InsightLense · writing

What your agents did, and what they were not allowed to do

Design guides on agent observability and accountability, and field notes from Vsure — a specialty P&C carrier whose staff are AI agents, which we run to find out where our own products break. Every number below comes from its ledger.

17 posts 4 sections Concepts first, worked examples last

Features and concepts

START HERE · 7 posts

How the platform is put together, and the ideas it is built on. Applicable whatever your agents actually do.

The thesis

Agentic AI is only interesting when it can be told no

An agent that cannot be refused is not autonomous — it is unguarded. Why authority has to live in the product rather than the prompt, and what you lose when you throw refusals away.

11 of 17 runs escalated 9 min read
The model

Three records, one thread: what a model-call log cannot tell your auditor

An agent produces three different kinds of evidence, and three different people come looking for them. Most tools keep only the first.

3 record types, 1 correlation id 10 min read
Instrumentation

Point your existing OTLP exporter at us and your traces appear

Standard OpenTelemetry ingest, current GenAI semantic conventions and the legacy attributes people actually have in production — with no application change.

0 lines of application code 8 min read
Economics

Cost you can reconcile: why we derive spend from traces instead of storing it

Most cost dashboards are a second source of truth that quietly disagrees with the invoice. Here is the alternative — and the metric nobody else reports.

$ per bound policy 9 min read
Evaluation

One score primitive: humans, judges, code checks and users in the same table

Separate tables for human feedback and automated evals is the reason nobody ever compares them. One shape, four sources, attachable to anything.

4 sources, 3 shapes, 1 table 8 min read
Prompt management

Did the version you promoted actually perform better?

Every trace records the prompt version that produced it, so v1 and v2 can be compared on outcomes rather than on vibes — even across two products.

v1 vs v2 in production 8 min read
Design guide

L0 to L4: designing autonomy levels a product can actually enforce

An autonomy ladder is worthless if it lives in a system prompt. The mechanic that makes it real is a permission check that returns a downgrade.

5 levels, enforced server-side 11 min read