Agentic Risk Architecture
Framework (ARAF)

Public Methodology Document — Version 2.1

Version 2.1 Date February 2026 Review Annual Cycle Status Public Methodology

Why Venture Bench Built This

AI agent businesses are deploying autonomous systems at a pace that contract and governance infrastructure has not kept up with. The legal frameworks used to document what agents can do, who is liable when they make errors, and how enterprise customers can audit and override them were not built for autonomous operation. They were built for software that waits for instructions.

Venture Bench was founded following extensive experience in cross-border M&A, hyperscale commercial structuring, and enterprise contracting across multiple jurisdictions. That experience revealed a consistent pattern: organisations that close enterprise deals and raise institutional capital are not necessarily those with the strongest product, but those whose commercial-legal architecture withstands scrutiny when scrutiny matters most.

For AI agent businesses, that moment arrives earlier than founders expect and carries higher stakes than standard SaaS. An enterprise procurement team reviewing an agentic vendor is not just evaluating software capability. They are evaluating whether the vendor's contracts, governance structure, and liability architecture are adequate for a system that acts without being asked. Most are not.

Venture Bench built the Agentic Risk Architecture Framework to make that gap visible, measurable, and remediable before it becomes a deal-blocker or a capital constraint.
⚠️ Important Notice — Limitations

This document describes the ARAF methodology in general terms for informational purposes. It is not a substitute for a formal ARAF assessment conducted under engagement with Venture Bench.

Nothing in this document constitutes: (a) legal, financial, tax, or investment advice; (b) a recommendation to raise or not raise capital, or to execute or not execute commercial agreements; (c) a determination of financial viability, investment suitability, or creditworthiness; (d) a guarantee of enterprise contract execution, capital raising success, or commercial outcomes.

Posture classifications, engagement pathway descriptions, collapse vector concepts, and remediation timelines are structural governance concepts presented for illustrative purposes. They describe general patterns observed across engagements and do not constitute advice regarding specific transactions or decisions.

The framework is provided without warranty as to completeness, accuracy, or fitness for any particular purpose. Application to specific circumstances requires a formal engagement and independent professional advice.


Three Audiences, Three Use Cases

AudienceHow They Use ARAF
AI Agent FoundersIdentify structural governance gaps before they become deal-blockers or capital constraints
Enterprise ProcurementEvaluate whether an agentic vendor's governance architecture meets institutional requirements
Institutional InvestorsAssess whether governance and contractual posture aligns with investment-grade requirements

The descriptions above reflect typical use cases. Public methodology documentation is informational only. Formal ARAF assessments, classifications, and remediation roadmaps are provided only under written engagement with Venture Bench.


Agentic Durability Architecture (ADA)

ADA is the governance category for AI-native enterprises deploying autonomous or semi-autonomous agents into production environments. ADA defines how agentic systems remain contractable, investable, and insurable under enterprise and institutional scrutiny.

ADA is the category. ARAF is the diagnostic engine within it.

ARAF assessments produce conformance and maturity determinations against the ADA Institutional Standard v1.0, the public reference standard defining minimum structural, governance, contractual, and capital requirements for agentic enterprises.

ARAF stands for Agentic Risk Architecture Framework. It is the structured methodology used to assess, evidence, and remediate risk posture across agentic organisations, identifying structural exposure before it manifests as a deal-blocker, capital constraint, or operational crisis.

Why ADA Exists

AI-native companies face a structural challenge. Enterprise procurement teams and institutional investors increasingly evaluate: deployment control mechanisms, liability containment architecture, model governance discipline, data provenance integrity, and contractual resilience under autonomous operation.

Most AI companies optimise for product velocity before durability architecture is engineered. ADA exists to close that gap, and to make closing it a commercial accelerant rather than a compliance burden.


Five Structural Domains

ARAF evaluates five structural domains. Each domain is assessed using documented evidence, not founder representations.

Autonomy scales capability. Autonomy also scales liability, concentration, opacity, and governance exposure.
1

Autonomy & Control

What actions can the agent take without human approval, and under what control architecture.

