How should AI-assisted course authoring be governed?

From source material to draft

Workday's combined Learning and Sana proposition includes AI-native authoring. Source material provides the basis from which a draft is generated, while the model assembles an explanation or activity that still needs to be evaluated.

The generated output is therefore a new representation of the source, not the source itself. A reviewer needs to understand both the intended learning task and the underlying material to identify omissions, incorrect emphasis or unsupported statements.

Publication connects authoring to administration

Workday's administrator documentation separates Sana authoring and learner experience from Workday Learning's system-of-record responsibilities. Assignments, enrolments, learning records and reporting remain administrative concerns even when the content-production experience changes.

This explains why an authoring improvement is not only a matter of producing material faster. The approved version must fit the catalogue, the intended population and the evidence retained when learning is completed. Changes to published content can also create ongoing review and maintenance work.

Governance follows the lifecycle

NIST's generative-AI guidance addresses provenance, testing and human review. Applied here, those relationships connect approved inputs, the generated draft, the review decision and the version that reaches learners.

This article concerns ownership of the operating lifecycle. The separate generated-content evaluation article focuses on checking the output itself. The guidance below keeps that distinction while asking whether the complete production process improves, including the work needed to review, support and update a course.

This explanation draws on What changed with Workday Learning, powered by Sana? — Workday evidence, How should Workday Learning AI tutors be evaluated? — NIST evidence, Workday Learning remains the system of record in the integrated Sana model, Workday and Sana responsibilities are split across administration, authoring and learner experience, Workday L&D reports three years of Sana Learn use before integrated rollout. The evidence limits and publication history remain below.

What changed

Generative authoring is moving into mainstream learning platforms, reducing the technical barrier to producing and revising learning content.

Why it matters

Faster drafting can leave more content waiting for review. Plan enough time to check accuracy, assign ownership and maintain published courses.

Use cases & proof

Where this holds up in practice

These examples show how a specific claim is supported by documentation or an implementation. Read their limits before applying them to your own environment.

Proof pattern · Evidence C

Workday Learning remains the system of record while Sana changes authoring and learner experience

Scenario

An organisation adopts Workday Learning, powered by Sana either as a new Workday Learning deployment or by adding the Sana capability to an existing Workday Learning estate.

Architecture pattern

Workday Learning system of record, catalog, assignments, enrollments, reporting and permissions → integrated Sana provisioning → Sana Create authoring and Sana in-lesson learner experience → status and completion updates retained in Workday Learning.

Observed / documented outcome

Workday administrator documentation defines a split operating model between Workday Learning and Sana. Workday also reports that its own L&D team used Sana Learn for three years across a global workforce before planning the integrated Workday Learning, powered by Sana rollout.

What this proves

Workday Learning, powered by Sana is not simply a front-end replacement. Workday Learning remains the system of record for core learning administration while Sana changes authoring and in-lesson experience, creating real security, provisioning, content-lifecycle and operating-model decisions for adopters.

What it does not prove

It does not prove that migration is low effort, that every Workday tenant is entitled or ready to enable the capability, that Workday’s internal Sana Learn outcomes will transfer to other organisations, or that Workday’s internal team had completed production deployment of the fully integrated offering when its case was published.

Apply with these constraints

Confirm subscription entitlement, tenant activation, Workday/Sana tenant linkage, security domains, provisioning, creator roles, content migration and reuse expectations, reporting continuity, regulated or sovereign environment constraints, accessibility, audit requirements and measurable rollout outcomes.

Proof sourcesOfficial ArchitectureWorkday: Workday Learning remains the system of record in the integrated Sana model Official Content LifecycleWorkday: Workday and Sana responsibilities are split across administration, authoring and learner experience Named Internal Use CaseWorkday: Workday L&D reports three years of Sana Learn use before integrated rollout

Last checked 18 Sept 2026 · Reviewed before publication.

Implementation implications

Prerequisites

Approved source material, named content owner, review workflow, accessibility standard, versioning approach, retention policy and rules for restricted or regulated content.

Architecture implications

Generated content should remain traceable to source material, approval state and published version. Avoid treating the model output as the authoritative source.

Risks

Confident inaccuracies, stale policy content, inaccessible generated media, loss of provenance, uncontrolled duplication and apparent productivity gains that reappear as review or maintenance cost.

Testing implications

Compare generated output with source material, test accessibility, verify assessments and answers, track reviewer corrections, measure end-to-end production time and sample content again after later source changes.

Source register