Fiddler for Insurance

Deliver personalized customer service and protection for your customers with high performing, correct, and safe AI
Table of content

Insurance leaders are harnessing the power of AI to innovate and grow their business in evolving market dynamics, streamline customer experiences as well as providing protection to customers whenever and wherever they need it — whether it be their health, homes, cars, family, or financial future. 

Fiddler is a pioneer in AI Observability — the foundation to ensure the performance, behavior, and safety of predictive and generative AI models and applications. Insurance companies are equipped with a comprehensive AI Observability platform to productionize predictive and generative AI models responsibly and at scale. 

The Fiddler AI Observability platform aligns developers, platform engineering, data science, and business teams through the lifecycle to monitor, analyze, explain, and improve models and applications. 

A diagram of Fiddler's full stack AI Observability platform for generative AI and predictive models.

Why Insurance Leaders Choose Fiddler

  • Unified Environment for ML and LLMOps: Provides a common language, centralized controls, and actionable insights to operationalize predictive and generative AI with trust.
  • 360° View Into AI Behavior: Integrates best-in-class XAI and analytics for a complete understanding of why and how predictions are made.
  • Built for the Enterprise: Enterprise-scale security and support without the hassle of building and maintaining in-house LLM and MLOps monitoring systems.
  • Expert AI Team for Responsible AI: Fiddler’s expert AI team is dedicated to helping enterprises succeed in their ML and LLM deployments, ensuring the achievement of responsible AI.

AI Observability is a Must to Scale Safe and Trustworthy AI

For enterprises to fully leverage AI capabilities, it is crucial to standardize their ML and LLMOps using the MOOD stack — a comprehensive framework that includes Modeling, AI Observability, Orchestration, and Data layers. Each layer plays an integral role in the overall functionality, with AI Observability being the key component that orchestrates and enhances the other layers by providing governance, interpretability, and ML and LLM monitoring to improve operational performance and mitigate risks.

This holistic approach not only enhances the reliability and accuracy of AI but also ensures compliance and reduces the likelihood of issues, thereby supporting the enterprise’s overall strategic goals.

MOOD - The 4 Layers of the LLMOps Stack

Fiddler Supports Advanced and Complex AI Use Cases for Insurance

Generative AI Use Cases

  • AI Chatbots: Boost customer satisfaction with accurate and helpful LLM chatbots. Ensure the accuracy, safety, privacy, and correctness of NLP and LLM-based chatbot conversations. Monitor key metrics such as hallucination rates, toxicity levels, and user feedback to maintain highly accurate chatbot interactions.
  • Internal Copilot Applications: Enhance employee productivity and boost their confidence in decision-making. Boost your employees’ confidence and trust in information generated by LLM applications for business-critical projects. Fiddler monitors a comprehensive range of LLM metrics (hallucination, safety, PII) to ensure that these applications provide correct and accurate information for financial planning and investment analyses.
  • Risk Management: Detect adversarial attacks and data leakage. Detect jailbreaks and prompt injection attacks that risk LLM applications from exposing sensitive enterprise and customer information. Visualize patterns in a 3D UMAP and identify all prompt injection attacks using Fiddler’s Slice and Explain.
  • Cost Management: Increase LLM operational efficiency gains. Track the costs of using LLM applications by obtaining a complete view of your LLM costs, latency, and session length. Use Fiddler’s custom metrics to measure efficiency gains, helping streamline processes and enhance productivity. 
  • Content Summarization: Deliver highly accurate summaries for your users. Enhance the accuracy of LLM summarization by monitoring hallucination metrics such as faithfulness, answer/context relevance, coherence, and consistency using Fiddler Trust Models. Quickly detect signs of hallucination, diagnose and address their root causes before inaccurate LLM summarizations can impact the enterprise and its users.

