⌂ Atlas
govrn · internal · partnership working doc

The merge map.

govrn × Aperture — a vetting surface, before anything is implemented. Enter the access key.

Invalid key
govrn / internal · ↖ walkthrough Draft for discussion · nothing implemented · June 2026
Two systems · one product · first vetting pass

govrn × Aperture

Two builds that started from the same conviction — an AI estate is only governed when you can prove it, with a signed and independent attestation — and grew toward each other from opposite ends. govrn built the standard, the assessment, and the independent attestation path. Aperture built the live command center that produces the evidence. This page maps the merge: what overlaps, what's unique to each side, and how the two halves make one product — with Grace Hill as the first proof-of-concept partner.

01 · The merge board

Where it merges, overlaps, and stays distinct.

govrnassess · attest

The standard and the judgment layer — what an auditor, board, or regulator will accept.

Assessment instrumentproductized three-lens baseline: technology · security · AI
Standards crosswalksNIST AI RMF · CSF 2.0 · ISO/IEC 42001 · EU AI Act · OWASP · ATLAS
Attestation pathaudit-ready evidence; certification via accredited third parties — never self-certified
Product-AI reviewthe AI a company ships — model disclosure, oversight basis, regulated-domain review
Engagement methodrepeatable across clients: readout, playbook, gap-check
Cost & efficiency lensassessment-grade: modeled baseline, right-sizing, recoverable spend
TOGETHERthe shared spine

Both sides built these independently — convergence is the proof the thesis is right.

AI inventory / estate visibilitygovrn: AI-BOM & baseline · Aperture: live discovery — one inventory, two refresh rates
Spend & ROIone lens, two modes: modeled at assessment, live once the gateway connects
Policy-as-codewhich models touch which data, autonomy limits, PII — same rules, written once
Shadow-AI discovery & governed onboardingfind it, then sanction / govern / replace / block
Audit trailtamper-evident record of AI actions — the raw material of attestation
Framework alignmentboth map to NIST AI RMF and the EU AI Act by design
Privacy & honesty principlesmetadata not content · enablement not surveillance · nothing claimed live until connected
Aperturesee · operate

The command center — the living system the governed estate actually runs on.

Mission Controllive map of every agent, session, and workflow; real-time cost ticker
Agent registry & autonomyevery agent, its owner, its autonomy level; approval workbench; gated kill switch
Connector & MCP inventorywhat every connector touches, who owns it, whether it's healthy
Engineering productivitycycle time, churn, AI code share — adoption with proof, not vanity output
Intelligence layerAsk & Act command bar · anomaly detection · auto board-readouts · eval & quality
Capability catalogan app store of sanctioned agents, prompts, and workflows — adoption done safely

left: ink — the standard  ·  center: stitched — built twice, merged once  ·  right: teal — the command center

02 · One engagement arc

The merge in motion: assess → design → implement → operate.

Laid on the engagement arc both sides already use, the merge is almost embarrassingly clean: govrn is the front half, Aperture is the back half, and the build phase is where the two teams work as one.

Phase 1

Assess

The baseline: three-lens posture, AI-BOM, product-AI review, cost read, gap report against the standards.

govrn leads
Phase 2

Design

Operating model, decision rights, policy set, autonomy tiers — and the evidence interface spec.

govrn leads · joint
Phase 3

Implement

The command center stands up: gateway + spend, policy-as-code, audit layer, control plane — built on the vetted toolchain.

co-build
Phase 4

Operate

Aperture runs the estate: Mission Control, anomalies, readouts, evals — continuously emitting evidence.

Aperture runs
The evidence interface Operate ──emits evidence──▶ re-Assess  ·  the run-state feeds the next baseline, and the loop closes

Drawn as a starting point — the shape we would propose, not the shape we have decided. Section 06 is where it gets decided together.

Why the interface matters — a principle, not a preference: independent assessment can't rest solely on the governed system's own reporting — that's basic separation of duties, the same reason auditors don't grade their own books. So the merge keeps two distinct roles by design: Aperture produces the evidence; the govrn assessment consumes and judges it. That separation is precisely what makes the combined product credible to a board, an auditor, or a regulator — neither half can offer that alone.

03 · Resolving the overlaps

Built twice, merged once.

