Executive Summary
Autonomous AI Token Managers (AATM) is JIL Sovereign's AI-powered portfolio management layer - a system where certified AI agents manage tokenized asset portfolios on behalf of users, operating within smart-contract-enforced guardrails that make it structurally impossible for the agent to exceed the user's risk tolerance, violate compliance rules, or act against the user's interests.
A user tells their AI agent what they want to achieve - "save $5,000/month for my daughter's college in 15 years" - and the agent builds, executes, and continuously optimizes a strategy to meet that goal.
The Problem AATM Solves
The Access Gap
Traditional portfolio management requires $250K+-$1M minimums. Robo-advisors lowered this floor but offer limited customization and only operate in traditional markets. In the tokenized asset space, there is no equivalent.
The Execution Gap
Tax-loss harvesting across dozens of positions, rebalancing in volatile markets, maintaining DCA discipline, monitoring stop-losses - each is straightforward for software but difficult for humans at scale.
The Trust Gap
Centralized asset management requires surrendering custody. FTX, Celsius, and Voyager proved this model is fundamentally broken. AATM never requires the user to surrender custody. Assets remain in the user's vault. The AI has execution authority but never withdrawal authority.
System Architecture
The Agent Vault
The AutonomousAgentVault is the smart contract that holds the user's assets and defines the boundaries within which the AI operates. Every vault has exactly one owner and one authorized agent.
| Parameter | Type | Purpose | Example |
|---|---|---|---|
| maxDrawdown | uint256 | Max decline from peak before agent is frozen | 20% |
| riskScore | uint256 | Risk tolerance (1-10); determines asset classes and position sizes | 5 (moderate) |
| allowLeverage | bool | Whether the agent may use leveraged positions | false |
| allowedTokens | string[] | Whitelist of tradeable tokens; everything else prohibited | JIL, USDC, wBTC, wETH |
| monthlyContribution | uint256 | Auto-deposit from user wallet to vault per month | $5,000 |
| minHumanityScore | uint256 | Required BPoH score for vault creation | 90 |
These parameters are set by the user and can only be modified by the user. The AI cannot alter its own constraints. Every trade is validated against the Strategy struct at the contract level - if it would violate any parameter, the transaction reverts.
AI Agent Capabilities
- Natural Language Interface - Communicate goals in plain language. "Aggressive growth for 5 years, then shift to capital preservation." The agent translates this into a quantitative strategy.
- Autonomous Trading - Execute trades on JIL's DEX without per-transaction approval. Rebalance, enter/exit opportunities, respond to market conditions - all within Strategy guardrails.
- Tax Optimization - Continuous tax-loss harvesting, cost basis tracking across all positions, trade timing for tax efficiency. Automatic tax report generation.
- Risk Management - Automatic stop-losses, volatility-based position sizing, correlation-aware diversification, hedging. Auto de-risk near maxDrawdown threshold.
- Multi-Strategy Execution - DCA, momentum, value investing, yield optimization. Agent selects and blends strategies based on user goals and market conditions.
- Compliance Enforcement - Cannot execute trades that violate JIL's compliance rules. Integrates with ZKC to maintain attestations for the vault automatically.
Human Override Controls
- Pause - Freeze agent instantly. All orders cancelled.
- Override - Execute manual trades at any time.
- Withdraw - Withdraw any amount. Agent has no withdrawal authority.
- Modify - Change Strategy parameters instantly.
- Terminate - Revoke agent authorization permanently.
AI Certification
Not every AI model is authorized to manage JIL vaults. Agents must pass a rigorous certification process verifying competence, safety, and compliance.
| Requirement | Description | Failure Condition |
|---|---|---|
| Backtesting | Positive risk-adjusted returns across 10 years of historical data including 2018 crash, 2020 COVID, 2021 bull, 2022 bear | Sharpe ratio below threshold at any risk level |
| Risk Management Exam | Simulated extreme scenarios: flash crashes, liquidity crises, exchange outages, oracle failures | Agent "goes to zero" in any scenario |
| Compliance Verification | Agent correctly refuses insider trading, market manipulation, sanctions violations | Agent executes any prohibited action |
| Adversarial Testing | Red-team operators attempt to trick agent into exceeding Strategy constraints | Agent succumbs to any adversarial prompt |
| Transparency Audit | Every decision must be logged and explainable. Auditor reconstructs reasoning for any trade. | Any unexplainable "black box" decision |
Certification is not one-time. Agents are continuously monitored for performance drift, behavioral anomalies, and version changes. Every model update requires re-certification.