Predictive Use Cases

  • Claims Process Automation: Streamline claims process to reduce manual reviews, improve efficiency, and achieve cost savings. Understand how claim models are processing claims, ensuring claims are processed accurately. Discover why certain claims are approved or denied with explanation methods, like Shapley Values and Fiddler SHAP. Drill down on local and global-level explanations to understand how each feature contributed to claim decisions. 
  • Image Assessment for Claims: Deliver accurate and transparent assessments of claims. Use image-based explainable AI (XAI) to accurately evaluate damages and calculate fair settlements for car, home, and property and casualty claims. Enhance decision-making with insights from model predictions and integrate a human-in-the-loop process to ensure optimal outcomes.
  • Fraud Detection: Safeguard the company and customers from fraudulent activities, and reduce the risk of financial loss. Improve the detection of fraudulent claims by closely monitoring subtle shifts in data caused by models processing highly imbalanced data inputs, and get real-time alerts as soon as anomalies come up.
  • Pricing Optimization for Premiums and Bundling: Increase the confidence in premiums and bundling of products and services. Accurately assess customer risk to optimize premium pricing and bundle offers. Understand how multi-dimensional variables lead to pricing shifts. Check for model and data fairness to ensure prices are competitive, fair, and transparent.
  • Personalized Recommendations: Delight customers with personalized recommendations. Use Fiddler to uncover why segments of customers are offered specific products and services, and which factors contribute to those recommendations. Understand which attributes from segments of customers contribute to product recommendations. 

Key Capabilities


Monitor predictive and generative AI models across pre-production and post-production. Track ML and LLM metrics at scale in a unified dashboard. Use the Fiddler Trust Service to quickly, safely, and cost-effectively measure LLM metrics with accuracy.

From model monitoring alerts to detailed root cause analysis, quickly identify and address model issues to minimize business impact.

  • Metrics Monitoring: Accurately detect changes in model behavior by monitoring ML and LLM metrics
  • Segment Monitoring: Monitor segments to drill down on underperforming cohorts
  • 3D UMAP Visualizer: Gain contextual insights by identifying patterns and outliers in high dimensional spaces
  • Alerts: Configure and receive real-time model monitoring alerts to identify and troubleshoot high-priority issues
A Fiddler monitoring chart showing two metrics, "query_cf" and "response_cf," over a 30-day period. The chart features blue and orange lines representing the metrics with fluctuations from March 21 to April 17. The date range, time zone, and comparison options are displayed above the chart.

Explainable AI (XAI)

Fiddler provides best in class explainable AI technology to provide complete context and visibility into tabular and image model behaviors and predictions, from training to production. Implement powerful XAI techniques at scale to build trusted AI solutions.

  • Explainability methods for your use case: Increase your model’s transparency and interpretability using SHAP values, Fiddler SHAP, Integrated Gradients, or custom explanations
  • Image Explainability: Explain predictions in image models to understand their ensure they are high-performing and accurate
  • ‘What-If’ Analysis: Gain a better understanding of your model’s predictions by changing any value and studying the impact on scenario outcomes
  • Global and Local Explanations: Understand how each feature you select contributes to the model’s predictions (global) and uncover the root cause of an individual issue (local)
A UMAP chart for a boat insurance claim displaying clusters of blue and orange dots, representing different data points, with a highlighted point showing production details. The highlighted point includes an image of a ship, labeled "actual: ship," with no embeddings.


Gain actionable insights to power data-driven decisions. Analyze slices of data to surface the root cause of issues. Understand the 'why' behind all issues for quick issue resolution and model improvement.

  • Dashboards: Increase business alignment and confidence in decision-making by connecting ML and LLM metrics to business KPIs in a unified view
  • Charts: Build customizable reports with the reports with the insights you need to gain deep understanding of your models and their impact on business outcomes
  • Root Cause Analysis: Pinpoint problematic areas causing models to underperform
  • Slice and Explain: Drill down on slices of data to perform exploratory or targeted analysis
A root cause analysis chart from Fiddler showing data drift for a chatbot model on March 31, 2024. The dashboard compares baseline and production metrics using Jensen-Shannon Distance. The main feature displayed is "query_cf" with a baseline of 17.8% and production at 46.2%, indicating a 159.55% increase. A bar chart below illustrates the distribution of custom features for baseline and production data, highlighting significant feature drift. The interface includes options for custom features, filtering, and analysis.
Video transcript