concentration (% from this insurer) is consistent with a legitimate operation. 5. Tell me what I need to present to the DA's office and whether this meets the threshold for a criminal referral in this state. Call out where independent models are likely to disagree, and list follow-up documents a reviewer should request.
What to expect
- Claimant relationship network map
- Implausible estimate register
- Staged accident probability score
- Body shop revenue analysis
- Criminal referral checklist by state standard
Review before you act
- Validate this output against source files before relying on it: Map all relationship connections between the 34 claimants: shared addresses, phone numbers, social media connections, or employer.
- Validate this output against source files before relying on it: Identify repair estimates that are implausible for the reported damage type—flag estimates where labor hours exceed industry standard by more than 20%.
- Validate this output against source files before relying on it: Determine whether any vehicles were claimed as damaged in the same incident across multiple policies (staged accident indicators).
- Validate this output against source files before relying on it: Assess whether the body shop's revenue concentration (% from this insurer) is consistent with a legitimate operation.
- Confirm every cited figure, date, counterparty, or requirement against the attached originals — models compress and can drop a qualifier.
- Treat disagreement between models as a review item, especially on classification, materiality, and recommended next action.
- Do not authorize an operational, clinical, legal, credit, or enforcement action solely because the models agree.
Why compare models on this
For Insurance Claim Fabrication — Auto Body Ring, running the same attachments across independent models is useful because the hard part is classification and completeness, not fluency. The workflow is already designed to surface claimant relationship network map; implausible estimate register; staged accident probability score; body shop revenue analysis. Those are comparison artifacts — they only exist if more than one model runs. Typology labels (bust-out vs. first-party vs. third-party) and ring membership often diverge across models when data is incomplete. Divergence is a reason to hold and verify, not to auto-file a SAR.
See governed multi-model AI on your own prompt
Compare GPT-5, Claude, and Gemini side by side, with human review and a decision record built in.