2

Concentration & Dependency

Revenue concentration, customer leverage, and upstream model dependency.

3

Governance & Escalation

Incident classification, escalation authority, and AIOC discipline.

4

Contract & Liability Architecture

Whether the contract stack contains, rather than exposes, the risk profile.

5

Data & Model Integrity

Data sensitivity, provenance documentation, and training data governance.

ARAF does not publish scoring thresholds or internal calibration mechanics. This preserves methodological integrity and prevents gaming.

ARAF produces: a structural risk posture assessment, a collapse vector analysis identifying the most likely legal or commercial failure pathway, a remediation roadmap with sequenced architecture priorities, and a procurement-ready governance summary for enterprise and investor use.


Risk Domains Interact Under Stress

Agentic exposure rarely emerges from one dimension in isolation. It compounds through interaction effects. ARAF maps these interactions into a collapse vector profile: the point where structural stress would most likely manifest under enterprise scrutiny or capital diligence.

Compounding PatternWhy It Escalates
High autonomy + weak contractual containmentAn autonomous action error against inadequate liability architecture creates direct, uncapped exposure.
Revenue concentration + opaque model governanceA dominant customer with leverage can extract terms that compound exposure further.
Cross-border data flows + unclear training provenanceRegulatory exposure becomes compounding and difficult to remediate post-incident.
Rapid autonomy scaling + static contract infrastructureContracts signed for an L1 product end up governing an L3 deployment. The gap emerges silently.
Collapse Vector Limitation

Collapse vector analysis identifies structural patterns that have historically correlated with commercial or legal stress under enterprise scrutiny. It does not predict specific outcomes, guarantee that identified risks will materialise, or warrant that unidentified risks do not exist.

Collapse vectors are illustrative and directional. They are not failure predictions, credit assessments, or suitability determinations. Reliance on collapse vector analysis for specific investment, capital, or commercial decisions requires independent professional advice.


Four Posture Classifications

🟢 Low Exposure

Durability architecture is proportionate to current operational scale. Posture is structurally aligned across core domains.

Engagement pathway: Proceed with enterprise contracting. Phase 1 hardening may be implemented to support scaling or procurement acceleration. It is not a prerequisite.

🟡 Managed Exposure

Material risks are present but containable with targeted remediation. Structural alignment achievable without pause.

Engagement pathway: Proceed with enterprise engagement, with a structured 30 to 90 day remediation plan implemented in parallel. No pause required.

🔴 High Exposure

Structural gaps materially constrain enterprise contracting or institutional capital readiness. Architecture misaligned to autonomy scale.

Engagement pathway: Organisations at this posture level typically benefit from completing Phase 1 contract hardening before executing major enterprise agreements. Timeline typically 4 to 8 weeks. Specific timing and sequencing should be determined with independent professional advice based on the organisation's circumstances.

Systemic Exposure

Compounding risks materially undermine contractability, insurability, or institutional capital suitability. Cross-domain structural fragility present.

Engagement pathway: Organisations at this posture level face material structural constraints that institutional investors and enterprise procurement teams are likely to identify. Addressing principal remediation architecture before major capital or commercial transactions reduces the risk of deal disruption or adverse terms. Specific decisions regarding timing should be made with independent financial and legal advice.


Non-Negotiable Structural Risks

ARAF identifies non-negotiable structural risks that materially constrain enterprise contracting or institutional capital readiness. Where such risks are present, the recommended engagement pathway escalates irrespective of strengths elsewhere in the profile.

ARAF does not publish escalation mechanics or internal thresholds. These conditions are illustrative, not exhaustive. Assessment findings are specific to the client's configuration and context.

Commitment Authority Indicators

Agent durability depends on how commitment authority is architected. The critical question is not whether an agent acts autonomously. It is whether the agent can create, modify, or discharge obligations on behalf of the business or its customers.

Commitment authority indicators are contextual. Illustrative indicators that elevate assessment posture include:

An agent that executes workflows autonomously but routes all commitment decisions to a human operates under a fundamentally different risk profile than one that can create binding obligations unilaterally. ARAF identifies which applies and calibrates the contract architecture recommendations accordingly.


Two Engagement Pathways

⚠️ Snapshot Limitation

The ARAF Snapshot is a preliminary triage instrument only. It provides directional indicators, not definitive classifications.

The Snapshot: does not assess contract infrastructure or liability architecture; does not determine whether identified collapse vectors are protected; is not suitable for enterprise conformance claims, investor diligence reliance, or formal procurement representations.

A Full ARAF Audit is required for definitive classification and remediation roadmap.

ARAF Snapshot

Purpose: Triage instrument for early-stage assessment

Designed for: Early-stage founders, pre-enterprise preparation, initial investor readiness screening

Domain coverage: Selected domains. Directional indicators only.

Classification: Directional. Not a definitive classification.

Timeline: 3 to 5 business days


AIOC: Organisational Infrastructure for Ongoing Governance

ARAF remediation often includes establishment of an Agentic Integrity Oversight Committee (AIOC). The AIOC provides the organisational infrastructure for ongoing ADA governance, ensuring that risk posture is maintained as the product and customer base evolve.

AIOC Functions

AIOC Review Trigger Conditions

Governance Exhibit Use

A one-page summary of the AIOC Operating Card is designed to be appended to enterprise MSAs as a governance exhibit or used in vendor risk questionnaires. This makes AIOC formalisation directly legible to enterprise procurement teams at the point of contract negotiation.


Enterprise Remediation (Anonymised)

The following case study is anonymised and presented for illustrative purposes only. Outcomes described reflect the specific circumstances of that engagement and are not guarantees of similar results. ARAF assessment outcomes, remediation timelines, and commercial impacts vary based on organisational context, market conditions, and execution.

Initial Posture

A mid-stage AI company operating semi-autonomous workflow agents engaged ARAF prior to enterprise expansion. ARAF indicated a systemic exposure posture across four dimensions: revenue concentration in a single enterprise customer exceeding structural resilience tolerance; autonomous execution without documented escalation pathways or override controls; liability clauses misaligned with the agent's actual authority scope; and training data provenance gaps that could not be substantiated by documentation.

Remediation Architecture

Over 90 days, the company implemented a sequenced remediation programme: contract stack redesign with liability containment aligned to the agent's autonomy level; escalation authority framework with documented override controls and AIOC formalisation; training data documentation audit and provenance gap remediation; and governance artefact preparation for enterprise procurement use.

Reported Impact

Structural posture (autonomy, concentration) is managed. Contractual posture (infrastructure, liability treatment, governance artefacts) is engineered.

AI Processing Transparency

ARAF assessments may utilise software tools, including AI-assisted technologies, to support analysis and documentation review. Such tools augment professional judgement; they do not replace it. Final classifications and recommendations reflect professional assessment by qualified personnel.

Independent Advice Required

ARAF assessments evaluate structural governance posture. They do not replace independent legal, financial, tax, or accounting advice. Organisations should engage qualified professionals for advice on: regulatory compliance obligations, securities law and capital raising requirements, tax structuring and implications, contract negotiation and execution, and insurance coverage and placement.

ARAF outputs are designed to inform professional advice, not substitute for it.


Methodology Governance

ParameterDetail
Current versionv2.1, February 2026
Review cycleAnnual minimum; interim updates for material regulatory or market developments
Calibration inputsCohort assessment data (anonymised); enterprise procurement feedback; institutional investor diligence feedback
Methodology integrityScoring logic, thresholds, and calibration maintained as controlled internal documentation

Ownership and Usage

ARAF, including the methodology, posture classifications, collapse vector concepts, and engagement pathway framework, is the intellectual property of Venture Bench Pty Ltd. Citation for internal reference purposes is permitted with appropriate attribution identifying the version number and date. Reproduction, adaptation, commercialisation, or creation of derivative methodologies is prohibited without prior written consent. All rights reserved.