SigmaScope

The problem

Modern Portfolio Theory is well understood: choose the mix of assets with the highest expected return for a given level of risk. Building it to the standard a bank expects is harder than the math. The numbers have to be right, reproducible and explainable. Every change has to be traceable to a reason and a reviewer.

What the AI team built

SigmaScope is a portfolio construction and analysis platform for advisors.

• Clients with their own logins, each seeing their own portfolios.
• Portfolios in versions: a benchmark, the actual holdings, and ideas to compare against them.
• The numbers for every version: each asset’s return, volatility and beta, and the full covariance matrix.
• An efficient-frontier explorer for trying allocations along the frontier.
• A data-validation gate: a version can’t be finalized until its data passes.
• Fundamentals and sentiment explorers for the assets in a portfolio.
• US and Canadian treasury rates as the risk-free asset.

It was built by AI developers working under the rules that became Agent-Governed Development.

What it taught

Writing the code was not the hard part. Keeping a team of AI developers honest was. It took the same things a good IT department relies on:

• Roles with clear authority. Each agent knows what it may decide on its own and what needs someone else.
• An order of work. Feedback first, documentation second, tasks and tests third, code fourth, validation fifth, a retrospective sixth.
• No one approves their own work. A change is approved by a role that didn’t make it.
• A person who can step in. Some decisions always go to a human.
• A written record, so the next agent, and the next auditor, can see why.

How it became AITLA

SigmaScope generated the context for AITLA. AITLA turns those lessons into policies a machine can enforce, so the next team of AI developers starts with the rules instead of discovering them.

Read how those rules work now in the worked example, or the architecture they became in AITLA.