Decision Record Builder · governed release 02

Preserve the reasoning—not just the answer.

Create a versioned record of one bounded AI, workflow, pilot, or vendor decision: who owns it, which alternatives were considered, what evidence supports it, who may be affected, what would stop it, and when it must be reviewed again.

Go directly to the builder ↓

Private working space

Build the record

Nothing entered here is transmitted to Aloha AI. Avoid confidential, privileged, personal, or regulated information unless your own authority and handling requirements are resolved.

Restoring your local record… Jump to download and reset controls ↓

0/11core fields recorded0 of 11Completeness is not quality, approval, safety, or compliance.

01Decision and authority

Record title *Name this bounded decision.VersionUse a new version when reasoning or evidence changes.Current statusFor example: under review, paused, rejected, bounded test, or approved by the named owner.Accountable owner *Name the person with authority for the decision and its consequences.Required reviewersName capable reviewers and the questions each must examine.Decision under consideration *State the exact choice—without assuming AI is the answer.

02Context and alternatives

Purpose and intended outcome *What needs to become true, and for whom?Permitted scope and explicit exclusions *Define users, setting, data, duration, and what this decision does not cover.Alternatives considered *Include the current process, redesign, human-led work, delay, and no-action where relevant.

03Evidence and affected people

Evidence register *List each dated source, what proposition it supports, and its limitations. Do not paste confidential material.Affected people and participation *Who may benefit, bear error, repair failure, need access, or have relevant knowledge? Record how their views entered the decision.Expected benefitsDescribe bounded, observable benefits—not promotional claims.Potential negative impactsInclude foreseeable errors, unequal burdens, rights or livelihood effects, and downstream repair work.Data, permission, and provenanceIdentify data classes, sources, authority, quality limits, retention, and prohibited inputs.

04Test, stop, and review

Measures and evidence needed to proceedWhat would be observed, compared, and independently reviewed? State the baseline.Stop conditions *What event, threshold, missing permission, failed safeguard, or unexpected impact requires pause or termination?Fallback and repair path *How will work continue safely, and who repairs harm or incorrect output?Next review dateUse YYYY-MM-DD when a date is known.Earlier review triggers *List changes in use, people, data, vendor, model, law, performance, incidents, or context that reopen the decision.Unresolved questions and dissentPreserve what is unknown, disputed, or awaiting specialist review.

Primary framework desk

Why these fields exist

NIST organizes AI risk work around govern, map, measure, and manage, emphasizes defined roles and lifecycle documentation, and states that its actions are not a universal checklist. GAO organizes accountability around governance, data, performance, and monitoring. This builder translates those framework themes into a portable working record; it does not certify framework conformance.