Signals
How Isomer detects risks across your claims book — the detector library, detection scopes, and signal output.
Isomer Signals is the risk detection product built on Isomer Core. It continuously evaluates the claim graph against a library of 85+ detectors and emits a structured signal for every match. No configuration required to start — connect an inbox and the detectors run.
How detection works
Each inbound message updates the claim graph. Signals evaluates the updated graph against every applicable detector. When a detector's criteria are met, it emits a signal.
A signal is not a raw alert — it is a structured output carrying everything needed to act:
| Field | Description |
|---|---|
| Source evidence | The triggering message, document, or page |
| Urgency | How quickly the signal requires a response |
| Confidence | How strongly the evidence supports the detection |
| Configured action | Notification, CMS writeback, or Actions workflow |
Detector library
Detectors are organized into seven categories.
Bad faith & litigation
Signals related to litigation activity, attorney involvement, bad faith exposure, and adversarial legal tactics — including venue selection, demand strategies, and reptile theory framing.
Examples: time-limited settlement demand, initial attorney representation notice, bad faith demand letter.
Fraud, abuse & evidence integrity
Signals related to fraudulent activity, organized schemes, suspicious billing or treatment patterns, evidence manipulation, and referral networks that suggest abuse of the claims process.
Examples: medical-legal referral pattern, staged loss indicators, duplicate billing.
Coverage & policy
Signals related to coverage disputes, policy interpretation, limits and reinsurance thresholds, misclassification, reservation of rights, and silent or unintended exposures.
Examples: policy limits demand approaching threshold, coverage dispute language, reservation of rights trigger.
Severity & complexity
Signals related to claim severity and handling complexity — medical severity escalation, diagnostic complexity, multi-system injuries, biopsychosocial factors, structural damage assessments, total losses, and patterns that indicate specialized handling.
Examples: catastrophic injury indicators, multi-system trauma, total loss threshold.
Compliance & regulatory
Signals related to regulatory investigations, government inquiries, sanctions screening, and statutory obligations that carry compliance risk.
Examples: OFAC match, regulatory inquiry notice, statutory deadline approaching.
Aggregation & systemic events
Signals related to multi-claimant events, common cause scenarios, emerging contaminants, and patterns indicating exposure extending beyond a single claim.
Examples: common-cause indicators, catastrophe event clustering, emerging contaminant filing.
Reputation & external scrutiny
Signals related to adverse media coverage, public attention, social media activity, and external scrutiny that could affect brand or public standing.
Examples: media mention, high-profile plaintiff attorney, social media activity on a claim.
Detection scopes
Every detector operates at one of three scopes.
Single-message. Fires on a single document when the graph has enough structure. Examples: a time-limited demand, a spoliation notice, a critical-severity FNOL.
Claim-lifecycle. Emerges only across multiple events on the same claim over time. Examples: a benign FNOL followed by escalating medical records and an attorney letter forming an indemnity-severity pattern. Detection is only possible after the claim develops — the single-message scope would not catch it.
Cross-claim. Requires pattern analysis across claims, parties, or portfolio data. Examples: organized fraud rings, medical-legal referral patterns shared across claimants. Cross-claim scope is what separates a signal platform from a document classifier.
Line of business applicability
Every detector carries an LOB applicability tag. Detectors can apply to a single LOB or multiple:
- Personal Auto
- Commercial Auto
- Homeowners / Property
- Commercial Property
- General Liability
- Workers' Compensation
- Professional Liability / E&O
- Umbrella / Excess
- Universal (all lines)
Where a signal behaves differently by LOB, the detector definition carries LOB-specific guidance. For example, attorney representation is common and expected in Workers' Compensation but is a stronger adversarial indicator in Homeowners.
Signal output and routing
When a detector fires, the signal is routed based on your configured response:
- Notification — email or Slack alert to an adjuster, supervisor, or team.
- CMS writeback — structured data written back to your claims management system.
- Actions workflow — triggers a multi-step workflow in Isomer Actions for cases requiring more than a notification.
Confidence thresholds are configurable. Low-confidence firings route to human review rather than auto-executing — no silent actions.