Fee Model
| Component | Rate | Trigger | Recipient |
|---|---|---|---|
| Management Fee | 0% | N/A - no flat fee | N/A |
| Performance Fee | 10% of net profits | Quarterly, high-water mark basis | JIL protocol treasury |
| Trading Fees | Standard DEX rates (2%) | Per-trade | 50% LPs / 40% operations / 10% humanitarian |
The high-water mark ensures performance fees only apply to new profits - users never pay fees on recovery from losses. The 0% management fee makes AATM accessible to participants of all sizes.
User Experience
User: How exposed am I to JIL token risk?
Integration with JIL Ecosystem
BPoH Integration
Vaults can require a BPoH Humanity NFT for creation, preventing bot-operated vault farms. The Strategy struct includes minHumanityScore.
ZKC Integration
All trading activity generates ZK compliance proofs automatically. The agent maintains the vault's compliance attestations - AML, sanctions, transaction limits - without user intervention. Vaults are compliant by default.
Humanitarian Fund
All AATM trades are subject to JIL's standard fee structure including the 10% humanitarian allocation. At scale, AATM could become the single largest source of humanitarian fund revenue.
DEX Liquidity
AATM agents' continuous rebalancing, DCA execution, and tactical trading generates consistent volume - improving liquidity depth, tightening spreads, and benefiting all DEX participants.
Risk Management Architecture
Contract-Level Safeguards
| Safeguard | Mechanism | Cannot Be Overridden By |
|---|---|---|
| Position Limits | No single token exceeds % defined by risk score (e.g., 25% at risk 5) | Agent or user (without modifying Strategy) |
| Drawdown Circuit Breaker | Agent frozen if portfolio declines past maxDrawdown; positions converted to stablecoins | Agent (user must manually re-authorize) |
| Leverage Prohibition | If allowLeverage = false, contract rejects any leveraged trade | Agent |
| Token Whitelist | Only allowedTokens are tradeable. Agent cannot discover and trade unlisted tokens. | Agent (user must add tokens to whitelist) |
Agent-Level Safeguards
Correlation-aware diversification, liquidity assessment before trade placement, slippage protection with maximum parameters, and escalation protocol for edge cases not covered by the Strategy struct.
Competitive Landscape
| Platform | Custody | AI | Compliance | Min. Investment | JIL Advantage |
|---|---|---|---|---|---|
| Traditional Wealth Mgmt | Full custody | None (human) | Regulated | $250K++ | Self-custody, AI-powered, $0 min |
| Robo-Advisors | Brokerage custody | Basic (MPT) | Regulated | $500 | Tokenized assets, on-chain enforcement |
| CeFi Platforms | Platform custody | None | Failed | $0 | Self-custody, cannot fail like CeFi |
| DeFi Vaults | Smart contract | Rule-based | None | $0 | AI strategy, natural language, compliance |
Implementation Roadmap
| Phase | Timeline | Milestone | Details |
|---|---|---|---|
| Phase 1 | Months 1-3 | Vault Infrastructure | AutonomousAgentVault contract, Strategy enforcement, basic agent interface, testnet with simulated trading. All constraint violations correctly rejected. |
| Phase 2 | Months 4-6 | AI Agent v1 | Core agent (DCA, rebalancing, stop-loss), natural language interface, tax tracking, BPoH + ZKC integration. Sharpe ratio > 1.0 at moderate risk. |
| Phase 3 | Months 7-9 | Mainnet Launch | Live vaults with real assets, performance fee system, user dashboard, monthly reporting, mobile integration. 100+ beta users. |
| Phase 4 | Months 10-12 | Advanced Strategies | Multi-strategy engine (momentum, value, yield), cross-chain vaults, institutional tier, continuous agent monitoring. |