A model signal becomes a committee-authorized decision — and a desk-ready draft.
A quant model proposes a factor tilt in a governed sleeve. Walk it through the mandate, counter-evidence, and committee sign-off to a broker-handoff draft the trading desk reviews — SigmaX never places the trade.
What this demonstrates: The committee proves what was allowed, what evidence supported it, and who signed off — and the desk receives a reviewable draft, not an executed order.
- 1. Inputs
- 2. Signal
- 3. ActionIntent
- 4. Mandate
- 5. Preflight
- 6. Compliance
- 7. Authorize
- 8. Outcome
Confirm the account context
SigmaX starts from a Capital Twin: the account model that carries positions, mandate, constraints, and context. Here is the sample context for this walkthrough.
- Sleeve
- Endowment Global Equity
- IC-governed, benchmarked
- Signal source
- Multi-factor model, weekly refresh
- Proposed change
- +3% quality-factor tilt, funded from broad beta
- Tracking-error budget
- ≤ 3.0%
- Current 2.1%
- Governance rule
- Tilt changes > 2% require an IC vote
- intent_id
- ai_demo_investment_committee
- twin
- Endowment Global Equity — IC-governed sleeve (Capital Twin)
- proposed_by
- Quant research model
- route
- —
- reason_code
- pending
- evidence
- 4 measured · 3 derived · 1 inferred
Sample packet. SigmaX produces a decision record and a non-executing route — it never places a trade, moves money, or routes customer capital.
This is an illustrative sample with fabricated data. SigmaX analyzes and governs; deterministic systems validate; a human authorizes; capital remains governed. SigmaX does not trade, custody assets, move money, or route customer capital.
Every proposed capital action, governed the same way.
Whatever AI or team proposes the action, SigmaX runs the same mandate, Preflight, compliance, human-authorization, route, and audit sequence you just walked through. SigmaX analyzes and governs; a human authorizes; capital remains governed.