ZTA Labs

Research Management

Governed AI, inside the investment workflow.

Research Management is the first market-confirmed workflow on the ZTA foundation. It connects approved evidence, governed model analysis, Model Council, Human Judgment, decision records, and outcomes in one controlled operating loop.

Why start here

Start with one governed workflow. Build from there.

Research sits upstream of many investment decisions. It is a strong proof point because it shows the full loop: approved evidence, governed analysis, Human Judgment, a decision record, and then a measurable outcome.

Research should not disappear when the decision is made. The system should preserve how the view was formed, what evidence supported it, and what happened afterward.

01 · Approved evidence

Filings, internal notes, and other approved sources enter the governed environment.

02 · Governed model analysis

Approved models analyze independently under policy-aware routing.

03 · Model Council

Elevated-risk work receives independent challenge and blinded evaluation.

04 · Human Judgment

The investment professional accepts, edits, rejects, or selects the alternative that best fits the firm’s framework.

05 · Decision record

Evidence, responses, scores, rationale, and outcomes are preserved as part of the institutional record.

Product surface

The evaluator recommends. The investment professional decides.

The product surface makes disagreement visible rather than hiding it inside a single merged answer.

Research Management · AQN credit deterioration assessment
Risk Tier 3 · Elevated

Assessment question

Does recent evidence warrant changing the internal credit view from Stable to Deteriorating?

Governance: Model Council required.

Responses and evaluation

Response A · 4.10

Cautious, evidence-backed, but does not treat liquidity deterioration as decisive.

Response B · 4.42

Selected by the analyst. Gives greater weight to deteriorating liquidity and covenant headroom.

Response C · 4.55

Evaluator recommendation. Strong on support, less aligned to the firm’s downside-first framing.

Response D · 3.91

Underspecified on risk transmission and less persuasive on the recommendation.

What becomes valuable over time

Visible disagreement

The analyst sees where responses diverge, which is often the information that should prompt a closer look.

Structured judgment

A reason code and rationale make the resulting signal reusable rather than anecdotal.

Governed learning record

Evidence, responses, evaluation, judgment, and outcomes can later be validated and promoted into institutional learning.