Footprints

Framing

The problem

Action Audit gives each user a personal undo of their own agent, but no one can answer the org-level question of which AI touched what, for whom, and when, across every user at once.

The pattern

Add an operator-facing record of every AI action — actor, tool, resource, outcome, time — that any admin can search, filter, and export across all users, separate from the per-user undo timeline.

Why chat breaks here

Chat transcripts are user-scoped and stack-shaped, so reconstructing what happened across many operators means manually pulling and stitching sessions instead of querying one record.

Risks

Volume can drown the signal, and over-collecting every user's actions can breach the very privacy boundaries the record was meant to protect — scope retention and access deliberately.

Avoid when

A single user operates the agent and a personal undo timeline already answers every question — the org-wide record is overhead no one queries.

Use when

When something goes wrong across a team's AI, one queryable record answers "who did what, to what, when" in seconds — a per-user undo timeline never can.

DOPE evaluation

Directability

Investigators filter by actor, tool, resource, and outcome to scope a review

Observability

Every AI action across the org is queryable, not just user-scoped sessions

Predictability

Action records use a stable schema so cross-time and cross-user comparisons stay meaningful

Explainability

Each entry links back to the prompt, plan, and tool calls that produced the action

In the wild

• LangSmith (LangChain) — Production observability for LLM apps. Every run shows the prompt tree, tool calls, latency, token cost — filterable across users, projects, and time. The org-wide query surface compliance teams reach for first.
• Datadog LLM Observability (Datadog) — Hooks into the same APM dashboards that ops already trusts. Per-trace prompt/response, cost, latency, user id, and PII detection. Treats LLM calls as a first-class span type, not a wrapper around a normal HTTP call.
• AWS Bedrock model invocation logs (Amazon) — Native CloudWatch + S3 logging for every Bedrock invocation: caller identity, request, response, guardrail decisions. Pairs with CloudTrail for the audit layer compliance teams already operate on.
• OpenAI Audit Logs API (OpenAI) — Org-admin endpoint for actor / action / object queries across the workspace — user invites, API key rotations, project membership, login methods. Less granular than LangSmith on the prompt level but the canonical "who did what" trail.

FAQ

When should I use the Footprints pattern?

When something goes wrong across a team's AI, one queryable record answers "who did what, to what, when" in seconds — a per-user undo timeline never can.

When should I avoid the Footprints pattern?

A single user operates the agent and a personal undo timeline already answers every question — the org-wide record is overhead no one queries.

What problem does Footprints solve?

Action Audit gives each user a personal undo of their own agent, but no one can answer the org-level question of which AI touched what, for whom, and when, across every user at once.

Why is chat the wrong fit for this?

Chat transcripts are user-scoped and stack-shaped, so reconstructing what happened across many operators means manually pulling and stitching sessions instead of querying one record.

Related patterns

• Extends: Action Audit & Undo — Audit gives the user reversible undo; Footprints is the org-level record across every user, every action.
• Often paired with: Agentic Dashboard — The control room reads from Footprints; without it, supervision has no memory beyond the live moment.
• Often paired with: Constraint Rules Engine — Rules without an audit trail are vibes. Footprints proves the constraints actually held in flight.