PGAIAGENTIC WEALTH

Global Strategy
Whitepaper v3.0

Agentic Wealth Platform · Multi-LLM + AI Skills × RWA
US Headquarters · Co-led by US Institutions & Singapore Sovereign Funds

Confidential · 2026 · Institutional Use Only
4+ LLMs 8+ Skills 11 Partners SEC+MAS Regulated US HQ

Contents

  1. Executive Summaryp.4
  2. The Thesisp.8
  3. Multi-LLM Architecturep.14
  4. The Skill Meshp.22
  5. Real-World Assetsp.30
  6. Institutional Partnersp.36
  7. Regulatory Frameworkp.42
  8. Custody Architecturep.46
  9. Fee Structurep.50
  10. Roadmapp.52
01
Executive Summary

PG-AI is a US-headquartered, AI-native private wealth platform co-led by Goldman Sachs, BlackRock, Temasek Holdings, and GIC. It orchestrates multiple frontier large language models with specialist AI skills to autonomously manage portfolios of real-world assets — custodied by five Tier-1 banks and regulated under both SEC and MAS jurisdictions.

Four Pillars

Intelligent4+ frontier LLMs (GPT-4o, Claude Opus, Gemini Ultra, Llama 4) reason in concert. 8+ specialist AI skills execute 24/7 on user-signed mandates.
StableUS Treasuries, investment-grade credit, gold, infrastructure RWA. A yield floor designed to survive every cycle.
SecureFive-bank custody: JPMorgan, Goldman Sachs, UBS, BNY Mellon, DBS. Ring-fenced SPV, multi-sig release, human override.
BackedCo-led by Goldman Sachs & BlackRock (US) + Temasek & GIC (Singapore). SEC + MAS dual regulatory coverage. Apollo, UBS, Deutsche Bank, BNY Mellon, DBS, JPMorgan as operational partners.

Core promise: Intelligent. Stable. Secure. — Four minds think. Eight skills act. Eleven giants guard.

02
The Thesis

The structural failure of private banking

A human team of 5–8 specialists works business hours, five days a week. Wealth sleeps 48 hours every weekend, gated by geography, audited once a year. This rigidity costs HNW individuals 100–300 bps annually in missed opportunities and timing drag.

AI Agents as the new operating system

PG-AI replaces the human team with an AI-native stack: multiple frontier LLMs provide the reasoning layer; specialist AI skills provide the execution layer. LLMs never trade directly — they reason about market state, user intent, and risk policy, then dispatch bounded skills to act. The result: 24/7/365 operation, sub-second response, zero emotional bias.

Why multi-model, not single-model

No single LLM is best at everything. GPT-4o excels at macro synthesis. Claude Opus leads in financial reasoning. Gemini Ultra brings multimodal capability. Llama 4 offers low-latency inference. PG-AI's routing layer selects the best model per task in real time. Skills stay the same; intelligence improves every quarter.

Why US headquarters + Singapore sovereign partnership

The United States provides the deepest capital markets, the most liquid asset pools, and the most sophisticated institutional investor base. Singapore provides the world's most trusted regulatory framework for cross-border asset management and two sovereign wealth vehicles (Temasek, GIC) with $800B+ in combined assets. PG-AI bridges both — US-headquartered, Singapore-guaranteed, globally distributed.

03
Multi-LLM Architecture

The routing layer

A model-agnostic router evaluates every task against a performance matrix updated weekly: task type, latency, context demand, cost, and accuracy.

LLMStrengthAssigned Tasks
GPT-4oBroad macro synthesisMarket Intel, Portfolio Construction
Claude OpusStep-by-step financial reasoningCompliance, Tax, Explainability
Gemini UltraNative multimodalEvent Watch, Earnings Analysis
Llama 4Low-latency inferenceRisk Guardrails, Real-time Monitoring

Model isolation

Each LLM operates in its own sandbox. Inter-model communication is mediated by the router, which strips sensitive data. User PII is never transmitted — only anonymized task parameters.

Continuous benchmarking

Weekly evaluation suite. New model versions enter shadow-mode: processing real tasks in parallel without affecting live portfolios. If a new version outperforms, it auto-replaces the incumbent.

04
The Skill Mesh

Architecture

Multi-LLM core at center. 8+ specialist skills orbit it — each bounded, auditable, policy-governed with formal spec, permission set, and audit trail.

The eight core skills

SKILL 01
Market Intel

Macro data, earnings, yield curves, news → live regime signal.

SKILL 02
Risk Guardrails

Drawdown cap, VaR, concentration rules enforced before any order.

SKILL 03
Compliance

Jurisdictional eligibility, suitability, Travel Rule — pre-checked.

SKILL 04
Tax Optimization

Tax-efficient lots, wash-sale avoidance, residency-aware routing.

SKILL 05
Rebalancing

Mandate-bounded drift correction. TWAP/VWAP execution.

SKILL 06
Custody Reconciliation

3-second signed holdings from 5 custodian banks via read-only API.

SKILL 07
Event Watch

Earnings, central banks, geopolitical → defensive mode escalation.

SKILL 08
Portfolio Construction

Target allocation across defensive (Treasuries, gold) and offensive (equities, infra) engines.

Skill lifecycle

New skills: spec review → sandbox → shadow-mode → committee approval. Deprecation: orchestrator re-routes automatically. Portfolio never disrupted.

05
Real-World Assets

Dual-engine strategy

Defensive engine provides yield floor. Offensive engine captures growth when signal permits.

