How the Data Suitability Recipe Works

The Data Suitability recipe builds an AI Use Case and captures its evidence. Its side panel sums the job up in one sentence: "Turn a model's input data into a governance record an auditor can check: bind the assets, fix a baseline, and capture a Data Suitability evidence record." This page covers the mechanics that span the whole flow. The steps themselves are detailed on Define the Use Case and Capture the Evidence, and the record they build on AI Use Cases.

The Record List

Opening the recipe from the Recipes page lands on a list of the AI Use Cases you can see, headed Data Suitability. Each row shows the record's name and description, how many datastores, containers, and checks it binds, and its certification badge. From here you can:

Click Start a new record to begin a fresh use case at the first step.
Open a row's actions menu and click Open in Recipe to continue that record's saved run.
Click the row itself to open the AI Use Case page, where you review its evidence rather than edit it.

An empty list reads "No model risk records yet. Start a new record to bind a model's inputs and emit its first attestation." This list is also where AI Use Cases live in the application: there is no separate AI Use Cases page in the left side panel. All AI Use Cases, at the top of the recipe's side panel and on the AI Use Case page, brings you back here.

Phases and Steps

The recipe runs in two phases. The side panel lists them, and the header above the step content names the current phase and its caption, followed by the numbered steps of that phase.

Phase	What it does	Steps
Define the Use Case	What the model is, what it reads, and the standard its inputs are held to.	Model, Inputs, Standard
Capture the Evidence	Fix a baseline, record the certified evidence, and retain it for governance review.	Baseline, Evidence, Complete

Back is available on every step after the first, and every step you have already reached stays clickable. Steps ahead of you lock while the current step's requirement is unmet: a model name and risk tier, at least one datastore and one container, a captured baseline, and a generated attestation. The Standard step has no requirement, since both the check set and the expiry schedule are optional.

Each Next on the first phase saves that step to the AI Use Case before moving on. The first Next creates the record; the later ones update it. The recipe only advances when the save succeeds, so what you see on the next step is always what is stored.

Progress Lives on the AI Use Case

This recipe keeps no progress in your browser. Each run is a recipe run saved on the AI Use Case, holding the step you are on and a small amount of run state: when the baseline was captured, which attestation the run produced, and a fingerprint of the definition that attestation was generated from. Everything else, such as the name, bindings, check set, and schedule, is read back from the record itself, so a saved run can never show you configuration someone else has since changed.

Progress is saved automatically a moment after each change, and the side panel footer shows Writing to AI Use Case #N with the record it writes to. When a save fails, the footer says so, and the recipe carries on; the next change tries again.

Because the run is on the record, the Recipes page card cannot tell whether you have something in progress. It always reads Get Started. To continue, use Open in Recipe on the record list or Continue Recipe on the AI Use Case page. See Entry Points and Resuming.

What the Run Produces

Output	Where it lives	Notes
AI Use Case	The Data Suitability record list and its own page	Created on the first Next, then updated step by step. Carries the name, risk tier, description, bound datastores, containers, and checks, the expiry schedule, and the certification state.
Recipe run	The AI Use Case, under Latest Recipe Run	The saved progress. Marked complete when you reach the Complete step.
Baseline	The AI Use Case	Today's quality scores, profile identifiers, and row counts for every bound container, captured on the Baseline step. Starts a certification cycle.
Attestation	The AI Use Case, on its Attestations tab	The immutable evidence record generated on the Evidence step. Regenerating adds a new one that supersedes the previous.
Exports	Your download folder	The attestation as JSON, the authoritative evidence record, or as a self-contained HTML report.

The recipe never reads or writes your data. It reads the platform's metadata about the bound assets and writes only the AI Use Case and its attestations. See What is recorded, and what is not.

Certification Comes Before Evidence

Certification is the governance approval; the attestation is the immutable evidence of that approved state. The Evidence step therefore refuses to generate an attestation until the AI Use Case is Certified, so that a Draft or In Review decision is never preserved in the artifact. Getting there is a two-person job by design:

An author with the Author team permission on every bound datastore builds the record and clicks Submit for Review.
A Manager with the Editor team permission on every bound datastore, or an Admin, reviews it and clicks Certify, optionally setting a validity date.

Qualytics does not send an approval notification, so the author shares the page with the approver. Both actions are available on the Evidence step and on the AI Use Case page. See Certify an AI Use Case.

Changing the Definition Later

The definition can be edited after evidence exists, from the recipe or from Edit in Recipe on the AI Use Case page. Two consequences follow:

The attestation the run produced is tied to the definition it was generated from. When the definition changes, the recipe no longer treats that attestation as current for the run, and the Evidence step asks you to generate a new one.
Changing which containers are bound after a baseline was captured blocks generation with "Cannot generate an attestation because the asset scope changed after baseline capture" until you capture a new baseline, which returns the record to Draft and starts a new certification cycle.

Existing attestations are never changed by any of this. They stay readable on the Attestations tab, each marked current or superseded.

No AI, No Working Assets

Unlike Entity Resolution and Data Reconciliation, this recipe offers no AgentQ suggestions and no How It Works explainer panel, and its side panel has no Working Assets drawer. The only asset it touches is the AI Use Case named in the side panel footer.