← Go back

Guide // MandateGuard

What is an evidence export for agent spending audits?

By the Ingenious AI team · Published 2026-11-14 · Reviewed 2026-11-14

▸ Run the free live demo See pricing

Free covers 1 agent and 1 policy with the full authorize flow, signed receipts and the approval inbox — no card required to start.

Short answer

An evidence export is a portable pack of your MandateGuard receipt history plus the chain verification verdict, produced for readers who were not there: auditors, insurers and customer vendor-risk reviewers. Its credibility comes from completeness and self-proof — every decision in the period, allowed and denied alike, in a hash-chained structure whose consistency can be recomputed — plus the policy context needed to interpret the decisions. Produce it routinely at period end, verify the chain in the same sitting, and test it against the five standard questions before filing it away.

What is an evidence pack?

An evidence pack is a portable bundle of a workspace's mandate records: the receipt history for a period plus the chain verification result, produced by MandateGuard's export on Pro and Team plans. It exists for readers who were not present when decisions were made — auditors, insurers, finance leadership, a customer's vendor-risk review — and who need the records to stand without your narration. Verification lets those readers confirm the receipts are internally consistent and unedited.

Why exports exist: readers who were not there

Internal dashboards serve people who can ask follow-up questions. Audits, insurance reviews and customer vendor assessments are conducted by people who cannot — and who are professionally obliged to discount anything that amounts to trust us. An evidence pack is designed for them: a self-contained bundle where the receipts carry their own integrity proof. The chain verification is the centrepiece, because it converts the pack from a printout (which could have been edited before printing) into a structure whose internal consistency anyone can recompute. When you design your export routine, optimise for that reader: complete periods, no cherry-picked subsets, verdict included, and task references that a stranger can follow to business objects.

What a good pack contains

At minimum: every receipt in the period — allowed, approved and denied — with vendor, amount, task reference, decision and reason; the chain verification verdict and receipt count; and enough policy context to interpret the decisions, namely the caps, allowlist and threshold in force. That last element is easy to forget and hard to reconstruct later, because policies change: a denial that looks odd under today's cap may be obvious under March's. Record policy snapshots with your period-end routine, even if only as a dated note beside the export. If an incident occurred in the period, include the freeze and unfreeze timeline; the chain shows it, and naming it pre-empts the reader's hardest question.

When to produce it

Produce packs on the calendar, not on demand: at each month or quarter end, in the same sitting as the chain verification, stored with the period's financial records. Two triggers justify an extra export: a spending incident, where the pack freezes the evidence while memories are fresh, and the start of an external review, where producing the routine pack instantly is itself a credibility signal. Organisations that assemble evidence only when asked pay twice — once in staff time reconstructing context, and once in the discount reviewers apply to freshly assembled records. The pack's value compounds when it is boringly routine.

The five questions a pack must answer

Test every pack against the questions a competent reviewer will ask. What did the agent spend, with whom, for what stated purposes? What did it try to spend and get refused, and why? Which spends did a named human approve? Can we be sure this record was not edited after the fact? And what were the rules in force when each decision was made? Receipts answer the first two; approval records answer the third; chain verification answers the fourth; policy context answers the fifth. If any answer requires your team's recollection, the pack is incomplete — fix the routine (usually task references or policy snapshots) rather than annotating around the gap.

Handling and sharing discipline

A pack is sensitive operational data: it maps your vendors, volumes and control thresholds. Share it through the reviewer's channel, not by casual email; keep a note of who received which period; and redact nothing inside the receipts themselves — selective redaction undermines the chain story, so if a task reference contains something it should not, fix the reference convention going forward instead. Internally, restrict pack storage like other financial evidence. MandateGuard's own posture supports this discipline: tokens and keys are stored hashed, receipts contain no payment credentials, and workspace data export or deletion can be requested through support.

Export in the wider control set

Evidence export is one of the paid-plan controls on Pro and Team, sitting alongside vendor trust, duplicate detection, the kill-switch and the intent service. Its role in that set is memory: the other controls act in the moment, and export preserves their actions in a form that survives staff changes, tool migrations and time. Teams sometimes defer export setup until an audit looms; the better sequence is to enable the routine when the mandate goes live, so the first external question — whenever it comes — is answered from a shelf, not a scramble.

Five actions to take now

  1. Export at period end, not when the audit lands
  2. Verify the chain in the same sitting and save the verdict with the pack
  3. Store the pack with the period's financial evidence, access-controlled
  4. Check the pack answers the five standard questions before filing it
  5. Keep task references clean — the pack inherits their quality

Worked example: one pack answers a vendor-risk review

An insurtech reviewer assessing a small agency's agent deployment asks for proof of spending controls. The agency produces its quarterly pack: 342 receipts, chain verdict verified, denial log showing a freeze during a March tool incident, and 27 approval decisions with their vendor, amount and task facts. The reviewer's follow-ups shrink to two: the approver roster and the cap-setting rationale.

Compare the previous year, when the same agency answered a similar review with screenshots of chat logs and a spreadsheet reconstructed from statements — a three-week exercise that still could not evidence refused spends. The pack took one export and one verification run, because the evidence had been created at decision time rather than assembled under questioning.

Frequently asked questions

Who typically asks for an evidence pack?

External auditors, insurers underwriting agent-related risk, enterprise customers running vendor reviews, and your own finance leadership after an incident.

Does the pack include denied attempts?

Yes — receipts include denials with reasons and attempted amounts. That completeness is what makes the pack credible as control evidence.

Can the recipient verify it independently?

The pack carries the chain verification; the chain's hash structure lets the records be recomputed rather than trusted on your word.

How often should we export?

At each period end (monthly or quarterly) and after any spending incident. Routine exports are cheap; retrospective assembly is not.

Is export available on the free plan?

Export is part of the paid depth on Pro and Team. The free plan keeps the full receipt chain and verification, so history is preserved if you upgrade.

Keep reading

Put the policy in force: spend envelopes · intent registry · approval inbox · how the proof works · all MandateGuard guides