Evaluates new insurance claims for fraud indicators using rule-based checks and AI consistency analysis.

How it works 18 steps

Inspect the data fetches, transforms, gates, and output this recipe runs.

Process (18 steps)

■ late_days date_diff

Calculate days between claim date of loss and claim date reported

■ late_flag compare

Check if late days is greater than setting late reporting threshold days

■ freq_flag compare

Check if policyholder prior claims count is at least setting high frequency threshold

■ threshold_amt math

Calculate setting avg claim for type times 2

■ amount_flag compare

Check if claim amount is greater than threshold amt

■ hashed_claim hash_replace

Anonymize sensitive fields for safe AI processing

✦ desc_check ai-classify

AI classifies data into one of the given categories

■ restored_ai hash_restore

Restore anonymized fields back to real values

■ desc_flag compare

Check if desc check category does not equal consistent

■ any_flag any

Check if any of the conditions are true

■ count_2 math

Calculate late flag plus freq flag

■ count_3 math

Calculate count 2 plus amount flag

■ flag_count math

Calculate count 3 plus desc flag

■ risk_level switch

Map flag count to one of: 0, 1, 2, 3, 4

■ ai_ready all

Check if all conditions are true

✦ ai_explanation ai-prompt

AI generates a response from a custom prompt

skip: step.ai_ready equal false

■ checklist to_checklist

Format results as an actionable checklist

■ summary to_summary

Format results as a summary card

▼

Output

summary summary

checklist checklist

ai_analysis ai_explanation

Settings 4 configurable

Configurable at install. Defaults shown — change them anytime in Recued.

enable ai toggle = on

avg claim for type number = 10000

high frequency threshold number = 3

late reporting threshold days number = 14