← Quantitative Proof QUANTITATIVE PROOF · INVESTMENT

Portfolio construction where observed data, risk models and generated scenarios remain distinct.

FinanceGPT investment workflows can combine portfolio holdings, observed market history, quantitative risk measures and optimization with generated scenario evidence while keeping research outputs separate from observed data and execution authority.

Evidence typeGoverned workflow demonstration Action authoritySeparate

THE QUESTION

How can an investment team construct and stress a portfolio without treating generated research as observed market evidence?

THE EVIDENCE PATH

01Portfolio holdings

02Observed market history

03Risk calculation

04Portfolio construction

05Scenario evidence

06Human review

Observed / supplied evidence

What must be available.

Portfolio holdings, constraints and investment objectives.
Observed market history from the governed market-data layer for the selected instruments.
Risk and portfolio-construction assumptions such as limits, target exposures or optimization constraints.

Calculated

What owns the numbers.

Risk measures and portfolio optimization are calculated by quantitative portfolio and risk methods.
Observed data remains explicitly distinguishable from derived metrics and FinanceGPT Synthetic research.
Conditional risk and scenario outputs can inform review without being relabelled as observed market facts.

Generated

What AI may generate.

LQM-generated or other synthetic scenario evidence can be used for counterfactual research.
Generated scenarios remain research evidence and do not become calibrated probabilities unless a separate validated method establishes that meaning.

Authority

What governs what happens next.

Portfolio constraints and model assumptions remain explicit.
Generated research is labelled as synthetic and remains execution-ineligible by itself.
A portfolio decision can require human review before any rebalance proposal advances.

WHAT THIS DEMONSTRATES

Evidence-supported conclusions.

A portfolio workflow can combine observed evidence, quantitative risk and generative research without collapsing them into one data class.
Portfolio construction can remain mathematically governed while natural language makes the workflow easier to interrogate.
Research output and transaction authority remain separate.

WHAT THIS DOES NOT CLAIM

Boundaries stay explicit.

It does not claim that generated scenarios are observed prices or future outcomes.
It does not claim investment performance or alpha.
It does not authorize a rebalance or trade.