PipeLedger AI

PUBLICATION CONTROL

Some agents are insiders. Most aren't.

Every dataset moves from extraction to release through review checkpoints and a governed publish step you control.

The problem

For listed and pre-IPO companies, an AI that surfaces an unreleased number isn't a convenience — it's a disclosure event. Publication control isn't optional; it's the legal boundary between governed and ungoverned AI.

How it works

Held for insiders, released to everyone on schedule

Q2 2026 close · agent clearance map
Q2 held · MNPI lockdown · release_at Jul 30, 2026, 11:59 PM ET
Zone 1
Cleared insiders

Full current data, including the held quarter.

CFO Agent Controller Agent Close Agent Consolidation Agent FP&A Agent
Sees Q2 the moment it's approved
Zone 2
Operationally sliced

Same released data as everyone — but the held quarter is sliced to their own domain.

Sales Agent
Released · through Q1Full access

The same published dataset as everyone — full company-wide history, up to the last release.

Held · Q2 (MNPI)Operational slice
can see
opportunitiescontractsbookings
can't see
company profitearnings forecast
Zone 3
Everyone else

Public datasets only — the last released quarter.

HR agents Ops agents
reads{org}_publishedQ1 2026
CFO
CFO holds the gate
On release day, lifts the wall for everyone.
01
Two published datasets

A public dataset ({org}_published) everyone reads, and an insider dataset ({org}_insider_published) that holds approved figures before their scheduled release. A mart approved during a locked period publishes to the insider dataset with a frozen release_at date.

02
Every query routed by three checks

Is there an active, not-yet-released insider publication for this mart? Is the person or agent asking cleared for insider data? Has the clock passed release_at? Only then does the figure resolve.

03
Freeze and revoke

Any snapshot can be frozen in place or revoked with immediate effect. All active sessions lose access at the moment of revocation.

Release calendar
Scheduled per fiscal quarter — named reviewer recorded in the audit log
Insider dataset
{org}_insider_published — held until release_at
Immutable log
Every approval records reviewer identity, action, and UTC timestamp — cannot be altered