Home›Products›ClassifierIQ

ClassifierIQ

the assets, classified — and remembered.

A SOV might list 400 properties in 12 formats. A slip uses shorthand every Lloyd's broker understands and no generic model does. ClassifierIQ classifies the assets within — schedules, slips, SOVs, loss notes and loss narratives — and remembers every correction your underwriters and pricing teams make.

Book a Demo → All Products
CLASSIFIERIQ LIVE
Classified
4,821
↑ 12% this week
AI Accuracy
97.4%
↑ vs 94.1% last month
Avg Confidence
91%
HITL queue: 3 items
AssetClassificationConfidenceStatus
Industrial Boiler — 2MWMachinery98%Done
Chemical Storage TankHazardous Mat.95%Done
Offshore Platform CraneMarine / Eng.72%Review
The Differentiator

How the memory loop works.

No generic classifier remembers. ClassifierIQ does. Over time, your instance reflects how your syndicate, your MGA, your pricing team thinks about risk — not a market average.

01
Classify

ClassifierIQ classifies the asset from your schedule, slip, SOV, or loss narrative. Confidence score and XAI trace attached to every output.

02
Correct

Your underwriter or pricing analyst reviews. If the classification is wrong, they correct it and record the mapping. The correction is captured immediately.

03
Remember

The correction is stored in the memory loop. Next time a similar asset arrives, ClassifierIQ applies your firm's mapping — not a market average.

Over time: your model. Your logic. Your institutional memory.
Built into the platform — not locked in someone's head.
100% AI accuracy on reviewed batches
Classification in under 60 seconds
XAI reasoning on every output
Learns from every human correction
HITL queue for low-confidence items
Key Capabilities

Five features that define
the ClassifierIQ platform.

We've highlighted what matters most. To explore the full platform — agents, workflows, feedback loops, and API integrations — book a demo.

Dashboard & Analytics
Live KPIs — schedules run, assets classified, hours saved, AI accuracy and average confidence. Weekly performance chart tracks accuracy improvement over time.
Custom AI Agents
Build, publish and manage class-specific classification agents. Each agent is configured for your taxonomy, tuned on your data, and improvable through feedback.
Batch Results
Every classification batch stored with Batch ID, confidence score, time taken, and status. Drill into any result for the full XAI explanation and source reasoning.
XAI Result Detail
Every classification includes asset category, type, confidence score, and a plain-English reasoning trace — understanding, confirmation, code mapping, and confidence justification.
Coming Soon
Visual Workflow Builder
Drag-and-drop workflow canvas for multi-agent classification pipelines. Watch this space — launching shortly.
See the Full Platform
API integrations, reference data, agent feedback loops, business class configuration, and more. Book a 30-minute demo on your actual data.
Book a Demo →
Live Platform

ClassifierIQ in action.

A working platform running in production. Select a capability above to explore. Want to see it on your data? Book a demo.

