Advenno built InsureBot, an AI claims platform with conversational FNOL, automated document processing, photo-based damage estimation, and real-time tracking. Claims processing dropped from 23 to 7.4 days and customer satisfaction jumped from 3.1 to 4.5/5.
The Challenge
Guardian Insurance Group served 320,000 policyholders across personal auto, homeowners, and commercial lines, processing 42,000 claims annually. The claims experience was the single most important touchpoint in the insurance relationship — the moment when the promise of coverage becomes real — yet Guardian's process was designed around internal efficiency rather than customer experience. Filing a claim required calling the FNOL center during business hours and spending an average of 14 minutes on the phone providing details that could have been captured digitally. After filing, the policyholder entered an information black hole — they had no way to check claim status without calling the same center, and these status inquiry calls consumed 32% of the center's total capacity, creating a self-reinforcing cycle of hold times and frustration. Document collection was the biggest bottleneck: after FNOL, a letter was mailed listing required documents. Policyholders had to gather police reports, contractor estimates, medical records, and photos, then mail, fax, or email them back. An average of 3.2 back-and-forth communications were needed before a claim file was complete — each adding days to the timeline. Adjusters reviewed every document manually, estimated damages using reference guides and personal experience, and made settlement decisions without data-driven benchmarking. The entire pipeline averaged 23 days from FNOL to settlement, and during that time the policyholder was dealing with the stress of whatever event triggered the claim — an accident, a storm, a break-in — while also navigating a slow and opaque process. The 3.1/5 satisfaction score among claimants was Guardian's worst across all customer interactions, and the 18% non-renewal rate among recent claimants cost an estimated $4.6M in annual premium revenue.
- 23-day average claims processing time from FNOL to settlement
- FNOL required a 14-minute phone call during business hours with no digital alternative
- 3.2 average back-and-forth document collection communications per claim adding days to timeline
- Zero policyholder visibility into claim status — status calls consumed 32% of call center capacity
- Adjuster damage estimates built manually from scratch without data-driven benchmarking
- 18% non-renewal rate among recent claimants costing $4.6M in annual premium loss
Our Solution
Advenno built InsureBot as a platform that reimagines claims from the policyholder's perspective while giving adjusters powerful AI tools to work faster and more accurately. The conversational FNOL chatbot — accessible via Guardian's mobile app, website, and SMS — guides policyholders through claim filing 24/7 in an average of 6 minutes. The conversation adapts based on claim type and coverage, collecting required information through natural language rather than forms, with photo and document upload built into the flow. AI document processing uses OCR to extract data from uploaded documents — police report numbers, repair estimate line items, medical billing codes — and NLP to classify and validate each document against the claim type's requirements, automatically identifying what's missing and requesting it in a single follow-up rather than multiple rounds. For auto claims, a computer vision model trained on 180,000 vehicle damage images analyzes uploaded photos to generate preliminary damage estimates including parts identification, labor estimates, and total loss probability. Adjusters review and approve the AI-generated estimate rather than building one from scratch — typically adjusting 15-20% of line items rather than creating all of them. The policyholder portal provides real-time claim status with milestone tracking, adjuster contact information, uploaded document inventory, and settlement timeline estimates. Push notifications alert policyholders when their claim advances to a new stage, when action is needed, or when settlement is ready. A claims analytics dashboard gives management visibility into pipeline volume, processing times, adjuster workloads, and customer satisfaction by claim type and region.
- 24/7 conversational FNOL chatbot completing intake in 6 minutes across app, web, and SMS
- AI document processing with OCR extraction, NLP classification, and automatic gap identification
- Computer vision damage estimation from photos trained on 180,000 vehicle images
- Real-time policyholder claim tracking with milestone status and push notifications
- Adjuster workbench with AI-generated estimates for review and approval rather than creation from scratch
- Automated document gap identification requesting all missing items in a single follow-up
- Claims analytics dashboard with pipeline, processing time, workload, and satisfaction metrics
Our Approach
Claims Journey Mapping
Mapped the complete claims experience from FNOL through settlement by analyzing 500 recent claims across all three lines of business. Identified the 7 major delay points in the pipeline, quantified each in days of processing time added, and discovered that document collection accounted for 48% of total processing time.
AI Model Development
Trained three specialized AI models: the conversational FNOL agent on 12,000 historical claim transcripts, the document processing engine on 85,000 claims documents, and the auto damage estimator on 180,000 labeled vehicle damage images. Each model was validated against expert benchmarks before deployment.
