AI agents in SAP SuccessFactors, governed like workers
Agents already validate payroll, screen applications and answer employee queries. Almost nobody records them where work is planned, costed and governed. This is a data model and validator that does, with every check mapped to the EU AI Act.
What you get
- An effective-dated
cust_DigitalWorkerMDF object design with owner, kill-switch owner, position, cost centre and expiring permissions - 15 CHK-AGT checks with console, JSON and HTML output and a CI-friendly exit code
- Decision scopes mapped to Annex III point 4 and the Article 5 prohibitions
- A 36-item audit workbook for HR, deployer and provider obligations
Why not create agents as employees?
It is the obvious shortcut and it breaks four things at once.
- Headcount and statutory reporting. Agents inflate FTE, gender pay gap denominators and workforce analytics.
- Payroll replication. EC Payroll and third-party payroll integrations pick up employment records. An agent with a job record is a payroll defect waiting to happen.
- Licensing. Many HCM licences count employee records.
- Meaning. Contract, notice and leave fields fill with junk.
The model instead
Agents get their own object that references the same foundation data, positions and people as employees, so they show up in org, cost and capacity views without polluting employment data.
Plus a composite cust_AgentPermission child with mandatory expiry, and references to legal entity and cost centre.
EU AI Act dates for HR, as amended
Regulation (EU) 2026/1744 (the Digital Omnibus on AI) was published on 24 July 2026 and entered into force on 27 July 2026.
The CHK-AGT check catalogue
| ID | Severity | Rule | Basis |
|---|---|---|---|
| CHK-AGT-00 | ERROR | Required fields, enumerations and dates valid | Data quality |
| CHK-AGT-01 | ERROR | Accountable owner is an active employee | Art. 26(2) |
| CHK-AGT-02 | ERROR | Kill-switch owner is an active employee | Art. 14(4)(e) |
| CHK-AGT-03 | ERROR | Employment decision scopes force high-risk classification | Annex III point 4; Art. 6(3) |
| CHK-AGT-04 | ERROR | High-risk agent has a human oversight mode | Art. 14; Art. 26(2) |
| CHK-AGT-05 | ERROR | High-risk agent keeps logs for at least six months | Art. 26(6) |
| CHK-AGT-06 | ERROR | Workers' representatives informed on or before go-live | Art. 26(7) |
| CHK-AGT-07 | ERROR | Permissions expire within 365 days of grant | Least privilege |
| CHK-AGT-08 | ERROR | No permission active past its expiry, including on suspended agents | Least privilege |
| CHK-AGT-09 | ERROR | Assigned position is active and staffed AGENT or HYBRID | Position control |
| CHK-AGT-10 | ERROR | Cost centre exists in foundation data | Cost allocation |
| CHK-AGT-11 | WARNING | Periodic review not overdue | Governance |
| CHK-AGT-12 | ERROR | No overlapping effective-dated records | Data integrity |
| CHK-AGT-13 | WARNING | Owner not accountable for more agents than they can oversee | Oversight capacity |
| CHK-AGT-14 | ERROR | No active prohibited practice | Art. 5(1)(f) |
"High-risk" means declared high-risk or carrying an employment decision scope, so a misclassification cannot hide the oversight, logging and consultation gaps behind it.
Run it in a minute
git clone https://github.com/SahirVhora/sf-agent-workforce cd sf-agent-workforce python -m agent_workforce.validate \ examples/demo_with_findings.json \ --as-of 2026-10-10 --html report.html python -m pytest -q
Python 3.10+, standard library only. Exit code 0 means clean, 1 means errors, 2 means unreadable input.
Free download: EU AI Act HR audit checklist
A working Excel checklist for HR and HRIS teams: 36 requirements with article references and applicable dates, a triage of 12 common HR AI use cases, and a summary that tracks completion by area.
Questions
Should AI agents be created as employees in SuccessFactors?
No. Creating agents as users with job records inflates headcount and statutory reporting, risks payroll replication, can consume per-employee licences and fills employment fields with meaningless data. Model them as their own effective-dated MDF object that references positions, cost centres and accountable people.
When do the EU AI Act high-risk rules apply to HR systems?
From 2 December 2027 for stand-alone Annex III systems, including employment and workers' management, under Regulation (EU) 2026/1744. Prohibitions have applied since 2 February 2025 and Article 50 transparency duties since 2 August 2026.
Does the EU AI Act apply to UK employers?
It can. Article 2 brings providers and deployers outside the EU into scope where they place AI systems on the EU market or where the system's output is used in the EU, for example an HR system used for EU-based workers or candidates.
Which HR uses of AI are high-risk?
Annex III point 4 covers recruitment and selection (targeted job adverts, filtering applications, evaluating candidates) and decisions affecting terms of work, promotion, termination, task allocation based on individual behaviour or traits, and monitoring or evaluating performance and behaviour.
Part of the SF Compass transformation assurance suite.