Show Why the Decision Was Made.
Govern is the capital-decision proof layer for teams that have to defend an allocation choice to a committee, an auditor, or a regulator. It assembles what is known, what is missing, and what still requires human review — as a traceable record, not a narrative.
AI analyzes. Deterministic systems validate. Humans authorize. Capital remains governed.
Know whether a capital action fits the mandate — before it reaches execution.
Change the amount, the proposed action and the rail it requests, the evidence, or the mandate. SigmaX evaluates the proposed action against a sample capital mandate and shows what passes, what requires review, and what stops.
No outcome is looked up — the verdict is computed from the controls below, not selected from a list. The mandate and the portfolio figures themselves are pre-set samples.
The mandate, the portfolio figures, the evidence scores, and the market and cost assumptions listed beside the checks are illustrative. There is no real client, account, or holding behind them, and the instrument is a generic asset label rather than a security.
Your amount and your evidence scores are what the gate compares. This is not a copy of the gate the governed API runs kept in step with it — it is the same module. The API imports it and adds what a browser has no business holding: the server clock, the caller's identity, and the durable record a determination is sealed into.
Nothing on this page can move money. There is no order path, no transfer path, no broker connection, and no external send in the code that produced the result below.
Set the mandate and the action
A managed household account at a wealth advisory firm. Cash floor $100,000. Single name capped at 15% of the portfolio. Anything at or above $25,000 needs an advisor approval, and one advisor may approve up to $175,000.
A capital-linked purchase, requested as a broker handoff draft — a ticket a human reviews, never an order SigmaX submits.
The amount is not a label. It is compared against the per-action risk budget, it sets the cash left after the action, and it sets the projected weight of the position — three separate limits, each of which can decide on its own.
A mandate can require that the case against the action was examined too. Lower this below the standard and the decision holds, whatever the amount.
The mandate decides
The mandate allows the action but not without a named human authorizing it. SigmaX identifies who; it never approves on their behalf.
This is an evaluation, not an authorization or execution instruction.
- passNo capital or execution authorityMandate Engine v1 can classify allowed next steps but cannot move capital or submit orders.
- passDecision packet requiredEvery ActionIntent must trace to a DecisionPacket.
- passDecision evidence standardEvidence score 78; mandate minimum 65.
- passCounter-evidence standardCounter-evidence score 52; mandate minimum 35.
- passScenario reviewScenario score 66; mandate requires scenario review.
- passAsset allowancepublic_equity checked against mandate allowed/forbidden asset classes.
- passInstrument eligibilitySINGLE-NAME-EQUITY checked against mandate instrument eligibility.
- passAction allowancebuy checked against blocked mandate actions.
- passLive execution disabledRequested rail is not a live execution rail.
- passRail allowancebroker_handoff checked against mandate-approved rails.
- passAccount scopesample-household-managed-1 checked against accounts in scope.
- passMandate expirationMandate carries no expires_at and is explicitly recorded as perpetual.
- passStop conditionsNo active stop condition supplied.
- passLiquidity reserveLiquidity reserve is within mandate.
- passCash after actionCash after action 165000; mandate minimum 100000.
- passPer-action risk budgetRequested notional 15000; mandate max 250000.
- passSingle-asset concentrationProjected concentration 10.8%; mandate max 15%.
- passAsset-class notional capNo asset-class notional cap applies to a public_equity money move.
- passRole permission / per-role limitAt least one permitted approver role covers this move, or the move requires no approval.
The engine runs these gates, but nothing here observed their inputs. They are not governance findings and are not counted as passes — and the record below says the same thing the panel does: each is warn, classed unknown, and listed under missing_requirements. They are outstanding, not blocking: neither decides this outcome.
- declaredMarket realityNot observed. This page reads no venue and no market calendar, so nothing was supplied to this gate and it decided nothing. It is recorded as an outstanding requirement, not as a pass.
- declaredSlippage limitNo fill estimate supplied. Nothing here quoted a price, so the cost gate has nothing to decide on and the engine records that rather than a cost.
Change one number, and the answer changes
Every amount below was run through the same engine that produced the result above, on the same evidence scores you set, across this mandate's own amount range — and the label beside it is what the engine returned. Nothing in this table is written copy: move either evidence slider and the whole table re-decides with the headline, and editing a limit in the sample mandate moves these numbers.
Each row is the first amount this slider can be set to that returns that answer — the scan steps by the slider's own $1,000, so the amount below the row genuinely returns something else.
