Svenska Dokumentmallar
AI Agent Governance – Mandate, Human Attestation, and Rollback | Word/PDF/Excel
AI Agent Governance – Mandate, Human Attestation, and Rollback | Word/PDF/Excel
- File format
- DOCX, PDF, XLSX
- Document language
- Five work templates in Swedish; Swedish User Guide; English User Guide. Excel in Swedish.
- Number of files
- 7
Digital download. No physical product is shipped.
Couldn't load pickup availability
About this template
AI agent governance – mandate, human review and rollback helps Swedish companies document what autonomous AI agents are allowed to do, which actions must be approved by a human, and how the agent is tested, stopped and reviewed. The package does not provide technical API restrictions, legal authority or automatic compliance.
Quick facts: Five Swedish working documents, one Swedish guide and one English User Guide (seven standalone documents), each in editable DOCX and searchable/printable PDF, plus one Swedish Excel workbook (XLSX) with nine tabs. 15 customer files in a digital ZIP package. The PDF files are not interactively fillable.
Package contents
Five Swedish working templates, one Swedish guide and one English user guide, all in Word and PDF. An Excel workbook with nine meaningful tabs. A total of 7 standalone documents and 15 files.
What the files contain
Agent and tool inventory, risk/data protection screening, mandate decision, action limits and human review, pilot testing with emergency stop and rollback, plus incident/change decision. Excel tracks status and evidence without making decisions or executing actions on the user's behalf.
When is the package suitable?
Example: an e-commerce company wants to let an AI agent suggest customer refunds but require human review before any payment is made; or an IT manager wants to show that the agent is stopped from making unauthorized tool calls. The package is not suitable as actual agent software, security control, legal authority, completed GDPR impact assessment or AI Act certification.
Documents and usage
| Document | Usage | Language | Format |
|---|---|---|---|
| 01 Agent Inventory | Identify systems, tools, personal data and risk classification | SV | DOCX / PDF |
| 02 Mandate Decision | Establish permitted tools, prohibited actions and responsibilities | SV | DOCX / PDF |
| 03 Review Matrix | Distinguish between suggestions, human approval and actual execution | SV | DOCX / PDF |
| 04 Pilot Test and Rollback | Test limits, denied action, emergency stop and rollback | SV | DOCX / PDF |
| 05 Deviation and Review | Document incident, correction, retesting and new decision | SV | DOCX / PDF |
| 06 Swedish Guide | Finalization, limitations and references | SV | DOCX / PDF |
| 07 English User Guide | English instructions for the Swedish documents | EN | DOCX / PDF |
| 08 Excel Workbook | Nine tabs for status, documentation and audit trail | SV | XLSX |
Excel in practice
Tabs: Overview, Agent Register, Mandate, Action Log, Pilot Tests, Risks and Incidents, Changes, Decisions, Read Me. Formulas count documented decisions, open incidents and tests, and flag, among other things, “EXECUTED WITHOUT REVIEW” when an executed action lacks recorded human approval. These features have been tested with fictional examples; the customer workbook contains no test data. Excel cannot block actual API calls or determine legal compliance.
Workflow
- Inventory the agent, tools, data sources, roles and intended action.
- Assess risk, prohibited AI practices, relevant AI Act classification and data protection.
- Establish internal mandate, human decision limits and technical restrictions.
- Test concrete scenarios, including lack of review and emergency stop.
- Document observed results, evidence references, deviations and manual decisions.
- Review if the model, tool or usage changes.
Legal prerequisites
The AI Act (EU) 2024/1689, consolidated 27 July 2026 has separate rules for AI competence (Article 4), prohibited uses (Article 5) and applicable high-risk AI systems. Articles 14 and 26 on specific high-risk requirements do not automatically apply to all AI agents; check actual classification, role and application date. For personal data, the relevant provisions of GDPR apply, including Articles 5, 22, 25, 28, 32 and 35. IMY's guidance on automated decisions highlights specific limitations for decisions that significantly affect individuals. The company's own monetary limits are not a general statutory AI agent limit.
Other documents that may be more suitable
AI Governance refers to overall AI governance and system registers. AI Policy provides general internal rules. IT System Decommissioning is used when an IT system is being retired. This product specifically governs the agent's actions during operation.
Frequently Asked Questions
Can the agent sign contracts using the template?
No. The mandate is an internal decision-making basis, not a power of attorney or a technical assignment of authority.
Are all AI agents high-risk?
No. It depends on the system, use, role and application date.
Are the PDF files fillable?
They are searchable and printable but lack interactive form fields. Edit the DOCX files.
Are the working templates in English?
No. The five working templates are in Swedish; the English file is a user guide.
Does Excel provide automatic legal or technical control?
No. Statuses and formulas are administrative support. The actual agent testing and legal assessment must be done separately.
Sources and transparency
Version 1.0, AI-assisted legal source review 10 November 2026. No external legal, lawyer or authority review is claimed. The product is not available for purchase in DRAFT, and the ZIP file has not been linked to Shopify Digital Products; digital delivery is intended only after later activation.
