Training that knows what you decided
Every LMS holds generic content that goes stale, because no LMS knows what your organisation actually did. This one is sitting on the record of it, and can draft the course from the decision and its outcome.
AI writes the course. It does not get to make things up.
Drafting is bounded to sources a human approved, and every claim is checked against that list on the server. The model saying it cited something is a hint, never the answer.
Drafting from approved material only.
Courses that come from your own record
A decision was made, an outcome was measured, and the lesson is sitting in the ledger. DecisionLearn drafts a course from it, so training reflects what the organisation actually did rather than a generic library.
It refuses before it offers
Not every decision should become a course. A policy check runs first and defaults to blocking, so a redundancy consultation cannot turn into a training module. A false positive costs one un-offered course; the other direction cannot be undone.
A full LMS, not a viewer
Authoring, assessments, assignments, certificates, SCORM, translations and reporting. It is the system of record for who was trained on what, and when.
A complete LMS underneath
The decision-seeded courses are the reason to choose it. These are the reasons it can be adopted.
AI that drafts from your sources, not from memory
The generator is never asked to know anything. It is handed material a human approved and told to teach from that and nothing else.
- Sections, then pages, then content, then images, in that order, because a picture chosen mid-page is chosen blind
- Illustration runs only once every page exists, so it can set the scene, show the mechanism, close on the outcome, and skip pages that need no picture
- Every claim carries the source it came from, and the citation is checked server-side against the approved list
- A claim citing something outside that list is demoted to an uncited flag, exactly like a claim with no citation at all
- Draft from a document you already have, from a decision and its outcome, or from a recurring pattern the platform noticed
- In the editor: draft one module as blocks, write assessment questions about content that already exists, or polish a block's wording without changing what it claims
Four gates, each with a named human
An AI-drafted course reaches a learner only after people sign off, and the gates are recorded in the audit ledger.
- Gate 0, proposal: accept, defer or reject the recurring decision pattern a course would be built from
- Gate 3, review: approve the drafted content, with the reviewer's identity recorded
- Gate 5, publish: set the audience, and whether the course is required or recommended
- Gate 9, retire: withdraw a course from circulation while preserving every completion record
- Uncited flags are graded, not absolute: past its risk tier's allowance a course cannot be approved until a human cites or cuts the claim
- A visible review-due date, so a course going stale is something somebody can see
Authoring a course people will finish
Most corporate courses fail on production quality, not subject matter. This is a real authoring studio.
- A block based course builder: text, images, video, embeds, files, and assessments in one flow
- Sections and pages, organised before a word is written, then written in a screen built for writing
- Draft a course from a document you already have, or from a decision and its outcome
- AI illustration and image generation for sections, inside the same budget and permissions as every other AI call
- Cover design with saved templates, one of which can be the tenant default
- Course templates and folders, with archiving rather than deletion
- Co-authors, in-course feedback, and learner notes
Assessment and completion
The part that turns a course into evidence somebody was actually trained.
- Assessments with attempts, grading, and pass thresholds
- Progress tracking per learner, per course, per assignment
- Certificates with custom designs and saved templates, issued on completion
- Certificate titles snapshotted at issue, so a later rename cannot rewrite history
- Completion gates, so a course can require what came before it
Assigning it to the right people
Assignment is where most LMS programmes collapse into a spreadsheet of who has not done it yet.
- Assign by audience dimension, drawn from the same canonical directory as everything else
- Assignment settings and due dates, so overdue has a definition
- Manager digests, so the chasing does not fall to one administrator
- Invites, offers and embeds for learners outside the tenant, without giving them an account
- An admin overview of what is assigned, started, overdue and complete
Standards, languages and reporting
The requirements that decide whether an LMS can be adopted at all.
- SCORM package import and export, so existing content is not stranded
- An LRS for statements, so completion data can leave for a wider system
- Course translations, with attempts recorded in the language they were taken in
- Reports with filters, and an export of the view you are actually looking at
- PII handling on learner data, with the platform's retention and legal hold applied
The loop nobody else can close
Decide, measure the outcome, teach what was learned, and record who was taught.
Decisions
A course can be drafted from a decision and cite its outcome.
Outcomes
What the decision produced becomes the lesson.
Compliance
Completion records become evidence in an attestation.
Agents
Reachable over MCP, under the same permissions a person has.
Careful about what it turns into a course
Before a decision is ever offered as training, it is checked for sensitivity and for whether it affects people. The check defaults to refusing, and the rules are evaluated when the question is asked rather than read from a flag that may have gone stale.