PMOpack
Verification

Any tool can write it. This is the layer that checks it.

Any AI tool can write a governance document. PMOpack is the verification layer that checks it — one reconciled set of numbers across the period’s board artefacts. Here is exactly how it works, and exactly what it checks.

One figure. One source. The same number everywhere.

The same number in your SteerCo pack, your RAID log and your status report — or a flag telling you precisely where they diverge. No more three documents quietly disagreeing.

01LINEAGE + VERIFIED SOURCE

Every figure from an uploaded file has a paper trail — and, where the PM has marked it verified, who signed it off and when.

Trace any number back to the file you uploaded and the PM’s verification note, where one has been recorded — see exactly which document backs it and, where the PM has marked it verified, who signed off and when. Not “we used your file” — “this file is verified, by this person, on this date.”

12 SepVERIFIED
sourcephoenix_plan.xlsx · go-live
uploaded and marked verified by the PM
value12 Sep
✓ matches your plan
02THE WATERMELON TEST

We catch the report that lies by looking healthy.

The headline RAG is reconciled against the underlying risks, issues and figures. A programme green on the cover and red in the detail does not get published.

Green cover, red detail
cover RAG Green
data migration Red
→ blocked · cannot publish
03INTEGRITY RESULT + COMPLETENESS

Every critical flag is resolved or accepted on the record — with a 0–100 completeness score beside it.

A critical flag is either resolved or accepted with a written justification that is recorded, not hidden. The completeness score and the integrity check are two separate, deterministic answers — an unresolved critical always blocks publish, whatever the completeness score reads.

Completeness86/100
critical2must resolve or accept
warning1
verified to source11
04APPEND-ONLY AUDIT TRAIL

Append-only by construction. Every exception a written reason.

Append-only by construction — enforced by triggers in the committed database schema, service-role callers included. Accepting a critical flag requires a justification stored permanently with the snapshot version. Every change to the audit record writes a new row; corrections supersede, never silently edit; publishing fingerprints the pack with SHA-256 and freezes that fingerprint from then on. That digest is shown on the artefact's coherence report, so you can check it against your own copy.

15:24published · RAID Log P4 · integrity ✓
15:22exception · F3 accepted · note logged
15:21integrity run · rule engine · 2 critical
15:20generated · SteerCo P4 · v3
append-only · triggers in the committed schema
The verification path

What happens to a figure from an uploaded file

  1. 1

    Pulled from a named source

    File, field and, where the parser captured one, the row — recorded, never retyped.

  2. 2

    Checked by the engine

    Coherence checks — does it add up, does it agree with every other artefact?

  3. 3

    Recorded before it's used

    Every column mapping is shown to you for review and confirmation before anything is written. Your confirmed mappings and the transformations applied are saved with the snapshot.

  4. Printed with its source

    Printed with its source, verification status and integrity result.

Methodology

What we check — and what we don't.

WE CHECK

PMOpack runs a deterministic rule engine over the whole pack: dates reconciled across artefacts, money reconciled to the Budget Tracker, RAG statuses checked against the evidence beneath them, decisions linked to owned actions, and registers checked for quality and escalation. Fixed, inspectable, identical every run.

WE DON'T

We don't make the governance judgement for you, and we don't use an AI to decide whether a figure is right — the rules do that, deterministically. The PM stays accountable; PMOpack makes sure they're accountable for figures that actually reconcile.

Exactly which checks run on which artefact isn't a marketing claim — it's a page inside the product, rendered live from the same rule registry the engine executes. It can never disagree with the engine, and an artefact type with no registered coverage says so, plainly.

See the live rule coverage (sign-in required)
Public proof — after beta

Anonymised, integrity-checked SteerCo, RAID and Status artefacts will be published here once beta customers consent.

Beautifully formatted is easy.
Correct is the hard part.

That's the part we built.