app.marzallabs.com/classifieriq/dashboard
CQ
ClassifierIQ
Classification · Risk
MAIN
Dashboard
AI Agents
Agent Results
A
admin
Dashboard
Classification quality, accuracy trends and human-loop impact
Run AI Agent
Agent Results
AI AGENT PLATFORM
Schedules Run
12
All time: 12
Assets Classified
567
All time: 567
Est. Hours Saved
23.6h
All time: 23.6h
AI Accuracy
100%
from 10 reviewed
Avg Confidence
84.2%
This month: 72.4%
Weekly Performance (12 weeks)
Accuracy improvement (AI confidence vs human approval) and time saved per week
AI Confidence
Human Accuracy
Feedback Coverage
10 of 567 assets reviewed (1.8%)
Recent Schedules Last 5 batch runs across all agents
DateAgentInsuredAssetsAvg ConfStatus
2026-03-13Claims Classifier—1786.5%
Completed
2026-02-25Claims Classifier—10084%
Completed
2026-02-25Claims Classifier—1778.9%
Completed
2026-02-25Claims Classifier—1781.1%
Completed
2026-02-25Claims Classifier—10086.6%
Completed
AI Agents
Create and manage your custom AI agents
+ New Agent
Claims Classifier (Copy)
Draft
Claims
Classifies insurance claims into your target categories and codes using expert reasoning, RAG knowledge base, and human feedback learning.
AI PlatformLLM AgentAgentic
Updated 15/03/2026 by admin
Run
Results
Edit
Claims Classifier
Active
Claims
Classifies insurance claims into your target categories and codes using expert reasoning, RAG knowledge base, and human feedback learning.
AI PlatformLLM AgentAgentic
Updated 15/03/2026 by admin
Run
Results
Feedback
General Asset Classifier
Active
General
A broad-purpose classifier for general insurance assets. Uses chat completion model with standard taxonomy mapping.
AI PlatformLLM AgentChat Completion
Updated 22/02/2026 by admin
Run
Results
Edit
Results — Claims Classifier
All batch schedules for this agent
Batch Schedules 12 batches
Batch IDInsuredDateAssetsDoneFailedAvg ConfTimeStatus
44a25f74—2026-03-13 16:551717086.5%58s
Completed
b83f4db6—2026-02-25 08:08100100089%3m 58s
Completed
29c2400b—2026-02-25 08:031717087.6%58s
Completed
48382bed—2026-02-25 07:211717081.1%58s
Completed
38ee666e—2026-02-25 07:191717078.9%58s
Completed
← All Batches·Batch 44a25f74✓ Completed
Results — Claims Classifier
Batch 44a25f74 · 17 results
Export Results
InputAsset CategoryAsset TypeConfidenceTimeDateActions
CMD WASTE - YARD 6PropertyYard Property90%10.8s13/03/26
View
NEW ROAD SCALEProperty——17.8s13/03/26
View
VIADUCT AND ACCESS RAMPSProperty——14.1s13/03/26
View
ARTESIAN WELL 01PropertyYard Property88%9.0s13/03/26
View
ARTESIAN WELL 03PropertyYard Property90%9.2s13/03/26
View
Result Detail — CMD WASTE - YARD 6
Input
{"items": "CMD WASTE - YARD 6"}
asset_category
Property
asset_type
Yard Property
confidence
90%
XAI Reasoning
(1) Asset Understanding: 'CMD WASTE - YARD 6' indicates a designated outdoor area for C&D waste management.

(2) Category Confirmation: Falls under 'Property' category. Yard Property type selected as it broadly covers designated outdoor areas within a facility.

(3) Code Mapping: Code 270 — Open Storage. Suitable for outdoor waste management areas.

(4) Confidence Justification: 90% (High). Clear interpretation, strong alignment between asset function and assigned category. No conflicting human learning data.
Result ID: efd1109... · Batch: 44a25f74 · 10.8s · 13/03/2026, 16:55:14
This is a preview of key screens. The full platform includes agent feedback, reference data, API configuration, business class rules, and more.
Use Cases

Built for any classification challenge
in specialty insurance.

ClassifierIQ handles the classification workflows that consume underwriter and analyst time — at batch scale, with auditability at every step.

Claims Classification
Claims classification and triage
Classifies insurance claims against your taxonomy and maps to your target schema. Handles hundreds of claims per batch, with confidence scoring and XAI reasoning on every classification output.
Schedule Processing
Underwriting schedule automation
Processes raw schedule data — extracting, classifying, and mapping each asset to your target schema. Low-confidence items auto-routed to the HITL queue before entering any system.
Data Enrichment
Reference data mapping
Maps classified items to your construction codes, categories, or any proprietary taxonomy. Enriches with reference data at classification time — no manual lookup.
Continuous Learning
Accuracy that improves in production
Every human correction in the HITL queue is captured and fed back to the agent. No retraining cycles required — the model learns from your reviewers' judgement automatically.
By the Numbers

Classification at insurance scale.

100%
AI accuracy on human-reviewed classification batches in production
84.2%
Average confidence score across all classifications this month
XAI
Full reasoning trace on every output — asset understanding, category confirmation, code mapping
HITL
Low-confidence items auto-routed for human review before entering any downstream system
Engineering Standards

Built to the standards
regulated industries require.

Every Marzal Labs product carries the same non-negotiable engineering standards. ClassifierIQ is no exception.

Explainable AI on every classification
Every output includes a plain-English reasoning trace. Reviewers can validate the asset understanding, category confirmation, and code mapping — not just the result.
Human-in-the-Loop mandatory
Low-confidence classifications never reach downstream systems without human review. HITL is architecture, not an option — built into the classification workflow at a configurable threshold.
Continuous learning from feedback
Every correction made in the HITL queue is captured and used to improve agent accuracy. Accuracy compounds over time without manual retraining cycles.
Deployed in your environment
Your data stays in your infrastructure. We deploy into your cloud tenant — your taxonomy, your controls, your security team's oversight, your API integrations.

See ClassifierIQ on
your actual data.

30 minutes. Your real assets. We run a live classification batch and walk you through the results, confidence scores, and XAI reasoning in real time.

No setup required  ·  Your taxonomy, your infrastructure  ·  Available now