- $0Approval requiredno check blocked at this amountthe amount this control opens on, $15,000, returns this
- $68,000Blocked by mandatestopped by: Single-asset concentration
- $81,000Blocked by mandatestopped by: Liquidity reserve, Cash after action, Single-asset concentration
- $251,000Blocked by mandatestopped by: Liquidity reserve, Cash after action, Per-action risk budget, Single-asset concentration
The output, unedited
This is the result object the gate returned, complete and unedited — verdict, reason codes, every check, the required approvals, and the authority stamp. It is not a summary of the decision above; it is the decision above. It re-runs on every change you make, and it is deterministic: the same inputs produce the same bytes on any machine.
Why the timestamps are not now. checked_at and produced_at both read 2026-08-01T12:00:00.000Z because the authorization clock is fixed and published, not read from your device. A gate must never be decided by a clock the governed party supplies. Both sample mandates are recorded as perpetual, so the fixed stamp changes no outcome — and a fixed stamp is also what keeps the server and browser renders byte-identical.
{
"mandate_check_id": "mandate_check_1b2c9388",
"mandate_id": "mandate_sample_advisory_household",
"money_move_id": "move_sample_advisory_household_buy_15000",
"verdict": "needs_approval",
"allowed": false,
"blocked": false,
"needs_evidence": false,
"needs_counter_evidence": false,
"needs_fit_check": false,
"needs_preflight": true,
"needs_approval": true,
"paper_only": false,
"handoff_ready": false,
"execution_not_permitted": true,
"reason_codes": [
"approval_required"
],
"missing_requirements": [
"Market reality",
"Slippage limit"
],
"next_allowed_actions": [
"advisor_task"
],
"mandate_version": 3,
"checked_at": "2026-08-01T12:00:00.000Z",
"checks": [
{
"check_id": "check_b9847c78",
"label": "No capital or execution authority",
"status": "pass",
"detail": "Mandate Engine v1 can classify allowed next steps but cannot move capital or submit orders.",
"evidence_class": "derived"
},
{
"check_id": "check_18f90acb",
"label": "Decision packet required",
"status": "pass",
"detail": "Every ActionIntent must trace to a DecisionPacket.",
"evidence_class": "derived"
},
{
"check_id": "check_a2301a39",
"label": "Decision evidence standard",
"status": "pass",
"detail": "Evidence score 78; mandate minimum 65.",
"evidence_class": "derived"
},
{
"check_id": "check_d4553d00",
"label": "Counter-evidence standard",
"status": "pass",
"detail": "Counter-evidence score 52; mandate minimum 35.",
"evidence_class": "derived"
},
{
"check_id": "check_23009866",
"label": "Scenario review",
"status": "pass",
"detail": "Scenario score 66; mandate requires scenario review.",
"evidence_class": "derived"
},
{
"check_id": "check_25d2e25e",
"label": "Asset allowance",
"status": "pass",
"detail": "public_equity checked against mandate allowed/forbidden asset classes.",
"evidence_class": "derived"
},
{
"check_id": "check_4f19f51b",
"label": "Instrument eligibility",
"status": "pass",
"detail": "SINGLE-NAME-EQUITY checked against mandate instrument eligibility.",
"evidence_class": "derived"
},
{
"check_id": "check_b8e82a8e",
"label": "Action allowance",
"status": "pass",
"detail": "buy checked against blocked mandate actions.",
"evidence_class": "derived"
},
{
"check_id": "check_1a231315",
"label": "Live execution disabled",
"status": "pass",
"detail": "Requested rail is not a live execution rail.",
"evidence_class": "derived"
},
{
"check_id": "check_eec18e1e",
"label": "Rail allowance",
"status": "pass",
"detail": "broker_handoff checked against mandate-approved rails.",
"evidence_class": "derived"
},
{
"check_id": "check_d30b6a19",
"label": "Account scope",
"status": "pass",
"detail": "sample-household-managed-1 checked against accounts in scope.",
"evidence_class": "derived"
},
{
"check_id": "check_8e8005fa",
"label": "Mandate expiration",
"status": "pass",
"detail": "Mandate carries no expires_at and is explicitly recorded as perpetual.",
"evidence_class": "derived"
},
{
"check_id": "check_1e2cf1c7",
"label": "Stop conditions",
"status": "pass",
"detail": "No active stop condition supplied.",
"evidence_class": "derived"
},
{
"check_id": "check_fa4aa39e",
"label": "Liquidity reserve",
"status": "pass",
"detail": "Liquidity reserve is within mandate.",
"evidence_class": "derived"
},
{
"check_id": "check_19855701",
"label": "Cash after action",
"status": "pass",
"detail": "Cash after action 165000; mandate minimum 100000.",
"evidence_class": "derived"
},
{
"check_id": "check_6e7b1103",
"label": "Per-action risk budget",
"status": "pass",
"detail": "Requested notional 15000; mandate max 250000.",
"evidence_class": "derived"
},
{
"check_id": "check_45b9e020",
"label": "Single-asset concentration",
"status": "pass",
"detail": "Projected concentration 10.8%; mandate max 15%.",