EngineAsset ClassesObjective
DefensiveUS Treasuries, IG credit, allocated goldStable yield floor across cycles
OffensiveGrowth equities, private credit (Apollo), AI-infra RWAAlpha on high-confidence signals

Asset selection gates

Liquidity (48h exit) → Custodial eligibility (≥2 of 5 banks) → Compliance clearanceRisk budget (VaR envelope). Every asset must pass all four.

06
Institutional Partners

Co-led by US institutions (Goldman Sachs, BlackRock) and Singapore sovereign funds (Temasek, GIC). Eleven global Tier-1 partners form PG-AI's operational backbone.

Co-Lead · US Anchor
Goldman Sachs & BlackRock
Goldman Sachs ($2.8T AUM) + BlackRock ($10.5T AUM). Joint Wall Street anchors for capital markets, prime brokerage, ETF distribution, and strategic advisory.
Co-Lead · Singapore Sovereign
Temasek Holdings & GIC
Temasek + GIC, Singapore's two sovereign funds. $800B+ combined. Long-horizon capital, APAC distribution, and MAS alignment.
Prime Custody
JPMorgan
JPMorgan. $34T under custody. Settlement and clearing.
Private Credit
Apollo Global
Apollo. $671B AUM. Institutional credit pipeline.
Wealth Infra
UBS
UBS. Global wealth rails and HNW distribution.
European Gateway
Deutsche Bank
Deutsche Bank. EU custody, FX, and AIFMD administration.
Asset Servicing
BNY Mellon
BNY Mellon. $46.5T under custody. Attestation API.
APAC Banking
DBS Bank
DBS. Southeast Asia's largest bank. MAS-native operations.
Regulator · Guarantee
MAS — Monetary Authority of Singapore
Monetary Authority of Singapore. APAC regulatory anchor. Signed protection agreement covering asset segregation, daily reporting, freeze authority, and full recoverability.
07
Regulatory Framework

Dual-Jurisdiction Safety Net

United States (Primary): SEC Registered Investment Adviser. Delaware holding entity. Goldman Sachs and BlackRock as co-leads. FINRA compliance. SOC 2 Type II audit.

Singapore (APAC): MAS Capital Markets Services license. Signed protection agreement. Asset segregation, daily reporting, immediate freeze authority, full recoverability.

JurisdictionAuthorityLicenseStatus
United StatesSECRegistered Investment AdviserRegistered
SingaporeMASCapital Markets Services (CMS)Licensed
European UnionAIFMDAlternative Investment Fund ManagerAuthorised
UAEADGM FSRAAsset ManagementLicensed
Hong KongSFCType 9 Asset ManagementFiled
S · E · CRegisteredUS HQ
M · A · SRegulatedGuaranteed
S · P · VRing-FencedAssets
08
Custody Architecture

Five-bank diversified custody

CustodianRoleJurisdiction
JPMorganPrime custody, settlementUS / Singapore
Goldman SachsPrime brokerage, capital marketsUS / UK
UBSWealth infrastructure, cross-borderSwitzerland / SG
BNY MellonAsset servicing, attestation APIUS / UK
DBS BankOperational accounts, MAS-nativeSingapore

Three-party multi-sig

Any transfer above mandate threshold requires: custodian bank + PG-AI LLM orchestrator + independent human compliance officer. Compliance officer retains unilateral veto.

Real-time attestation

● LIVE ATTESTATION · VERIFIED 3s AGO

> jpmorgan.custody $ 1,204,331,890 · LOCKED
> goldman.prime $ 891,204,556 · SEGREGATED
> ubs.wealth $ 682,019,447 · SEGREGATED
> bny_mellon.assets $ 441,023,187 · ALLOCATED
> dbs.operational $ 218,554,902 · MAS-REG
> reserve_ratio 100.67% · OVER-COLLATERALIZED
> sec.filing CURRENT ✓
> mas.daily_report SUBMITTED ✓
> your_leaf INCLUDED ✓
09
Fee Structure

Zero management fee. 20% performance fee only above the investor's historical high-water mark. If AI doesn't make you money, we earn nothing.

FeePG-AIPrivate BankHedge Fund
Management0%0.75–1.50%1.50–2.00%
Performance20% above HWMRare / 10%20% above HWM
Entry / Exit0%0–3%Lock-up
10
Roadmap
2026 Q2

Mainnet Launch

8 core skills, 4 LLMs, SEC + MAS compliance. First institutional mandates. Five-bank custody live. Goldman Sachs + BlackRock + Temasek + GIC co-launch.

2026 Q3

Skill Expansion

ESG screening, alt-data ingestion, derivatives overlay, client reporting. Apollo private credit pipeline activated.

2026 Q4

Global Corridors

EU via AIFMD. BlackRock ETF distribution integration. Deutsche Bank European custody.

2027 Q1

Hong Kong Launch

SFC Type 9 activation. Greater China distribution. GIC infra co-investment program.

2027 Q2+

Open Skill Ecosystem

Third-party skill marketplace. Temasek governance advisory. Global AUM target: $10B+.

A
Appendices

A. Glossary

LLMLarge Language Model — frontier AI model for reasoning.
SkillBounded, auditable execution module in the Skill Mesh.
MandateUser-signed risk envelope defining permissible AI actions.
RWAReal-World Assets — tokenized traditional instruments.
SPVSpecial Purpose Vehicle — ring-fenced entity holding client assets.
HWMHigh-Water Mark — peak NAV above which performance fees apply.

B. Risk Disclosure

This whitepaper is for informational purposes only and does not constitute investment advice. AI-driven management carries risks including model error, data latency, infrastructure failure, regulatory change, and market risk. All investments involve risk of loss. Consult your own advisers.