Companies using AI to decide on credit, employment, insurance, pricing or access to services.
Who uses it
IT, data, compliance, audit, PMO, security and asset owners
Usual owner
Data governance; legal
Formats
DOCX, PDF
Languages
Español, English
Approximate length
6 pages
Length
Standard. A complete policy, procedure, plan or guide, with roles, deadlines and evidence.
Document type
Registers, datasheets and records. Tables and forms to keep track of what exists and what was decided.
Document code
APG-INV-003. APG is the AI Policy Generator collection, INV the area (Inventory) and the number identifies the document. Use it to cite the document in your document system.
Preview
Excerpt generated from the document: the "How to use this document" section, the title and the first two sections, plus the full table of contents. The rest of the text is not shown.
APG-INV-003 · Version 1.1.0 · sources as of November 4, 2026
How to use this document
What it is for. This is a standard and template to log significant decisions about people made fully or partly with AI or other automated processes: with what data, with which system version, what output it gave, what the human reviewer did, what was communicated to the individual and how their requests for explanation or review were resolved. It is used to explain decisions, handle complaints, demonstrate human oversight and audit the system.
Who should own it. Usually the Data protection officer (or the country's equivalent role) together with Legal, and the system owner of each decision process. Enter the function in [RESPONSIBLE FUNCTION] and the person in [OWNER].
Decisions to take first. Decide which decision processes are in scope (Section 2), whether the log is generated automatically by each system or centralized (Section 6), which retention periods apply to each process (Section 10) and who may access the log (Section 11).
Fields to complete. Complete one process sheet (Section 4) for each decision process and configure the Section 5 fields in the system that generates the log. The "Example" columns show a complete fictitious credit decision.
Country variants. This document has no country variants. Individuals' rights are described in the "AI User Rights Notice" and their handling in the "AI Transparency and User Explanation Policy" and the "Automated Decision Review Procedure". Retention periods vary by country and industry: the table in Section 10 includes rows for the European Union, the United States (Colorado, California and federal employment) and Colombia (health care); keep those that apply and complete fields 4.9 and 4.10.
Review. Ask privacy and legal to review Sections 2, 7 and 10, and security to validate Section 11. If the log contains personal data, include it in your record of processing activities.
Approval. Have it approved by the [AI COMMITTEE] or management, record the approval in "Version history" and communicate the standard to system owners and the technical teams that must implement it.
Automated Decision Log Standard
Field
Value
Organization
[LEGAL ENTITY NAME]
Owning function
[RESPONSIBLE FUNCTION]
Log owner
[OWNER]
Approved by
[APPROVED BY]
Effective date
[DATE]
Review frequency
[REVIEW FREQUENCY]
Log location
[COMPLETE: system or repository]
Privacy contact
[PRIVACY EMAIL]
1. Purpose
1.1 [ORGANIZATION NAME] keeps a complete, tamper-evident and searchable log of significant individual decisions made fully or partly by AI systems or automated processes.
1.2 The log makes it possible to:
a) explain to each affected person, specifically, how the decision affecting them was reached; b) handle requests for human intervention, review, objection or appeal; c) demonstrate that the planned human oversight exists and works; d) reconstruct and audit decisions after a complaint, an incident or an inspection; e) detect errors, drift or discriminatory effects across all decisions.
2. Scope
2.1 The log covers decisions about identified or identifiable individuals that:
a) are made without meaningful human involvement; or b) are made with human involvement, but where the output of an AI system or an automated score, whether in-house or from a third party, is decisive or has a significant influence;
and that produce legal effects or similarly significant effects on them, for example in:
granting, terms or limits of credit, insurance or financial services;
hiring, recruitment, task allocation, evaluation, promotion, discipline or dismissal of workers or candidates, including management by digital platforms;
personalized prices or terms with significant effect;
access to essential services, housing, education, health or benefits;
suspension, restriction or closure of accounts;
fraud detection that blocks transactions or services.
2.2 [OPTIONAL] The organization also logs, with the minimum fields, lower-impact automated recommendations used in sensitive processes, even if they do not meet the condition in 2.1.
2.3 Each decision process in scope is identified in the Section 4 list and in the AI system inventory.
Full table of contents of the document:
1. Purpose
2. Scope
3. Log structure
4. Decision process sheet
5. Individual decision log
6. Log generation rules
7. Explanations and individuals' requests
8. Human oversight
9. Monitoring across decisions
10. Retention
11. Security and access
12. Standard review
Implementation aids
Related regulatory references
Version history
Related rules
Rules in the registry whose obligations relate to this base document, based on our reading of the sources. It does not mean they apply to you.
Explain the logic of automated decisionsRes. AAIP 4/2019, criterio sobre decisiones automatizadas (Ley 25.326, art. 15 inc. 1)In force(supporting)Verified in the official source on November 4, 2026
AI does not replace clinical judgmentRes. MinSalud 1644 de 2026, art. 23.1 y 23.4In force(supporting)Verified in the official source on November 2, 2026
Traceability and audit of AI outputsRes. MinSalud 1644 de 2026, art. 23.2 y 23.6In force(related)Verified in the official source on November 2, 2026
Providers of Annex III high-risk systemsReglamento (UE) 2024/1689, arts. 6(3)-(4), 8-17, 43, 47-49, 72-73Adopted, applies later(supporting)Verified in the official source on November 2, 2026
Safeguards against solely automated decisionsRGPD, arts. 13.2.f), 14.2.g), 15.1.h) y 22; STJUE C-634/21, SCHUFA, 7-dic-2023; STJUE C-203/22, Dun & Bradstreet Austria, 27-feb-2025In force(supporting)Verified in the official source on October 31, 2026
Monitor AI selection tools for adverse impact42 U.S.C. § 2000e-2(k); 29 CFR 1607.4(D) (regla de los cuatro quintos)In force(supporting)Pending confirmation in the official source (last checked: November 4, 2026)
Keep compliance records for three yearsSB 26-189, conservación de registrosAdopted, applies later(related)Verified in the official source on November 2, 2026
Reference template for the impact-ratio bias audit report (in New York City an independent auditor performs it), AI hiring policy, AI-use notice to candidates, human oversight and a decision log, related to New York, Illinois, California and Colorado employment rules.
Model risk with inventory and materiality, adverse-decision explanation letters, human review, a decision log, technical model documentation and the EU AI Act Art. 26 checklist for high-risk systems.
Every sellable template in the catalog in a single pack: base policies, procedures, registers, notices and matrices to adapt with your own team, at a flat price below the sum of the individual documents.