Insight Lense · Agent observability & business audit

Prove what your agents did — and what they weren’t allowed to do

The moment an agent can change a record, a log of model calls stops being enough. InsightLense keeps the reasoning, the authority it acted under, and the business record that changed — as one thread, from the account back to the decision.

OpenTelemetry‑native ingest Runs in your own environment Agent, human and system actions in one ledger
The model

Three records, one thread

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

For the engineer

The run

Every step an agent took, in order — think, call a tool, read the result, think again. The sequence is the explanation.

  • Step-level tokens, latency and cost
  • Inputs and outputs per call
  • Model and prompt version in force
For the supervisor

The authority decision

What the agent asked to do, what it was permitted to do, and who it had to ask. A refusal is a result, not an error.

  • Allowed, refused, escalated or downgraded
  • Authority held vs authority required
  • Who the decision was raised to
For the auditor

The business event

What actually changed. A reserve moved from $50,000 to $250,000 on this claim — by an agent, at this time, for this reason.

  • Old value, new value, amount
  • Agent, person or system — distinguished
  • Immutable, retained, exportable
Traceability

One thread, end to end

Two identifiers travel with every record. Correlation is the business thread — minted once when an account first exists and never changed, however many systems it crosses. Causation names the immediate parent, so a ledger entry leads back to the reasoning that produced it.

Without them, each system keeps an excellent record of its own half of a conversation. With them, one query answers what happened to an account, across every service, by machines and people alike.

// everything on one account, in order
09:14:02  insightuw  submission_received  agent BR-AON-1
09:14:31  insightuw  cleared             agent UW-TBECK
09:18:44  insightuw  referred           escalated → UW-LEAD
11:02:10  insightuw  quoted             human UW-LEAD  $1.24M
14:37:55  insightcw  reserve_posted     agent CL-DOKAFOR  $250k
14:38:02  insightcede cession_created    system
// each row links to the run that caused it
The test

Five questions, one query each

The measure of an observability layer is not how much it collects. It is whether the questions people actually ask can be answered without a spreadsheet.

“Everything that happened to this account, in order.”

Auditor. One thread across every system that touched it, human and machine actions together.

“Why did the agent settle at that number — and what was it refused?”

Supervisor. The reasoning trace, plus every authority gate the run hit on the way.

“Which actions were taken by agents rather than people this quarter?”

Compliance. Actor kind is a first-class dimension, not a field buried in a payload.

“Did that event reach the next system, and how late?”

Operations. Queue depth is not the number that matters — the age of the oldest unpublished event is. A backlog of two hundred seconds old is a busy system; a backlog of one four hours old is a broken one.

“What did it cost to handle this piece of work?”

Finance. Model spend attributed to the business object, not only to the actor or the month — because an agent decides for itself how many calls a task takes.

Platform

What comes with it

Tracing, audit and the operational machinery around them — one platform rather than a tracing tool, a ledger and a cost dashboard stitched together.

Run traces

Parent and child spans, grouped into sessions, with token, latency and cost on every step.

Authority records

Escalation rate, refusals, and the level held against the level required — per actor, per capability.

Business ledger

Append-only events with old and new values, actor kind and reason, queryable by account.

Cost & budgets

Model pricing, per-actor budgets, forecast and breakdown — derived from traces, so the figures reconcile.

Alerts & incidents

Rules on failure rate, spend, latency and escalation collapse, with incident tracking and SLA targets.

Prompt version evidence

Every trace records which prompt version produced it, so you can see whether the version you promoted actually performed better. Authoring and approval live in InsightPrompts.

Scores & evaluation

One score primitive for every judgement — a reviewer, an automated check, a model judge or a user — on any run, span or business event.

OpenTelemetry native

Standard OTLP ingest at /v1/traces. Point an existing exporter at it and your traces appear — no application change.

SDK & API

Python SDK with a trace decorator, batched export and automatic instrumentation for Bedrock, OpenAI and Anthropic.

Positioning

Most tools stop at the model call

We speak the same standards as everyone else — OTLP in, scores, sessions, cost, prompt versions. The difference is what we keep after the model call returns.

A model-call log tells you

  • Which model ran, and what it cost
  • How long the call took
  • What the prompt and completion were

Useful for debugging. Insufficient the moment the software is allowed to change something.

An accountability record tells you

  • What the agent tried, and what the product refused
  • Who it escalated to, and what they decided
  • Which business record changed, and by how much
  • Whether a person or a machine was responsible

This is what a regulator, an internal auditor and a supervising manager each ask for — and none of them are asking about tokens.

Agentic AI is only trustworthy when it can be told no — and only auditable when that refusal was written down.

Enterprise deployment

Your data never leaves your environment

On-premise and private-cloud deployment. All LLM monitoring and analysis happens inside your perimeter.

In your environment

On-premise and private-cloud deployment options. All LLM monitoring and analysis happens within your network — nothing leaves your perimeter.

Compliance-ready

GDPR, HIPAA, SOC 2, and ISO 27001 — built-in audit trails, encryption, and access controls. Evidence collection automated.

Full audit trail

Every LLM call, every security event, every cost change — logged with timestamps, user, model version, and policy actions.

Ready to see what your agents are actually doing?

Tell us where your agents already act on real records — and what you would need to prove about those actions to an auditor. We’ll set up a hands-on walkthrough within 2 weeks.

Request a demo Contact sales
On-premise deployment Agent authority records Immutable business ledger Enterprise support