What does an insurance quote need to remember?

Current position — September 30, 2026

The answers entered and the price returned are a start. I am interested in a deeper question: can we still explain the quote after the product that generated it has changed?

Imagine a product definition changes between quotation and the next action. A question is revised. A rating parameter changes. An approval threshold moves. Looking at today’s configuration may tell us how a new quote should work. It may not explain the one already on the table.

The architecture I want to explore gives each transaction a durable connection to the definition under which it occurred.

That connection might be implemented with a retained definition or a reliable reference to an immutable version. The important property is that a later change cannot silently alter our understanding of an earlier transaction.

This becomes especially useful when agents operate the system. An agent needs a clear contract for what it can do now. A reviewer also needs evidence of the contract that applied then. Those are related responsibilities, but they answer different questions.

Consider a simple product change. Before approving it, I would want to ask:

• Which new transactions will use it?
• Which existing transactions remain tied to the previous definition?
• What happens if an old quote is continued after the change?
• Where will a human see and resolve any incompatibility?

These questions make the insurance database more interesting than a collection of current records. It becomes a place where the relationships between versions, decisions and outcomes remain intelligible over time.

There is a commercial question here for Facio. If generating interfaces and integrations becomes easier, can we provide a foundation that makes those changing systems dependable?

We should test that through an ordinary change, not just a clean first demonstration. Create a quote under one definition, change the product, continue the earlier transaction and inspect the evidence. Can someone understand the result without reconstructing the past from memory? Can another interface reach the same interpretation?

A software test alone would not prove that customers should buy the service. We would also need to measure how much investigation, reconciliation and operational work this foundation removes.

My working hypothesis is that an insurance core earns its place by preserving meaning as the business changes. The ability to generate the next screen is valuable. So is the ability to explain yesterday’s decision.