What was supposed to happen?
The intended sequence, written down before the fact.
Design Pattern · September 2026
Operational truth
A workflow that runs is not the same as work that happened. Zero Neo systems have to reconcile what the records say with what the evidence shows.
Under the pillar Zero Neo
Records are claims
A database row saying follow-up complete does not prove a meaningful follow-up occurred. A booking record does not prove the customer appeared. An assigned task does not prove the work was performed correctly, or at all.
Every operational system carries this gap. Automating a workflow does not close it; it usually widens it, because the record is now produced by the same mechanism that was supposed to do the work.
What has to be answerable
For every workflow the organization cares about, these questions should have answers without anyone having to investigate.
What actually happened?
The observed sequence, including the steps that were skipped.
Who or what performed it?
A person, a rule, or a model, named.
When?
Timing, including how long each step waited.
What evidence proves it?
An artifact that exists independently of the status field.
What exceptions occurred?
Counted and categorized, not buried in a log.
What outcome resulted?
Measured against something the business already tracks.
Continuous reconciliation
The useful role for AI here is not generating more status updates. It is comparing recorded state against observable evidence continuously and raising the discrepancies a human should look at.
This is also the honest test of whether an assembled environment is better than a purchased one. A packaged product with a clean audit trail beats an assembled one with none, whatever the licence saving.