B-TFAL examines a black-box AI decision and shows you exactly why it landed the way it did — and proves the reasoning is reproducible. Connect your model or upload your decision data to start. Or try a sample case to see how it works.
—
—
—
| Factor | Contribution | Value |
|---|
Same decision, same input, run five times. SHAP samples randomly each run so its explanation moves. B-TFAL returns the same answer every run by construction — the same property a regulator or auditor requires to verify a record.
—
B-TFAL audits a black-box AI decision. It tells you why the AI landed where it did, how confident we are in the reasoning, and whether the answer is reproducible — which is what compliance, audit, and regulator workflows actually need.
Every audit returns five panels in plain English: the outcome, how confident the model is, the reasoning trail, what would have to change for a different decision, and proof that the same input would return the same answer.
Toggle Internal view (operator-facing dashboard) or Applicant view (plain-language letter as the customer or patient would receive it) at the top of the page.
Turn on Design notes (top right) to see why each design choice was made — useful if you're evaluating the product surface as much as the engine.
—
—
—
—
The attribution and confidence values in this record are reproducible by deterministic construction. Re-running the engine on the unmodified input returns the same values. This certificate may be re-verified by the engine fingerprint above.