AI in enterprise learning: the capability map is already fragmenting.
Six major learning platforms are moving beyond generic “AI features”, but they are doing so at different layers: tutoring, adaptive delivery, authoring, search, administration, analytics and skills. This scan codes only what official material actually 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.
“Has AI” is no longer a useful buying criterion.
The reviewed platforms use AI for different tasks. Workday is combining AI tutoring, authoring, search and learning operations with HCM and skills context. Oracle is moving further into adaptive agentic course delivery and course design. Cornerstone is pairing a course assistant with an adaptive learning agent. 360Learning splits its companion into search, coaching, action and insights modes. Docebo describes agents that use enterprise knowledge; several of those features are planned or being released in stages. SAP has released conversational and compliance-oriented Learning capability while its next SuccessFactors release remains under active watch.
Implication: compare the tasks AI performs, the data it uses, its permissions and how outputs are reviewed.
Capability map
Documented means the reviewed official material explicitly describes the capability. Planned means the vendor explicitly presents it as upcoming or staged. Not established means this review did not find sufficient official evidence to mark the capability present; it does not mean the platform cannot do it.
| Platform | Learner Q&A / tutor | Adaptive delivery | AI authoring | AI search | Admin / action | AI insights | Skills context | Evidence status |
|---|---|---|---|---|---|---|---|---|
| SAP SuccessFactors Learning | Documented | Not established | Not established | Documented | Documented | Documented | Documented | Released capability plus active roadmap/release watch |
| Workday Learning powered by Sana | Documented | Not established | Documented | Documented | Documented | Not established | Documented | Generally available |
| Oracle Learning | Documented | Documented | Documented | Not established | Not established | Not established | Documented | 26C released; 26D capability documented |
| Cornerstone Galaxy | Documented | Documented | Documented | Not established | Documented | Not established | Documented | Spring 2026 release |
| Docebo | Planned | Not established | Planned | Planned | Planned | Planned | Planned | AgentHub / Enterprise Knowledge announced; availability still being staged |
| 360Learning | Documented | Not established | Not established | Documented | Documented | Documented | Documented | AI Companion modes introduced through 2026 |
Three patterns matter more than the feature count.
AI is moving into execution
Compliance nudges, assignment automation, administrative actions and agent workflows mean the governance question is shifting from “what can it answer?” to “what can it do?”.
Learning and skills are converging
Workday, Oracle, Cornerstone, Docebo and 360Learning increasingly connect AI learning experiences to skills data or development context. Weak skills data can therefore become a learning-quality problem.
Availability needs its own field
A released assistant, a documented future update and an announced agent platform are not equivalent. Procurement comparisons should keep capability and availability separate.
A reproducible evidence-coding rule, not a vendor score.
LT Horizon reviewed official product, release, documentation and support material available on 28 September 2026 for SAP SuccessFactors Learning, Workday Learning powered by Sana, Oracle Learning, Cornerstone Galaxy, Docebo and 360Learning. Each platform was coded against seven functional dimensions. A cell is marked only when the source explicitly supports that function. Marketing language was not converted into an adoption, quality or effectiveness claim.
Limitations: this is a public-evidence capability scan, not hands-on product testing. It does not measure answer quality, implementation effort, price, security posture, customer adoption, model performance or learner outcomes. “Not established” is an evidence state, not a negative product judgement. Vendor documentation changes quickly; this map should be treated as a dated baseline.
Official evidence reviewed
Cite the dated baseline, not a floating market claim.
Suggested citation: LT Horizon (2026), AI in enterprise learning: capability map, September 2026, published 28 September 2026, reviewed baseline date 28 September 2026.
Canonical URL: https://lthorizon.com/research/ai-learning-capability-map-2026
Reusable data: Download CSV. The dataset preserves the same evidence-state language used in the report: Documented, Planned and Not established.
Please retain the report date when quoting or reusing the map so later product changes are not mistaken for the September 2026 baseline.
Turn the map into evidence of change.
This baseline becomes more useful when it is repeated. LT Horizon will use later release evidence to record which planned capabilities became available, which changed scope, and where vendors expanded from assistance into action.
Read current Intelligence · See what LT Horizon is watching · Research method and standards