Let AI recommend, rebalance and execute. Keep every financial action inside a mandate you can replay.
A financial action is high-stakes and sometimes irreversible: one rebalance outside the mandate breaches a client agreement, one recommendation quietly promoted to execution moves real money on a stale approval, one confirmed on-chain transaction can never be recalled. Asset managers, wealth platforms and digital-asset desks are putting AI agents to work across the investment lifecycle — screening portfolios, drafting recommendations, computing rebalances, staging executions. KYE Financial AI Authority Console™ sits between the AI and the execution venue and decides, at the moment of each action, whether that action may proceed, under whose authority, inside which mandate, and with what evidence on the record.
A governed authority layer for AI across the financial action lifecycle.
- Action admissibility. Every consequential AI-assisted financial action — execute an investment recommendation, execute a portfolio rebalance, broadcast a DeFi transaction, bind a model-risk disposition, apply a model-driven limit change, distribute investment advice — is admitted only when a decision is recorded: which agent, which action, the intended effect, and the authority under which it proceeds. No recorded authority, no action.
- Mandate-scope and execution binding. A rebalance binds to the machine-readable mandate scope — instrument classes, concentration limits, turnover envelope — recorded before execution; an approved recommendation never carries forward into execution without a distinct execution-scoped grant, live at execution time. The console refuses an action outside its mandate — an AI agent cannot widen a mandate by its own reasoning, and approval at the recommendation stage is not authority at the execution stage.
- Named-principal finality on irreversibles. An AI-initiated on-chain transaction stays advisory until a named, currently-authorised human principal records approval — because a confirmed on-chain transaction has no recall path, the authority gate sits before broadcast. The accountability for irreversible financial execution is personal and named — it never transfers to the AI.
- Supervisory-review replay. Every governed action carries its bound chain (action record / mandate or execution scope / authority decision / decision map) and an ed25519 signature against the firm’s published keys, so a client dispute or a supervisory review can be answered by replaying exactly what executed and under whose authority — offline, from public keys alone.
Honest scope: KYE™ governs the AI’s authority and evidence, not the model and not the venue.
KYE™ does not run trading strategies, portfolio-optimisation or risk models, does not execute market or on-chain transactions itself, and does not give investment advice — those stay with your investment process, your models and your execution venues. KYE™ sits above them. What KYE™ does is govern the moment an AI agent acts: whether the action is admissible, under whose authority, inside which mandate, and with what record on the file. That maps cleanly to the duties your oversight already sets — the Securities and Futures Act licensing-and-conduct boundary, the MAS consultation guidelines on AI risk management (oversight, model lifecycle, record-keeping), MAS FEAT (fairness, ethics, accountability, transparency), and the Payment Services Act for digital-payment-token services. Each is a governed authority-and-evidence boundary, never a claim that KYE™ makes the investment decision for you. Authority finality, not the model.
Pick the tier that fits your investment function.
Solo
For a single desk or fund team. One mandate on synthetic data, four-week engagement, the full action-admissibility and named-principal finality boundary.
- Action admissibility across the financial action lifecycle
- Resolved authority on every consequential action
- Single-team workspace, synthetic / anonymised data only
- Signed end-of-pilot evidence pack
Practice
For an asset-management or wealth function. Multiple mandates, multi-user roles, machine-readable mandate-scope binding on every rebalance, six-week engagement.
- Multiple mandates and action types
- Recommendation / execution / mandate-owner / model-risk roles held by distinct named principals
- Mandate-scope binding required before any rebalance executes
- Signed end-of-pilot evidence pack
Network
For a bank with an AI programme, a multi-desk institution, or a digital-asset platform. Full lifecycle including the recommendation-to-execution boundary end to end, named-human finality on every on-chain execution, a closing dispute-replay exercise, ten-week engagement.
- All action types across every participating desk
- Recommendation approvals shown NOT carrying forward into execution without a distinct execution-scoped grant
- On-chain executions gated behind named-human finality
- Dispute determinations bound to a replayable authority chain, re-verified from public keys alone
Pilot fees credit one hundred per cent against the first-year annual licence when signed within sixty days of pilot close-out.
Portfolio oversight leads, model-risk officers, digital-asset operations leads, and compliance owners.
- Portfolio oversight lead. You own the mandates. KYE™ makes each mandate a machine-readable scope envelope and computes whether each AI-assisted rebalance or execution is admissible — so the AI helps without quietly trading outside what the client agreed.
- Model-risk officer. You own which models may drive which decisions. KYE™ binds every model-risk disposition and model-driven limit change to the model’s identity, version and validation state, with an evidence basis sealed before the disposition stands.
- Digital-asset operations lead. You own what gets broadcast on-chain. KYE™ keeps every AI-initiated on-chain execution advisory until a named, currently-authorised principal approves — the gate sits before the irreversible step.
- Compliance owner. You answer to oversight and to supervisory reviewers. KYE™ gives every AI-assisted financial action a recorded authority decision and a replayable evidence pack in one queue.