Human Stewardship. Autonomous Execution.
iPO Hub operates with a hybrid team: human principals who set policy and bear accountability, and autonomous agents who execute at scale — every action logged, every decision auditable.
The Accountable Principals
Our human leadership sets policy, bears legal accountability, and retains override authority over all autonomous operations.
Alex
Director & Founder
Sets platform policy, bears legal accountability, and retains override authority over all autonomous operations.
Legal Counsel
Legal Adviser
Reviews statutory obligations, advises on regulatory exposure, and approves compliance rule changes.
Data Protection Officer
ICO Compliance Lead
Oversees GDPR obligations, subject access requests, and data sovereignty architecture.
The 4-Tier Human-in-the-Loop Matrix
Every agent action is assigned a tier that determines how much human oversight is required before it executes. Tiers are set by confidence score, financial value, and legal consequence — not by the agent itself.
Tier 1 — Autonomous
The agent acts independently within its defined scope. Every action is logged to the append-only event store. No human approval is required unless confidence falls below 85% or a safety threshold is breached.
Examples: Routine compliance monitoring, tenant notification dispatch, inbound message classification.
Tier 2 — Supervised
The agent proposes an action and places it in the human review queue. A human operator must approve before the action is executed. Used for decisions with financial or legal consequence.
Examples: Invoice discrepancies over 10%, certification expiry escalations, SPV verification packs.
Tier 3 — Human-Controlled
The agent drafts or flags but cannot act. Dual human sign-off is required. Reserved for statutory notices, security overrides, and actions with irreversible legal effect.
Examples: Section 8 / Section 21 notices, security policy overrides, amounts over £5,000.
The Agent Workforce
Each agent operates within a strictly defined scope. No agent can custody funds, sign contracts, or act outside its authorised domain.
Triage & Classification Router
Tier 1 — Autonomous
Classifies every inbound communication and routes it to the correct specialist agent. Escalates to human review when confidence falls below 85%.
- ›Multi-channel ingestion
- ›AI-powered classification
- ›Confidence-gated escalation
Guardrail: Low-confidence classifications → human review queue
Compliance & Regulatory Officer
Tier 1 — Autonomous
Monitors gas safety, EICR, EPC, and legionella certification expiry across the entire portfolio. Creates escalation tasks 30 days before breach.
- ›Awaab's Law enforcement
- ›Certification expiry tracking
- ›Regulatory breach escalation
Guardrail: Critical expiries → Tier 2 supervised; legal action → Tier 3 dual sign-off
Financial Governance Officer
Tier 2 — Supervised
Performs 3-way invoice matching against work orders and contractor quotes. Flags discrepancies over 10% and amounts over £5,000 for human sign-off.
- ›3-way invoice matching
- ›SPV verification pack generation
- ›Tripartite boundary enforcement
Guardrail: All financial approvals require human sign-off
Communications Officer
Tier 1 — Autonomous
Drafts and dispatches tenant communications across email, SMS, and in-app channels. Every message carries an AI transparency badge and respects tenant communication preferences.
- ›Template-driven outbound comms
- ›Opt-in/opt-out enforcement
- ›AI transparency badging
Guardrail: All outbound comms carry mandatory AI disclosure
Security & Trust Officer
Tier 3 — Human-Controlled
Sanitises PII from agent contexts, detects prompt injection attempts, and enforces brand-safety rules on all outbound communications. Requires dual sign-off for any security override.
- ›PII sanitisation
- ›Prompt injection detection
- ›Brand-safety enforcement
Guardrail: Deterministic rule enforcement — no autonomous LLM reasoning
Every Agent Action Is Auditable
All autonomous decisions are recorded in an append-only event store. Human operators retain override authority at every tier. No agent can custody funds, sign contracts on behalf of iPO Hub, or act outside its defined scope.
Read our governance framework →