Anas HamadAtmet Brain6 min readHow Liva Insurance Used Atmet to Build a Company Brain Before Building Agents
Knowledge Stayed Inside the Company
When we started working with Liva Insurance, customer data, insurance policies, claims, correspondence, and internal files were spread across multiple places. Any Agent working inside the company needed the full picture before answering or taking action.
Instead of building a separate Agent for every task, we started by building a unified company Knowledge Base inside Atmet, then built Agents and Automations on top of it.
Atmet can run locally inside the company’s own infrastructure, keeping company knowledge and data within its environment while giving models and agents the understanding they need to work.
The Knowledge Base Was More Than Files
Policy and procedure documents matter in insurance, but they are not enough. Asking what documents a claim requires is different from asking why a particular customer’s claim is delayed.
The second question may require customer data, the policy, claim status, received documents, the latest customer communication, and actions taken by the team. Atmet connects these sources in one Knowledge Base that understands their relationships.
A customer becomes an entity connected to a policy, vehicle, claim, documents, correspondence, and actions, so every Agent can start from an existing understanding of the company.
The Team Could Build Agents
Once the company’s knowledge lived inside Atmet, the team no longer needed to build an Agent from scratch for every new use case.
They could write what they wanted an Agent to do, define its role and available tools, and let it work over the same Knowledge Base.
The team could build Agents to review long-open claims, analyze delays, prepare management summaries, or draft customer emails based on each claim’s status.
From Understanding to Action
A claims Agent can understand whether a claim is waiting for a customer document, an internal action, or team review.
For claims waiting on documents, it can prepare an email for each customer using the actual claim status, missing document, and claim number. The company can limit it to drafts or require employee approval before sending.
Another Agent can use the same knowledge for analysis, such as comparing processing times, finding recurring patterns, and preparing management reports. The knowledge is shared; roles, tools, and permissions differ.
What We Wanted to Build with Liva
We did not want Atmet to be another chatbot inside an insurance company or an automation tool that needed new setup for every task.
We built a brain that understands the company and connects its knowledge and sources in one Knowledge Base that can run locally. The Liva team can then build Agents simply by describing what they want them to do.
One brain inside the company, with every Agent working on top of it.