At ITC Vegas 2025, I had the opportunity to present alongside Mike Sze, Senior Director of Software Development at Verisk, where we spoke about the need to build a complex, enterprise-grade agentic AI strategy in less than three weeks.
The Challenge:
For most of the underwriters, their day job revolves around understanding the change in regulatory landscape and continuously updating and validating thousands of existing and new filings to stay compliant. These compliance artifacts arrive in unstructured documents often hundreds of pages long filled with references, version overlaps, and nested dependencies including updates needed on existing rules, loss costs, filings, policies etc from the circulars and memes.
Currently these updates are done manually cross-referencing documents, interpreting change in compliance decisions or wordings, and ensuring downstream dependencies on products and policies are intact. Each change triggers multiple internal review cycles with Actuarial, Regulatory, and Underwriting teams, stretching a routine filing update into a multi-week process.
Introducing Arivonix AI
Arivonix AI is an enterprise-grade Agentic AI platform that connects seamlessly to over 250 data sources with zero data movement and build data products or data insights using pre-built agents in real-time. It comes with 200+ pre-built Agentic AI pipelines, many of which are tailored specifically for insurers.
What I’ve seen make the biggest impact for our customers is how Arivonix breaks the traditional Agentic AI “black box.” in the insurance space. Too often, insurance teams build an Agentic AI solution that works for one datasource or line of business but fails when it needs validation from their business teams or take the solution for launch, thereby eroding trust across teams and potential loss of ROI. With Arivonix AI, product managers, SMEs, engineers, and DevOps all work together in one transparent workspace where every pipeline is visible, explainable, and auditable end to end. That’s exactly where our Agentic AI architecture makes a difference. Each agent is designed to handle a specific function in the data workflow, operating transparently and collaboratively within enterprise boundaries.
With Human-in-the-Loop validation and one-click deployment from development to production, we ensure that speed, accuracy, and trust co-exist, enabling our clients to scale AI securely and reliably.
How Arivonix AI Solved Verisk’s Challenge
When we started working with Verisk’s Core Lines team, the goal was simple but technically demanding to turn their vast, unstructured and siloed compliance documents into structured, connected, and actionable insights to help with the underwriting use cases.
Using Arivonix AI, we automated the entire process end-to-end. The platform ingested unstructured regulatory compliance documents, extracted key entities, standardized formats, and linked related information all within a single, visual pipeline.
The result was a dynamic knowledge graph with a semantic layer on top that unified Verisk’s proprietary forms, rules, loss costs, policies etc into one consistent relationship data model. What once required extensive manual effort became a repeatable, transparent workflow. This transformation not only improved data quality but also made the new data model instantly accessible to improve the filings process for insurers.
Arivonix AI’s Date-Centric AI Assurance
What makes Arivonix AI an enterprise-grade platform is that it has a no-compromise AI governance framework that helps our customers in highly regulated industries like insurance confidently. Every feature was designed to balance innovation with assurance.
The platform embeds AI security, zero-trust access controls, tri-secret encryption and nano-segmentation to ensure every data interaction is secure, governed and isolated.
We also built their capabilities directly into the platform with model transparency, guardrails, lineage tracking and cataloging, and Human-in-the-Loop oversight. This combination of governance, explainability, and control ensures that every AI output is accurate, compliant, and fully auditable turning automation into an enterprise-grade, trustworthy capability.
Agent Arivon: Your AI Data Engineer
At this conference, we also had the privilege to introduce a working demo of our new Agent Arivon, designed to turn business requirements into working Agentic AI pipelines and data products within minutes, so Insurers can focus on their strategy and less on building complex data pipelines and processes without the need to constantly update them with the change in events.
Agent Arivon interprets simple user requirements from tools like Jira or business briefs, understands the intent, and automatically builds the corresponding data workflow end-to-end. Agent Arivon delivers up to 90% of the pipeline instantly, allowing experts to fine-tune the rest through Human-in-the-Loop validation. The result is rapid, transparent, and reliable AI pipeline creation that bridges business context with technical execution. The Knowledge Graph outputs for a complex use case described earlier can now be done through Agent Arivon in minutes.
Visit www.arivonix.ai to know more.