"evidence_class": "derived"
},
{
"check_id": "check_0c6ec05c",
"label": "Asset-class notional cap",
"status": "pass",
"detail": "No asset-class notional cap applies to a public_equity money move.",
"evidence_class": "derived"
},
{
"check_id": "check_86c2237d",
"label": "Market reality",
"status": "warn",
"detail": "Market state was not observed (market hours unobserved; venue availability unobserved). This gate decided nothing.",
"evidence_class": "unknown"
},
{
"check_id": "check_834d67ae",
"label": "Slippage limit",
"status": "warn",
"detail": "No fill estimate was supplied, so this gate decided nothing.",
"evidence_class": "unknown"
},
{
"check_id": "check_38f4911b",
"label": "Role permission / per-role limit",
"status": "pass",
"detail": "At least one permitted approver role covers this move, or the move requires no approval.",
"evidence_class": "derived"
}
],
"required_approvals": [
"advisor"
],
"allowed_rails": [
"paper_trading",
"broker_handoff",
"advisor_task",
"watchlist",
"blocked",
"no_action"
],
"blocked_by": [],
"produced_at": "2026-08-01T12:00:00.000Z",
"authority": {
"capital_authority": "none",
"execution_authority": false,
"external_send": false,
"customer_action": false,
"customer_eligibility": "not_customer_eligible",
"broker_submit_path": "none",
"provider_calls": "not_called",
"secrets_policy": "names_only"
}
}This preview does not produce a Decision Record, and there is nothing here to export. A Decision Record binds a decision to an identified caller, a stored mandate version, a content address, and an append-only audit trail, so it can be reconstructed and checked for tampering later. That sealing happens in the governed API against a durable store. This page has no account, no identity, and no store, so it has none to give you — and it will not hand you a file that looks like one.
Those four values are read off the evaluation you are looking at, not written into this page. Sample data only. No external action, no broker connection, no custody, no transfer, and no capital movement occurs from this website.
AI analyzes. Deterministic systems validate. Humans authorize. Capital remains governed.
Compliance isn’t a policy binder here. It’s a gate every capital action clears.
Before a proposed capital action can move, SigmaX recomputes what it can check deterministically — your client mandate bands, position caps, and concentration limits — and holds the action until every review your obligations require (suitability, tax, trading conduct) is recorded against it. SigmaX gates and documents the decision; your firm’s people and systems make the regulatory determinations. What fails is blocked or held. What passes carries its proof. A named human authorizes. And all of it lands in a record built to withstand examination.
Every proposed action is recomputed against the signed governing document — allocation bands, allowed assets, position caps, prohibitions.
Sample: 71% − 9 pp = 62%, checked against the 45–65% band; a 15% position blocked by a 10% cap.
The decision record requires suitability documentation — rationale, alternatives considered, cost, and risk-profile fit — and authorization is held until a person attaches it. SigmaX records the determination and gates the action; it does not make the suitability call itself.
Sample: the rebalance file must carry all four before the advisor can authorize.
When the wash-sale rule applies, the firm marks it as a required review and SigmaX holds any replacement lot until that review is recorded. SigmaX gates and documents the decision; your team or tax system makes the determination.
Sample: one replacement held pending its recorded review — not passed through.
Concentration and exposure limits are enforced continuously; exceptions require sign-off, evidence, and an explicit expiry.
Sample: 6.2% vs a 5.0% issuer limit → a 48-hour, evidence-backed exception with a cure task.
The firm’s restricted-name policy and best-execution duty (FINRA 5310-style) travel with any desk handoff as required, recorded checks, and the handoff is held until they clear. SigmaX attaches and gates these obligations; the firm’s systems and people make the determinations.
Sample: the committee’s broker-handoff draft ships with the restricted-name check and best-execution requirements attached.
Records are complete, append-only, attributable, time-stamped, and retrievable — the qualities examiners test record-keeping against.
Shown below: pull the record, read it, verify its integrity.
Configured per firm — these are the sample rule sets used in the walkthroughs above. Your mandates and policies become the machine-checked standard.
And when the examiner asks — the proof.
How each decision was made, the sources behind every fact, and who authorized it — pulled as a query, not reconstructed from inboxes. Sample data, real product surfaces.
