← THE PORTAL
FINANCE OPERATING INFRASTRUCTURE · FINANCE SYSTEMS / DATA / METRIC GOVERNANCE

From metric debates to one governed data layer.

Today, the same number has three definitions and data is fixed report by report. In Future Finance, data products are certified, metric definitions are governed, every number carries lineage, and agent boundaries are enforced, monitored, and revocable {D} the layer that makes all of it defensible.
LEARNING RETURN 01 SIGNAL conflict detected · definitions diverge 02 EXPOSURE trust at risk · data quality sized 03 CHOICE certify {D} remediate 04 OWNER ACTION data / metric owner 05 OUTCOME is there one version? EVIDENCE GATE 06 · VALUE PROOF certified definition in force The governed loop, cast for this function – the same circuit as every Finance function, carrying metric conflict risk instead of a report.
01 · THE TRANSFORMATION

From metric debates to one governed data layer.

TODAY
Metric debates and data firefighting.
Three definitions of the same number.
Data fixed report by report.
Agent access ungoverned.
IN FUTURE FINANCE
Certified data products {D} governed metric definitions
Versioned calculations {D} Finance math in registered services
Lineage on every number
Agent boundaries enforced {D} monitored, logged, revocable
METRIC CONFLICT → ACTION ROUTE
02 · WHERE TO START – THE WORKFLOWS, RANKED

Four workflows. One operating pattern.

Every workflow in this function becomes the same governed loop – cast differently. Below, each one in full: what it becomes, who does what, what it needs, and where the human boundary sits.
RANKED BY · ownership (who holds the lever) · value (from this domain’s sizing) · autonomy ceiling (Tier 1 = human-only → Tier 4 = highest permitted autonomy) · control sensitivity · scope (core vs conditional)   weighting leans value + ownership
AGENTS prepare DETERMINISTIC SYSTEMS calculate HUMANS approve OWNERS execute FINANCE validates
1Tier 4 of 4
Finance data product management
Report-by-report data fixes become a governed data product lifecycle {D} products certified by allowed use, quality tested, and lineage carried on every field.
Who does what
Agentssurface data quality issues with lineage
Systemstest quality against rules continuously
Humanscertify the data products
ARCHETYPE policy-to-complianceCADENCE ongoingDATA lineage · quality rules · ownershipSENSITIVITY high
THE BOUNDARY · data products are human-certified; agents surface issues, systems test, certification is not automated.
OWNERSHIP Finance ownsVALUE High – enables the restAUTONOMY Tier 4 of 4CONTROL High sensitivitySCOPE Core
2Tier 2 of 4
Metric and semantic governance
Metric debates become governed definitions and calculation logic {D} conflicts identified, definitions enforced, and every official metric human-approved.
Who does what
Agentsidentify metric conflicts across sources
Systemsenforce the certified definitions
Humansapprove the definition and logic
ARCHETYPE policy-to-complianceCADENCE change-drivenDATA semantic layer · metric catalogSENSITIVITY high
THE BOUNDARY · definitions are human-approved; the semantic layer enforces, agents never create official metrics.
OWNERSHIP Finance ownsVALUE High – one version of truthAUTONOMY Tier 2 of 4CONTROL High sensitivitySCOPE Core
3Tier 2 of 4
Finance context management
Tribal knowledge becomes a governed context repository {D} policies, owners, contracts, and calendars curated and retrievable, with humans approving what enters.
Who does what
Agentsretrieve governed context on demand
Humanscurate and approve the context store
ARCHETYPE policy-to-complianceCADENCE ongoingDATA policies · owners · contractsSENSITIVITY medium
THE BOUNDARY · context is human-curated; agents retrieve, they do not author authoritative context.
OWNERSHIP Finance stewardVALUE Agent readinessAUTONOMY Tier 2 of 4CONTROL Medium sensitivitySCOPE Core
4Tier varies
Agent workflow stewardship
Unmanaged automations become a governed agent workflow lifecycle {D} agents generating logs, humans reviewing behavior, and autonomy assigned, monitored, and revocable.
ranked last: the governance that makes every other loop safe.
Who does what
Agentsgenerate logs and behavior telemetry
Systemsenforce autonomy tiers and monitoring
Humansreview agent behavior and set autonomy
ARCHETYPE policy-to-complianceCADENCE ongoingDATA agent logs · registriesSENSITIVITY high
THE BOUNDARY · autonomy is human-assigned and revocable; this is the stewardship that keeps every other loop governed.
OWNERSHIP Finance stewardVALUE Trust enablementAUTONOMY Tier assignedCONTROL High sensitivitySCOPE Core
03 · THE SIZING – FULL EVIDENCE TRAIL

The number, carried the way every claim is carried.

The figure on the front page arrives here as what it is – a governed packet. Range, basis, inputs, benchmarks, derivation, assumptions, and the strongest objection to it, all in one place.
FINANCE SYSTEMS SIZING PACKET
DATA-SZ-13 · DATA-QUALITY POOL
OUTSIDE-IN
RANGE
$3–12M / yr · direct · plus enabling (non-additive)
BASIS
Hybrid · Confidence: Low
WHAT IT IS
Two components. (a) Direct {D} data-quality and reconciliation effort, sized from a benchmarked data-management figure. (b) Enabling {D} a stated fraction of other zones’ value that is unrealizable without governed data; flagged NON-ADDITIVE and excluded from the total. The direct anchor is a self-reported survey estimate, hence Low confidence.
INPUTS
A benchmarked cost of poor data quality / data-management effort for a representative large-cap SaaS company; the enabling share applied to other zones’ realizable value
BENCHMARKS
Cost of poor data quality (Gartner; self-report caveat, haircut) · data-management effort benchmarks (APQC)
DERIVATION
a · finance-relevant data-quality/reconciliation effort ~$3M–12M/yr (heavily haircut)
b · 10–25% of other zones’ realizable value is contingent on governed data – a dependency overlay, EXCLUDED from the total
ASSUMPTIONS
(a) finance-data slice $3–12M (haircut, self-report caveat stated) · (b) enabling share 10–25% (pure assumption, overlapping, NON-ADDITIVE)
SENSITIVITY
For (a), the finance-data-slice assumption moves it most – halving gives ~$1.5M–6M/yrsanity bound: (a) cannot exceed the benchmarked cost of the finance-data function; (b) cannot be incremental to the pools it enables
THE ATTACK
“The direct figure is a vendor survey and the enabling piece double-counts.” — Both fair: (a) is heavily haircut with the self-report caveat stated, and (b) is flagged NON-ADDITIVE and held out of the total. The honest role of this layer is enablement, sized small and labeled.
OUTSIDE-IN · ILLUSTRATIVE · SUBJECT TO VALIDATION
Modeled on a Representative SaaS Company · outside-in, illustrative
A target-state vision · every value claim subject to validation