‹ Executive recipesExecutive & Chief of Staff · Agent Blueprint

Decision audit trail

Reconstruct why a decision was made, what was rejected, and who signed it. The reasons travel with the record, not with the person.

Best forPost-mortems, onboarding, audits, re-litigating a closed debate
PrimitivesDecisionsRationaleProvenance

Why this is memory-only: the alternatives, dissent, and rationale often disappear from the final decision artifact. A persistent memory layer preserves the conversation that produced the decision.

Side by side: token usage, with and without Paperbase

Memory only

Without Paperbase

Not recoverable from search alone

  1. Find the document that recorded the decision (if one exists).
  2. Read the surrounding meeting notes for hints of dissent.
  3. Ask the people involved to recall what else was considered.
  4. Reconstruct the rationale from whatever they remember.
  5. Lose the rejected alternative and the dissent entirely when the person leaves.

With Paperbase

Persistent memory layer

  1. Pull the Decision entity with its full Rationale.
  2. Ask Paperbase for the alternatives that were discussed and rejected, with who argued for them.
  3. Pull the interactions that led up to the decision, in order.
  4. Return the decision, the people, the rejected options, and the provenance, in one view.

Agent prompt

You are reconstructing a decision audit trail.

Using Paperbase:

1. Find the Decision matching {{description_or_id}}.
2. Return its Rationale, who made it, and when.
3. Pull the alternatives that were considered and rejected, with who argued for each.
4. Pull the interactions that led up to the decision, in chronological order.

Output a single view:

- The decision and its stated rationale.
- The options rejected, with the argument for each.
- The dissent or pushback, quoted.
- Full provenance: every claim traced to its source interaction.

Placeholders in {double_braces} are inputs the agent will ask for at runtime. Give the prompt to any agent connected to Paperbase, and the rest grounds in your own memory graph.

Want a deeper recipe for your team?

Send us the prompt, the source systems, and the workflow. We will draft the recipe and sign you into the sandbox to run it.