Your compliance team — or your examiner, with your permission — pulls the record with a scoped, read-only query.
GET /v1/trigger-engine/audit?from=2026-01-01&intent=ai_demo_wealth_advisory Authorization: workspace-scoped key · read-only · least privilege → 200 OK · 8 entries · workspace-scoped, read-only (sample)
GET …/audit/export.csv
The complete record as a file your examiner opens in a spreadsheet — injection-safe, deterministic columns.
GET …/impact-ledger/export.csv
What each decision changed, exportable alongside the audit trail it reconciles against.
Behind the record: append-only storage enforced at the database level · named actors on every entry · provenance references and evidence classes (measured / derived / inferred) on every fact · workspace-scoped, least-privilege access · CSV and API export · recompute-on-read integrity verification.
Sample data shown; the query surfaces are the product’s real read-only audit APIs. Standards are referenced as what the data is checked against, configured to your firm — SigmaX is software, not legal advice or a certification. SigmaX gates and records these checks; the underlying tax, suitability, and trading-conduct determinations are made by your firm. Your compliance team maps these controls to your regulatory obligations.
Proof before motion.
- AI has no capital authority on this surface.
- The current Govern surface is read-only and operator-supervised.
- Every decision record links back to its underlying evidence.
- Nothing is marked validated without passing the gate that defines it.
- No live capital authority is granted by this surface.
Nine ways to make a decision defensible.
Each capability is a discrete module. Together they turn informal conviction into a reviewed, traceable, human-authorized record. The live surface shown in product is labeled, and interactive previews use illustrative data.
- 01Decision Records
Every authorization decision, validation outcome, and mandate change is captured as a structured Decision Record — an audit-chained lineage from observation to outcome. When a committee asks why, the record answers. The packaged one-click Decision-Record export is designed and in build.
- 02Proof Frontier
The Proof Frontier maps each thesis against the evidence it still needs. It makes the boundary between proven, challenged, and untested explicit, so reviewers see exactly where conviction outruns evidence.
- 03Decision Intelligence
Maps theses and their supporting evidence as a navigable Decision Graph. Relationships between ideas, sources, and market events are visible as connected nodes, so the evidence base can be inspected rather than re-read.
- 04Drift Radar
Monitors whether current holdings or active theses have drifted from the stated mandate. Structured alerts surface when cumulative drift crosses the threshold you set, keeping decisions aligned with the framework that authorized them.
- 05Scenario Lab
Runs counterfactual and stress scenarios against open theses before capital is committed. The Scenario Lab models how different market conditions affect each thesis so risk is evaluated on evidence, not optimism.
- 06Mandate Monitor
Continuously checks live posture against the investment mandate it must honor. Mandate Monitor records where a position sits inside or outside its envelope and flags breaches for human review.
- 07Counter-Evidence Engine
Actively searches for the evidence that would refute a thesis, not only the evidence that supports it. Surfacing the strongest case against a decision is part of making that decision defensible.
- 08Impact Ledger + Audit
Compiles the decision lineage, evidence base, and validation history into an export an auditor can read end to end. The pack shows what was known, what was missing, and who authorized the outcome.
- 09Investment Committee PacketsIn build
Designed to assemble the Investment Memo, supporting Decision Records, benchmark results, and open questions into a committee-ready packet — so the meeting reviews evidence rather than reconstructs it. Packet generation is on the roadmap.
No capital authority. Human authorization required.
SigmaX shows how capital should move; it does not move, custody, route, or auto-execute capital. Govern prepares the evidence and the decision record. The authorization itself belongs to the authorized human, committee, or capital-governance office outside this surface. Market participation involves risk, including possible loss of capital.
SigmaX provides capital decision intelligence, scenario analysis, and authority workflow software. SigmaX does not provide investment advice. SigmaX does not place trades, move funds, route customer capital, or manage client capital. Shadow/simulated/paper/backtested results are not live trading results. No result is a guarantee of future performance. Human approval is required before any capital-linked action.
Where SigmaX uses AI, it is governed as assistive software: truthful capability claims, human oversight, traceable records, disclosure of AI assistance, and no capital authority. SigmaX designs AI controls with reference to applicable securities obligations, FINRA existing-rule guidance, SEC AI-claims discipline, NIST AI RMF-style risk management, and EU AI Act transparency readiness. This is compliance-readiness design, not a claim that every deployment or jurisdiction-specific use has completed legal review.
Defensible decisions, on the record.
Request access to evaluate Govern, or walk the sample to see company input, mandate result, output packet, and cost avoided. The walkthrough is a written example; the Capital Action Preview on the homepage recomputes from its controls.