The shared-spine items aren't duplication to eliminate — each was built from a different vantage point, and the merged version is better than either original:

Capabilitygovrn broughtAperture broughtMerged
Cost lensAssessment-grade model: spend baseline, right-sizing, recoverable estimateLive telemetry: gateway-level spend, budgets, alerts, forecastingOne lens, two modes — modeled at baseline, live once connected; the model becomes the forecast check
Governance & policyThe vetted toolchain: policy-as-code, machine-readable compliance evidence, LLM audit logging, telemetry standards — selected and framework-mappedThe operational design: governed-onboarding funnel, data-sensitivity classification, vendor/DPA gatingThe kit meets the funnel — our tool selections become the build's components; nothing researched twice
InventoryAI-BOM at assessment depth, incl. the AI the company shipsContinuous discovery of internal tools, agents, connectorsOne registry, two refresh rates — point-in-time depth + always-on breadth
Framework mappingWorking crosswalks across six standardsNIST AI RMF / EU AI Act policy mapping in the rules engineOne crosswalk library feeding both the assessment and the policy engine
ReportingThe readout: posture, gaps, remediation path, attestation statusAuto-generated weekly board readoutsOperational cadence + independent verdict — weekly from the platform, attested at re-baseline
04 · The first proof

Grace Hill: proof of concept → product.

Grace Hill is the right first partner for the merged system, for reasons that are already on the table: generative AI embedded in compliance-sensitive workflows, a regulated domain with a real external forcing function, AI-forward leadership, and a relationship where both halves of this merge are already trusted. Everything below is subject to scoping with Grace Hill — nothing on this page commits them to anything.

What the PoC proves.
  • The arc works end-to-end — baseline → design → command-center build → operate, one client, one story.
  • The evidence interface is real — the run-state feeding a re-baseline with live data instead of public signals.
  • Both risk surfaces covered — the AI the company uses (the command center's home turf) and the AI it ships (the assessment's) — the complete answer neither half gives alone.
  • The universal model holds — new tools and departments plug in without rebuilding, the test of productizability.
What makes it a product after.
  • Repeatability — the method is already multi-client by design; the PoC hardens the command center for tenant two.
  • The full-spectrum claim — see it, measure it, govern it, prove it: no single competitor covers assessment + attestation + live operations.
  • Reference evidence — a real regulated-domain deployment, instrumented from day one.
  • Two doors into every account — the command center speaks to the CTO; the assessment speaks to risk, compliance, and the board. One product, both buyers.
05 · What each side gains

The merge is additive in both directions.

Aperture gains from govrn:
  • The standards spine — six crosswalked frameworks its policy engine can enforce against from day one
  • The vetted build kit — policy-as-code, compliance-evidence, and audit tooling already researched, compared, and selected
  • The attestation path — its evidence becomes provable governance, the thing boards and auditors actually buy
  • A multi-client method and delivery muscle — the road from internal platform to product
govrn gains from Aperture:
  • The connected run-state — the monitoring our dashboard honestly marks "not connected" becomes real
  • Live telemetry under the cost lens — modeled baselines upgraded to gateway-level truth
  • The control plane — agent autonomy, approval gates, kill switch: governance that can act, not just attest
  • A builder inside the first account — the fastest possible proof of concept
06 · To vet together — before anything is implemented

The open questions.

  • The evidence interface spec. What exactly does the assessment consume from the run-state — formats, the compliance-evidence mapping, audit-log requirements, refresh cadence? This is the load-bearing joint; spec it first.
  • One data model. Aperture's universal model and govrn's AI-BOM/posture model need to become one schema — the inventory both refresh rates write into.
  • Product-AI placement. The AI a client ships (models, disclosures, regulated-domain review) — does the command center grow a product-AI surface, or does it stay assessment-side with Aperture feeding it telemetry?
  • Naming & architecture. Is Aperture the command-center module of the combined platform, a companion product, or the run-state tier of one brand? (No right answer yet — just an explicit one.)
  • The Grace Hill PoC scope. Which phases, what's in the first ninety days, what does "proven" mean — and what the engagement looks like commercially.
  • Partnership structure. Contribution, ownership, and roles formalized in writing while everything is friendly and small — the cheapest moment to do it.
Status: nothing is implemented. This page is the vetting surface — the shared map for the conversation. When the questions above have answers, the merge map becomes the build plan.