Data, Explainability, And AI¶

How your results become governed analytical products you can query, compare, and explain.

The platform serves governed analytical products, not raw runtime folders. Your results, the summaries built on them, and the analytics you run all sit on the same auditable substrate — which is exactly what makes them consistent and explainable. Available today.

BigQuery As Your Analytical Backbone¶

Your analytical truth lives in BigQuery, organised so each stage of the data's life is first-class and governed. Available today.

flowchart LR
    A[Intake] --> B[Governed objects]
    B --> C[Execution state]
    C --> D[Outputs and facts]
    D --> E[Serving summaries]
    E --> F[BI and consumption]

This split is deliberate: you can always trace a served number back through the serving summary, the canonical fact, the execution that produced it, and the governed input it came from.

Explainability, Grounded In Truth¶

The goal of explainability here is not to let a model invent financial meaning. It is to let your analyst understand a result, compare it to relevant baselines, and interrogate the governed context behind it.

flowchart TD
    A[Results and evidence] --> B[Canonical facts in BigQuery]
    B --> C[Serving and comparison views]
    C --> D[Explainable summaries]
    D --> E[Workbench narratives and Q&A]

When the narrative and conversational layers arrive, they will follow the same rule as everything else: the model explains governed facts; it never invents them.

The Kinds Of Questions It Will Answer¶

The designed explanation experience is built around the questions a risk analyst actually asks — for example:

• why did exposure change versus the prior approved run?
• what were the largest contributors to a market-risk movement, by desk and jurisdiction?
• what evidence and launch gate applied to this run?
• what readiness issues blocked the previous attempt?

Answering before the run (readiness) and after the run (results) together gives you one continuous operating narrative.

What This Means For You¶

• One auditable substrate. APIs, the workbench, and analytics all read the same governed facts — no divergent copies of the truth.
• Explanations you can defend, because they are grounded in evidence and cite their sources.
• Analysis stays in the warehouse, close to governed data, which keeps your data movement controlled.

Where The Platform Is Today¶

Governed analytical products, comparison, and insight cases are available today. The BigQuery ML signal layer and the Vertex AI narrative and Q&A experience are on the roadmap — see Capability Maturity And Status.