SAP SuccessFactors Learning
Learning agents can support enrolment, reminders and manager approval flows; Learning security remains role/domain based.
LT Horizon Research · 29 September 2026 · baseline v1.0
AI can now take action in learning systems. The control model matters as much as the capability.
This scan compares six enterprise learning platforms across action, permission boundaries, human confirmation, reversibility, auditability, source grounding and administrative control. It codes only what current official material supports.
Read the baseline alongside current release coverage.
The September comparison remains a dated evidence baseline. For release-specific October analysis, read the SAP Learning AI release guide and the Workday Learning powered by Sana assessment. Those articles retain their own evidence dates and limitations.
Check the actions AI can take.
Across the reviewed platforms, AI is no longer limited to search, summaries or recommendations. Official material now describes enrolment, reminders, approvals, catalogue changes, adaptive learning spaces, content generation and other workflow actions. But the public control evidence is uneven: some vendors document confirmation, permissions, reversibility or audit trails explicitly, while others describe agentic capability without exposing equivalent governance detail.
Implication: compare permissions and controls alongside the task an agent performs. Check who can trigger actions, what needs confirmation, what can be undone and what gets logged.
Action and governance baseline
Documented means reviewed official material explicitly supports the control or capability. Planned means it is presented as upcoming or staged. Not established means this scan did not find sufficient official evidence; it is not a claim that the control does not exist.
Scroll across the comparison to read every control. Each platform’s evidence note appears separately below.
| Platform | AI action | Permission boundary | Human confirmation | Reversibility | Auditability | Source grounding | Admin control |
|---|---|---|---|---|---|---|---|
| SAP SuccessFactors Learning | Documented | Documented | Documented | Not established | Not established | Documented | Documented |
| Workday Learning powered by Sana | Not established | Documented | Not established | Not established | Not established | Not established | Documented |
| Oracle Learning | Documented | Documented | Documented | Not established | Not established | Documented | Documented |
| Cornerstone Galaxy | Documented | Not established | Not established | Not established | Not established | Not established | Not established |
| Docebo | Planned | Documented | Documented | Not established | Documented | Documented | Documented |
| 360Learning | Documented | Not established | Documented | Documented | Not established | Documented | Not established |
Workday Learning powered by Sana
Workday documents domain-permission control, constrained access and admin enablement; reviewed material does not establish autonomous administrative action.
Oracle Learning
AI-generated audience criteria require specialist approval; agentic course creation stays draft until explicit activation and uses approved source material.
Cornerstone Galaxy
Official release material establishes agentic and administrative action, but this public-evidence scan did not find equivalent control detail.
Docebo
AgentHub is staged for wider availability; Docebo documents permission-aware action, audit trails, AI controls and confirmation in its MCP workflow.
360Learning
Action Mode executes catalogue changes, requires confirmation for higher-stakes operations and includes built-in revert; other modes expose source-backed answers.
Three governance patterns are already visible.
Check where people approve changes
Oracle keeps generated audience criteria and courses behind specialist approval or activation. 360Learning documents confirmation for higher-stakes actions and a revert function. Docebo’s MCP flow tells users to review requests before confirming.
Permissions must follow the action
Workday documents security-group and domain permissions around Sana access. Docebo states that agent actions remain permission-aware. SAP Learning retains its own role/domain security model. Agentic capability should not be evaluated separately from those permission boundaries.
Public governance evidence is uneven
Cornerstone’s public release material establishes agentic capability, but this scan did not find the same level of public detail for confirmation, reversibility or audit. “Not established” therefore identifies a question to ask the vendor before buying.
A control-evidence scan, not a safety score.
LT Horizon reviewed official vendor product, release, documentation and support material available on 29 September 2026 for SAP SuccessFactors Learning, Workday Learning powered by Sana, Oracle Learning, Cornerstone Galaxy, Docebo and 360Learning. Each platform was coded against seven governance dimensions. A cell was marked only when a reviewed official source explicitly supported it.
Limitations: this is not penetration testing, a legal assessment, hands-on product validation or a security certification. It does not test whether controls work as designed, whether audit logs are complete, or whether tenant configuration changes the outcome. Vendor marketing is not treated as proof of customer adoption or control effectiveness. “Not established” remains an evidence state.
Official evidence reviewed
Cite this as a dated governance baseline.
Suggested citation: LT Horizon (2026), AI action and governance in enterprise learning, September 2026, published 29 September 2026.
Canonical URL: https://lthorizon.com/research/ai-action-governance-2026
Reusable data: Download CSV.
Track when action boundaries change.
This baseline should be refreshed when a vendor adds a new executable action, introduces or removes confirmation, documents permissions or audit behaviour, changes availability, or publishes stronger control evidence.
Compare the AI capability map · Read current Intelligence · Research method and standards