FurtherAI

Everyone in insurance is chasing the same number: hours saved by AI. Dr. Alexander Vollert spent his career running transformation at @AXA and @Allianz, most recently as Group CEO & COO of AXA. He thinks that number is the smaller prize. His challenge to AI business cases: "You can write down a lot of big numbers... We can reduce the amount of work that's needed by 50, 60, 70%...but show me that you can realize it." Time saved is easy to measure. Better risk selection, fraud detection, and pricing are harder to measure but often matter more. They just take longer to prove. Our co-founder & CEO, Aman Gour, sat down with Alex to go deeper on all of it: value capture vs. value creation, why governance should speed things up instead of slowing them down, and what "owning AI" actually looks like for an executive.

We had a busy month of product ships. Introducing The Coverage, our new monthly recap of what we’re building for insurance teams. Over the last four weeks, we launched: - The Multi-Agent System, connecting specialized agents that pass work and context to one another, and to the people who own the decisions. - Computer Use 2.0, helping agents work across portals and applications without APIs. - Platform updates with 10+ new insurance integrations, EU hosting, new frontier models, and more. Catch up on everything we shipped in the first edition of The Coverage. Link in comments.

Yesterday, we launched the Multi-Agent System in @furtheraicom. A multi-agent system is a set of agents that each own a defined scope of work and hand off to one another, and to the people who own the decisions, across a full insurance process. The last 12 months have been about AI agents automating tasks across the insurance lifecycle: Underwriting: Submission intake, triage, risk assessment, rating, quoting. Distribution: Market submission, quote compare, proposal generation. Claims: FNOL intake, fraud detection, coverage verification, claims letter generation. Each AI Agent is good at its task. But complete processes still depend on people to move work and context from one task to the next. Each handoff adds effort, causes delays, and creates another opportunity for context to get lost. That's what a multi-agent system unlocks. An intake agent can process a submission and pass the full context to an underwriting agent. The underwriting agent then can pass relevant context to the rating agent and so on. AI Agents do the work. People stay responsible for decisions, exceptions, and relationships. Request a demo today!

“I’ve just hired you an underwriting assistant.” That’s how Ed LaFramboise introduced @furtheraicom to the underwriting team at Euclid Security Programs. The description gave the team a practical picture of the workflow. The assistant gathers information from submissions, applications, loss runs, third-party sources, historical results, and underwriting guidelines. It organizes that material for review and helps underwriters see the complete account sooner. The underwriter validates the information, applies experience to the risk, and makes the decision. Ed joined Nick Colis and our CEO @amangour30 at the Program Manager Conference to discuss Euclid’s first five months deploying AI inside a live MGA operation. Watch the full conversation bellow:

AI models are improving faster than many products built on them. A new model may reason more deeply, write better code, and handle more context. Yet those capabilities have limited impact when the application still confines the model to short requests and predefined tools. We encountered this while building @furtheraicom’s first assistant. It worked well for conversations, but complex insurance work unfolds across documents, systems, calculations, and multiple rounds of review. The agent needed enough time and freedom to complete the work, inspect the result, and correct its own mistakes. That led us to rethink the architecture. Each assistant session now runs inside a secure working environment with access to files, a shell, persistent state, and governed connections to the systems where insurance work happens. The agent can navigate documents, query structured data, run code, preserve its progress, and carry a task through to completion. This creates a compounding effect. As frontier models improve at reasoning, coding, tool use, and context handling, those advances can flow directly into the product. Senior Frontend Engineer Joseph Damiba wrote about the infrastructure behind this transition, the failure modes we encountered along the way, and how we are building an agent that improves with every generation of models. Full piece linked bellow: