Skip to content
AI Governance & Fiduciary Safety

26 Agents, Human Command, Fiduciary Duty Preserved

You are handing AI agents your LP data, fund documents, and investor communications. This explains how AssetBoxx keeps humans in command through structured automation, SEC-aware compliance, and enforceable governance.

~70% Structured CodeHuman Approval on EverythingPersonize Governance12-Week Trust Escalation
Human ControlGovernance & PersonizeStructured CodeAI Judgment~30% of operationsExec. Asst.Client Exp.DeliverySecurityKnowledgeOps Intel~70% Structured~30% AI

Not Autonomous AI. Structured Agents.

AssetBoxx agents are Structured Agents: deterministic code producing identical results every time, with AI invoked only where human judgment is required.

AssetBoxx runs structured automation on dedicated infrastructure: API integrations, rule-based pipelines, scheduled jobs, and compliance checks. AI is the exception, not the rule. This distinction is critical for fiduciary duty — your firm needs predictable, auditable operations.

Deterministic by Default

LP data processing uses structured rules: fund code matching, investor classification, document routing, and compliance flagging. AI weighs in only on ambiguous cases requiring judgment.

AI for Judgment Only

AI drafts market commentary in the GP's voice, synthesizes research data, and generates quarterly narratives. Everything else is structured code with identical, auditable results.

Zero-AI Agents Exist

Several agents are entirely scripts and API calls. Data aggregation, compliance scheduling, portfolio monitoring triggers: pure automation with zero AI costs and zero hallucination risk.

Predictable, Auditable, Defensible

Compliance checks, LP data routing, filing deadlines, and alert routing: identical output given identical input. No randomness, no hallucination, no drift. Audit-ready by design.

Structured vs. AI Ratio by Agent

StructuredAI
Data & Compliance Ops
100%
Security & Infrastructure
85%
Portfolio Monitoring
80%
LP Relationship Management
75%
Fund Operations
70%
Investor Communications
70%
Compliance Review
65%
Investment Research
40%
Content & Thought Leadership
35%

Approximately 70% of all operations across the 26-agent platform are structured automation. AI is the exception, not the rule. This matters for fiduciary duty and regulatory defensibility.

The Human Approval Framework

Every LP-facing and externally visible action requires human approval before reaching anyone outside your firm.

AssetBoxx never sends, publishes, or communicates externally without a human in the loop. For asset managers, this is not optional — it is a fiduciary requirement. The architecture enforces it.

Dual-Gate Approval

LP communications pass through the Compliance Review agent (Compliance Review) for SEC Marketing Rule validation, then to a human reviewer. Both gates must approve before anything sends.

Correction Learning

Edits are stored and used to improve future drafts. The system learns from your judgment, adapts to your firm's voice, but never bypasses the approval requirement.

Zero Unauthorized Communications

LP letters, investor emails, quarterly reports, marketing content, and LinkedIn posts sent without approval: zero. No code path bypasses this for external content.

Every Surface Covered

Investor emails, quarterly letters, market commentary, LinkedIn posts, website content, capital call notices, and LP portal updates all require compliance review plus human approval.

Triggerstructured

Market data update, quarter end, LP inquiry, or scheduled publication

Processstructured

Pipeline gathers data, enriches context, routes to appropriate agent

Generateai

AI drafts content with full context, market data, and firm voice

Compliance Gatestructured

the Compliance Review agent validates SEC Marketing Rule compliance, disclaimers, performance claims

Human Reviewhuman

Draft appears with full content, recipient, compliance status for approval

Action & Audithuman

Only on explicit approval does anything send. Full audit trail recorded.

4 structured steps
1 AI step
2 human steps

Compliance-First Governance

Every agent operates from the same source of truth with SEC-aware governance enforced at the infrastructure level.

The biggest risk in multi-agent systems for asset managers is not just inconsistency — it is regulatory exposure. AssetBoxx solves this with unified compliance governance across all 26 agents.

Unified Compliance Layer

Every agent operates under the same compliance framework: SEC Marketing Rule awareness, disclaimer requirements, performance claim validation rules, and recordkeeping obligations. the Compliance Review agent enforces these standards across every external output from every agent.

Policies Defined Once, Enforced Everywhere

Compliance constraints are defined once and enforced across all 26 agents. 'All performance claims must reference source data and time periods' applies to the Research Writer, LinkedIn Agent, and every other agent that produces external content.

Preventing Regulatory Drift

Without governance, agents drift in tone, compliance rigor, and disclaimer usage. The governance layer keeps SEC compliance standards constant regardless of which agent produces the content or how many are operating simultaneously.

LP Data Isolation as Governance

Fund A's LP data is isolated from Fund B's at the database level. Investor records, fund documents, and communications are separated by infrastructure controls, not prompt instructions an AI might ignore. This isolation is a governance requirement, not just a security feature.

Personize

Governance Powered by Personize

AssetBoxx uses Personize as its unified memory and governance infrastructure layer. Personize provides persistent memory grounded in real customer history and enforces governance policies across every agent and workflow, ensuring consistent, compliant behavior at scale.

Agent Boundaries and Fiduciary Permissions

Every agent operates within defined permissions aligned with fiduciary duty. No agent can expand its own access.

Read/Write Separation

Each agent has explicit access boundaries designed for fiduciary compliance. The Research Writer accesses market data but cannot send investor communications. The Compliance Agent reviews content but cannot alter fund performance data.

No Self-Expansion

Agents cannot expand their own permissions. New capabilities require supervised deployment with compliance review. Access is granted by humans through configuration, never requested at runtime.

Internal vs. LP-Facing

Internal agents (research, data gathering, monitoring) produce no external output and need no approval. LP-facing agents require both compliance gate and human approval on every output.

SEC Recordkeeping Enforcement

Every agent action that touches LP data or produces external content is logged with the detail required for SEC recordkeeping: which agent, which data, when, what output, who approved.

Cryptographic Accountability

Every action taken by every agent is recorded in hash-chained audit logs. Each entry carries a SHA-256 hash linking it to the previous entry. For any regulatory inquiry, you walk the chain backward from the result to the trigger: the data that informed it, the compliance checks performed, the human who approved it. The chain is tamper-evident. Entries cannot be edited, inserted, or deleted after the fact without detection.

Trust Is Earned, Not Assumed

Autonomy increases only after demonstrated reliability through four phases. LP-facing approval never goes away.

Weeks 1-3

Shadow Mode

All 26 agents observe and report but take no action. Every classification, recommendation, and draft is logged for review. Your team validates accuracy against your standards.

Weeks 3-6

Supervised Execution

Agents generate outputs with 100% human approval required. the Compliance Review agent compliance gate active on all external content. Corrections feed back into the system.

Weeks 6-10

Calibrated Trust

Low-risk internal operations run with reduced oversight: market data gathering, research compilation, internal analysis. All LP-facing actions still require compliance gate plus human approval.

Week 10+

Production Steady State

Full capacity with permanent dual-gate approval on all LP-facing and external communications. Internal research and data operations run autonomously. Monthly compliance posture reports document ongoing adherence.

LP-facing approval never goes away. Every investor email, quarterly letter, market commentary, and LinkedIn post always requires compliance review plus human approval. Trust escalation applies only to internal operations. This is a fiduciary requirement, not a preference.

Structured in Practice

How Structured Agents blend deterministic code with targeted AI for asset management operations.

Investment Research Writer

Market Commentary Pipeline

Structured Code (6 steps)
  • FOMC statement ingestion and parsing
  • FRED economic data API polling
  • Earnings data aggregation from verified sources
  • Credit spread and yield curve monitoring
  • Source attribution and citation tracking
  • Research RAG pipeline for institutional sources
AI Model (2 steps)
  • Market commentary drafting in GP voice
  • Quarterly narrative synthesis from multiple data sources
Compliance Review

SEC Marketing Rule Pipeline

Structured Code (6 steps)
  • Disclaimer library matching and insertion
  • Performance claim detection and source verification
  • Testimonial and endorsement classification
  • Fair balance validation rules
  • Recordkeeping metadata generation
  • Compliance status tracking across all content
AI Model (2 steps)
  • Nuanced compliance judgment on edge cases
  • Context-aware disclaimer recommendations
LP Relationship Management

Investor Communication Pipeline

Structured Code (6 steps)
  • LP contact database management
  • Distribution list maintenance
  • Communication frequency tracking
  • Capital call notice scheduling
  • Document delivery confirmation logging
  • Investor portal update triggers
100% Structured. Zero AI model costs.

See the Governance Framework Live

We show you approval flows, compliance gates, agent boundaries, and audit trails in the live system. Not slides. The production platform handling real fund operations.

Every claim in this document is verifiable in the live system.