Adjuster Experience Design
Designed the adjuster workbench in close collaboration with 15 senior adjusters to ensure AI augmentation felt like assistance rather than replacement. The interface presents AI-generated estimates with confidence scores and supporting evidence, making it easy for adjusters to accept, modify, or override any recommendation while maintaining professional judgment.
Policyholder Portal Development
Built the customer-facing portal and mobile app with a primary design goal of eliminating status inquiry calls. Conducted user testing with 60 recent claimants across demographics, optimizing the status timeline, notification frequency, and language clarity based on their feedback.
Controlled Deployment
Deployed to personal auto claims first (highest volume, most standardized), then homeowners, then commercial. Each line launched with a 4-week pilot comparing AI-assisted processing against traditional workflow, demonstrating improvements before full rollout. Adjusters who participated in pilots became internal advocates.
The Results
InsureBot transformed Guardian Insurance Group's claims operation from an industry laggard into a customer experience differentiator. Average claims processing time dropped from 23 days to 7.4 days — a 68% reduction — with the document collection phase (previously 48% of total time) compressed from 11 days to 2.3 days through AI-powered gap identification and digital submission. Customer satisfaction scores rose from 3.1 to 4.5 out of 5, and the non-renewal rate among recent claimants decreased from 18% to 7%, recovering an estimated $2.5M in annual premium retention. FNOL completion time dropped from 14 minutes to 6 minutes, with 73% of claims now filed through digital channels rather than phone — and 41% filed outside business hours, representing capacity that didn't exist before. Status inquiry calls decreased 74%, freeing the call center to focus on complex claims requiring human empathy and judgment. Adjuster productivity increased 156% — each adjuster now handles 2.5x more claims with AI-generated estimates reducing the time from initial review to settlement offer. The computer vision damage estimator's preliminary assessments were accepted by adjusters with minimal modification in 72% of auto claims, demonstrating strong accuracy while preserving the human oversight required in insurance. Guardian's NPS among claimants improved from -12 to +34, and the company featured InsureBot in its marketing materials as a key differentiator in policy acquisition campaigns.
Return on Investment
Technologies Used
Integrations
InsureBot changed our relationship with claimants from adversarial to supportive. When someone has just been in an accident, the last thing they need is a 14-minute phone call and a 23-day wait. Now they file in 6 minutes, track everything on their phone, and get resolved in a week. Our retention numbers speak for themselves.
Project Gallery
Lessons Learned
- Designing for the policyholder's emotional state was critical — claim filing happens during stressful moments, so the conversational UI needed to be calming and reassuring
- AI damage estimation needed adjuster oversight framed as empowerment rather than limitation — the tool helps them work faster, not replaces their judgment
- Document gap identification in a single follow-up rather than multiple rounds was the biggest processing time improvement
- Phased deployment by claim type allowed the team to calibrate AI accuracy on standardized auto claims before tackling more variable homeowner and commercial claims
Summary
Advenno built InsureBot, an AI-powered claims processing platform for Guardian Insurance Group serving 320,000 policyholders. Conversational FNOL, AI document processing, computer vision damage estimation, and real-time tracking cut processing from 23 to 7.4 days and lifted satisfaction from 3.1 to 4.5/5.
Key Takeaways
- Conversational FNOL reduced intake from 14 minutes to 6 minutes with 73% digital adoption
- AI document gap identification compressed the biggest bottleneck from 11 days to 2.3 days
- Computer vision damage estimates accepted with minimal modification in 72% of auto claims
- Status tracking eliminated 74% of inquiry calls, freeing the call center for complex claims
- Non-renewal rate among claimants dropped from 18% to 7%, recovering $2.5M in annual premiums
Frequently Asked Questions
Key Terms
- FNOL
- First Notice of Loss — the initial report filed by a policyholder when a loss event occurs, capturing essential details about the incident, damage, and parties involved.
- Computer Vision Damage Estimation
- AI technology that analyzes photographs of vehicle or property damage to identify affected components, estimate repair costs, and assess total loss probability.
- Claims Cycle Time
- The total elapsed time from First Notice of Loss to claim settlement or closure — the primary operational metric for insurance claims departments.
Facts & Statistics
Sources & Citations
- McKinsey: Claims of the Future
- J.D. Power: Property Claims Satisfaction Study
Building a mobile app or rethinking an existing one?
Platform choice and architecture decisions made in week one affect everything that follows. Share your use case and we can give you a straightforward perspective on the trade-offs.
Compare Notes With Our Team
