What is course coach and how do I use it?
What Course Coach checks, how to run a review, how the suggestions and one-click fixes work, and what the Insights tab shows.
Course Coach reviews a course you have built and gives you specific, actionable suggestions across accessibility, learning design, clarity, engagement and content integrity. Most suggestions come with a one-click fix the agent applies for you. A separate tab shows where your real learners are getting stuck.
There is no score and no grade. You get suggestions you can act on, each one naming the problem, the screen it is on and why it matters.
Where do I find it?
In the agent panel in the builder, which has three tabs:
- Chat is where you talk to the agent directly and ask it to make changes.
- Review holds the suggestions from a review of your course, under the heading Improve this course.
- Insights holds what your learner data shows.
Once a review has run, the Review tab carries a count of open suggestions, so you can see there is something worth looking at without opening it.
How do I run a review?
Open the Review tab and run the review from there. On a course you have just built there is nothing in the tab until you do, so this is the first step rather than something waiting for you.
At the top of the results you get a resolved counter and a progress bar, reading something like "0 of 5 resolved", which fills as you work through the list.
What does it look at?
Two kinds of checking, combined into one list of suggestions.
Objective checks. Mechanical, repeatable, and the same every time for the same course:
- Accessibility. Images and slideshow slides with no alt text. Link text that does not say where it goes, such as "click here" or a bare URL. A screen carrying real body text with no heading on it.
- Unfinished content. An image, video, audio, document or embed block that is still an empty placeholder.
- Question problems. A question with no text, no answer options, or no correct answer marked. A short answer question that accepts no answers. A multiple choice question with only two options, where a learner has a coin flip chance of guessing. A matching question without enough pairs.
- A predictable answer key. If the correct answer sits in the same position across most of your multiple choice questions, learners can game the quiz without knowing the content.
- Clarity. Reading level well above general audience level. Acronyms used repeatedly but never spelled out.
- Learning design. A screen that is a wall of text with nothing to look at or do. A course of several screens with no knowledge checks anywhere.
- Content that has drifted. A screen whose text was edited after its narration was generated, so learners listening hear the old wording. A course that has not been touched in more than six months. External links that are genuinely broken.
Expert judgement. Four specialists read the whole course, which is how it catches problems no single screen reveals:
- Alignment. Does the course assess anything it never actually teaches? Is anything taught and then never reinforced? This is the highest value catch, because a question testing untaught material is invisible until a learner fails it.
- Structure and cognitive load. Is a concept used before it is introduced? Does each step rest on the last, or does the course front-load complexity and jump around? Is practice spaced through the course or crammed at the end?
- Assessment quality. Do questions test understanding and application rather than recall? Do the wrong answers reflect real misconceptions? Does the feedback explain why an answer is right or wrong, rather than just saying "Incorrect"?
- Clarity, voice and engagement. Does the course set up why this matters at the start and consolidate at the end? Is it tied to the learner's real world? Jargon, inconsistent terminology, and stretches of explanation where an example would land better.
The results are deliberately short. Suggestions are checked before you see them and the weak ones are dropped, so you get the few worth your time rather than everything that could conceivably be said.
How are suggestions ranked?
Three levels, shown as filter pills above the list with a count on each: Correct, Improve and Refine, plus All. Each suggestion carries a coloured dot for its level, so you can see the shape of the list at a glance, and you can filter to one level when you want to work through the most serious ones first.
One rule underpins the whole thing and is worth knowing, because it is what makes the panel trustworthy. Only the objective checks can tell you something definitely affects learners, and those are the ones marked Correct. Anything based on judgement, or on learner data, never claims that certainty and appears as Improve or Refine. So when Course Coach marks something Correct, you can fix it without second-guessing.
What does a suggestion tell me?
The list shows short, specific titles, such as "Assessed temperature threshold not taught" or "Missing feedback on screen 7". Click one to open it, and it shows:
- What the problem is.
- Why it matters, in a sentence, usually naming the principle behind it, so you learn something rather than just clearing a list.
- The evidence, a specific quote or observation, so you can see what it is reacting to.
- Where it is. Suggestions are labelled with the screen number and title, the question number for a quiz question, or "whole course" for something that is not about one screen.
Only one suggestion is open at a time, so opening the next one closes the last. Clicking a suggestion also takes you to the spot it refers to, so you are looking at the thing being discussed.
How do the fixes work?
Most suggestions carry a fix button, labelled for what it will actually do, and the agent applies it with the same tools it uses when you ask it directly. The alternative is always Dismiss. Resolved suggestions animate out of the list and the counter and progress bar update.
Two things worth knowing:
- Fixes are verified rather than assumed. For the objective checks, Coassemble re-runs the check after the fix and only marks it resolved if the problem has actually gone. If the agent did not change anything, or asked you a question instead, the suggestion stays open and says so.
- Some suggestions have no fix, on purpose. Narration that has drifted out of sync, a broken external link, and a course that has gone stale are all things where guessing would be worse than telling you. Those are surfaced for you to act on.
What is the Insights tab?
The same idea, using your real learner data instead of reading the course. It surfaces things like a screen where a large share of learners stall and do not go further, a text-heavy screen people are moving past too fast to have read, a question most learners get wrong on their first attempt, a screen learners keep coming back to, low completion across the course, and a course taking much longer than the time you advertised.
Nothing appears until a few learners have been through the course. Below that the tab says "Too quiet to tell yet" with a Refresh button, rather than reading meaning into one or two people. This is why a brand new course shows nothing here.
Is Course Coach available on my plan?
Course Coach is not included on every plan. If it is not included on yours, you will not see the Review and Insights tabs, and everything else about building and editing courses works normally.