Chapter 41: The Routing Matrix
The company had not hidden the trap; it had labeled each wire with a polite noun.
On Friday, April 18, Samuel placed a redacted spreadsheet on the KRR data-room screen after screening counsel cleared it for limited merits review.
The file came from the WMG human-resources shared drive, not Northline's sealed export or its unopened directory.
Its evidence label was D06, and the native original remained with the HR custodian.
I could annotate the review copy but could not download, forward, or rename it.
Across the top, columns divided complaints into operational concern, reputation concern, and external sensitivity.
None said punish the employee, damage a reference, silence a critic, or remove anyone from employment.
That was the first defense the document offered anyone who wanted innocence by vocabulary.
Samuel asked me to read the fields as a communications professional, not a lawyer.
I started with intake source, complaint status, audience risk, and response owner.
Each sounded like ordinary work until the status menu reached one word: externalizable.
The adjacent rules displayed two automatic review destinations when that value was selected.
One destination was reference review.
The other was communications review.
Neither destination proved that a reference had been harmed or a public statement issued.
They showed that a complaint classification could place both processes into motion.
Samuel made me say the limit aloud before he let me continue.
"A trigger is not an outcome," I said.
"And a label is not intent," he replied.
We opened D05 beside the matrix, keeping both documents inside KRR's controlled workspace.
The protocol described escalation tiers, response owners, and review queues in similarly neutral language.
I mapped each D06 field to its closest D05 stage on a blank crosswalk.
Intake source connected to case opening.
Externalizable connected to the protocol's combined reference and communications checkpoint.
Executive sensitivity connected to a strategy escalation field without identifying any decision maker.
The crosswalk showed system design, not whether any person had used it as designed.
Samuel then loaded the nine protected record IDs without names or underlying case narratives.
I could see dates, status values, and project categories only.
Four IDs contained the externalizable value somewhere in their redacted routing history.
Three later displayed a reference-review event, and two displayed a communications-review event.
Those counts were sampling facts, not a companywide rate or proof of causation against anyone.
I marked every blank where an identity, outcome, or unverified inference would otherwise appear.
The ID overlapping Northline's project date carried no visible employee name or draft text.
I did not ask Samuel to reveal either.
Instead, I completed the field map and signed it as an analytical aid.
The signature did not turn my interpretation into KRR's conclusion, and Samuel countersigned that boundary.
Samuel assigned systems validation to his own technical team and locked my annotations separately.
He also requested the deployment log identified as L05 from the HRIS custodian.
Until that log arrived, D06 could not prove when this version entered production.
It also could not prove who approved it, who used it, or why.
No protected witness had explained the actual reference process yet.
The future witness line remained W02, unopened and unnamed to me.
I looked again at the phrase reputation concern and recognized my profession's favorite hiding place.
Polite language had once made incomplete facts feel disciplined rather than dangerous.
Now the language gave KRR a testable pathway instead of a verdict.
At 3:26, the custodian generated a review-session receipt and closed my access.
Samuel preserved the crosswalk, D06 hash, redaction map, and request for L05 together.
Before the screen went dark, he opened the version-properties panel for one final field.
The author line did not identify human resources.
It identified Andrew's strategy office.