Learning design

What is AI-native authoring?

An AI-native authoring tool is one built around AI from the start. You give it your source material and a short brief, it plans and writes a first version of the course, and your job moves from typing to checking and improving.

Updated

Learning design
Short answerIn a traditional authoring tool you start with a blank page and build every screen. Many of those tools now have AI buttons that write a paragraph or a quiz question. In an AI-native tool the order is reversed: the AI drafts the whole course from your sources, shows you its plan first, and you decide how closely it must stick to what your sources say.

AI-native and AI-assisted: the difference

AI-assistedAI-native
Where you startA blank course or a templateYour sources and a brief
What the AI doesHelps with one piece at a time: rewrite this text, suggest a question, make an imagePlans the course, writes every page and builds the questions
What you doBuild the structure yourself and ask for help inside itReview the plan, correct it, then edit the draft
Where the time goesBuildingChecking

Neither is wrong. If you enjoy crafting each screen and have the time, AI-assisted is fine. If you have forty procedures to turn into training before a go-live, starting from a draft matters more.

How AI-native authoring works, step by step

The steps below are how it works in Recourse. Other tools differ in the detail.

  1. Add your sourcesA video, a document, or both. In Recourse you can attach up to five. A source is optional, but a course about your own process should have one.
  2. Write a briefSay what learners should be able to do afterwards. You can add who the learners are, the objectives, a length (5, 15 or 30 minutes, or let the tool decide), a level, and whether you want a quiz at the end.
  3. Choose how closely to follow your sourcesVerbatim, Explain or Expand. This is the most important setting, and it has its own section below.
  4. Review the planBefore anything is written you get a storyboard: the lessons, in order, with their objectives. It sits beside a chat. The assistant may ask you a few questions about things your brief left open. You can edit the outline or ask for changes in plain words.
  5. BuildThe pages are written one by one and appear in the editor as they finish.
  6. EditChange anything by hand, or ask the assistant to change it. Add or swap blocks: text, images, video, accordions, tabs, flashcards, sorting, hotspots, timelines and questions.
  7. PublishShare by link or export SCORM for your LMS.

Verbatim, Explain or Expand: how closely to follow your sources

The biggest worry with AI-written training is that it says something your documents do not. This setting is the control for that. It decides how much the course may add to what you gave it.

SettingWhat it doesUse it when
VerbatimOnly what your sources say. Nothing is added.Policy, compliance, safety and legal content, where every statement must trace back to an approved document
ExplainYour sources, explained, with generic examples. This is the default.Most training. The facts are yours, and the course is allowed to make them easier to follow
ExpandYour sources plus general good practice, examples and scenarios.Topics where your source is thin and general knowledge helps, such as handling a difficult customer call

An example. Your source is a two page expenses policy.

  • Verbatim gives a course that states the limits and the approval steps as written, and asks questions on exactly those.
  • Explain gives the same rules, with a worked example of a claim to show how a rule applies.
  • Expand also adds general advice, such as keeping receipts as you go. That advice is sensible. It is also not in your policy, so someone should read it before it is published.

If you are not sure, start with Explain. Move to Verbatim when a reviewer needs to sign off every line.

What makes a good source

  • A recording of someone doing the task. For software training this is the richest source: it shows the screens and carries the expert's explanation, much of which was never written down. See from screen recording to interactive training.
  • The approved procedure or policy. The current version, not a draft.
  • An existing course. Old slides or a course from another tool can be the starting point for a new one.
  • A recorded training session. The questions people asked are a guide to what the course should cover.

A weak source gives a weak course. Three pages of accurate detail beat thirty pages of general background.

What to check before you publish

AI-native does not mean hands off. A person who knows the subject should check these things.

  1. Names and numbers. Menu paths, field names, limits, dates and codes should match your sources exactly.
  2. Questions. Each question should have one clearly right answer that the course itself teaches.
  3. Anything added. In Explain and Expand, read the examples. Are they true for your company?
  4. What is missing. Compare the plan with your source. Was a section left out?
  5. Tone and level. Does it sound right for the people taking it?

The storyboard step exists so you can catch most of this before the pages are written. Fixing a plan takes a minute. Fixing a finished course takes longer.

Authoring from Claude or ChatGPT

The same thing can be done from inside an AI assistant. Recourse connects to Claude and ChatGPT, so you can ask the assistant to plan and build a course, and the draft appears in your Recourse workspace to edit and publish. The steps are in how to create a course with Claude or ChatGPT.

Where AI-native authoring fits, and where it does not

Good fit: a lot of training to make from material that already exists, such as a system rollout, a product that changes every month, or a library of procedures nobody has turned into courses.

Weaker fit: a single flagship course with custom animation and a scripted story. A tool such as Articulate Storyline gives a designer finer control over every screen. See Recourse vs Storyline.

For software tasks, reading about clicks is not enough. Pair the course with a simulation so people practise the task as well as read about it.

How Recourse fits

Recourse is an AI-native authoring tool. It plans and writes a course from your recordings and documents, lets you choose Verbatim, Explain or Expand, and shows you the storyboard before it writes. It can also import a course from Articulate Rise (in beta, on the Pro plan) so that existing courses can be edited in the same place. It exports SCORM 1.2 on the Pro plan and is free to try for 15 days.

Questions people ask

What does AI-native authoring mean?
It means the authoring tool was designed around AI from the start. You give it source material and a brief, it plans and writes a first version of the whole course, and you review and edit. In an older tool with AI features added, you still build the course yourself and the AI helps with single pieces.
What is the difference between AI-native and AI-assisted authoring?
AI-assisted tools help inside a course you are building by hand: rewrite this text, suggest a question. AI-native tools produce the plan, the pages and the questions from your sources, and your time goes into checking and improving the draft.
Will an AI authoring tool make things up?
It can, which is why control over sources matters. Look for a setting that limits the course to what your documents say. In Recourse that setting is called Verbatim. Explain and Expand allow more to be added. Whatever the setting, a person who knows the subject should review the course before it is published.
What do Verbatim, Explain and Expand mean?
They are the three choices in Recourse for how closely a course follows your sources. Verbatim uses only what your sources say. Explain, the default, keeps to your sources and adds explanation and generic examples. Expand also adds general good practice, examples and scenarios.
What can I use as a source?
A video, such as a screen recording or a recorded training session, or a document, such as a procedure or a policy. In Recourse you can attach up to five sources to one course. You can also build a course from a brief alone, with no source.
Can I edit the course after the AI writes it?
Yes. The result is a normal editable course. You can change any text, move or replace blocks, add your own questions, or ask the assistant to make a change for you.
Does an AI-native tool replace an instructional designer?
No. It replaces the slow part, which is producing a first draft. Deciding what people need to be able to do, choosing the right sources, checking accuracy and judging whether the course will work are still a person's job.

Keep reading

Last reviewed 19 November 2026. Recourse publishes these guides for learning, customer education and change teams. Details drawn from public sources can change.