A Minimum AI Learning Entitlement: What You Can Build in One Canvas Week

A guest post on the HEPI blog, published on 9 October 2026, argues that every student should receive a minimum AI learning entitlement by the end of their first term. It draws on the HEPI and Kortext Student Generative AI Survey 2026, which, as the post reports it, found that 95% of undergraduates use AI in some way while only 48% feel their teaching staff are helping them build AI skills for their careers.

You may not be able to change institutional policy this term. You can decide what happens in one week of one Canvas course. This post sets out what is realistic in that time: access guidance, one discipline-specific verification task and a low-stakes practice activity. Together they make a modest, practical start on AI literacy for students.

What the proposal says, and what it does not prove

In the HEPI post, Roney Lima do Nascimento, an IB Diploma Programme mathematics teacher, describes an entitlement with four parts:

• Equal access to an institutionally supported AI tool, or an equivalent no-cost route.
• Discipline-specific teaching on verifying outputs, disclosure, data protection and the limits of the tool in that field.
• Low-stakes practice before assessments, whether AI is permitted or not.
• Clear, assessment-specific rules, with human review when misuse is suspected.

The proposed test of success is whether students can explain why they accepted, rejected or changed an AI output, rather than whether their work looks polished. That is a useful test for a lecturer, because you can build it straight into a task.

A balanced reading matters here. The post is advocacy, and its figures come from a single survey, with no detail given on sample or method. Two commenters raise fair objections: that an entitlement does not guarantee capability, and that learning the material without AI may be a better way to build the habit of checking. A week of activity will not settle either point. What it can do is give students a first structured chance to practise judgement, and give you evidence of what they can already do.

Part one: access guidance

Students cannot practise judgement with a tool they cannot reach, or are unsure they may use. Unequal access is also the first gap that opens up, so it makes sense to start there.

Create a single Canvas page near the top of your module, perhaps titled “Using AI in this module: the basics”. It needs only five short headings:

1. The tool: which tool your institution supports and how students get access.
2. The alternative: a no-cost route for students who cannot or prefer not to use it, where your institution has approved one.
3. The data: what students should never enter, such as personal data or other people’s work.
4. The rules: where AI is and is not permitted in each assessment, linked to the assessment briefs.
5. The contact: who to ask when something is unclear.

You are not setting policy here. You are pointing students to the guidance that already exists and saying plainly where it is silent. If your institution has no guidance yet, say so, and state what you expect in your own module.

This page belongs in the Prepare stage of PANDA. If the framework is new to you, start with my introduction to the PANDA framework, and see Preparing & Activating for how those early stages work.

Part two: one discipline-specific verification task

Generic AI training is easy to ignore. A task built on your own subject is harder to dismiss, because the errors matter to the student’s own work.

The idea is simple. Put a real question from your discipline to the tool your institution supports. Keep an answer that contains a genuine flaw, and note the flaw yourself before any student sees it. Then ask students to check the answer against your course sources and decide what to do with it.

Here is how that might look in different subjects:

• Engineering: a worked calculation with a sound method but a unit mistake.
• History: a summary that sounds convincing but cites a source that is not on the reading list.
• Business: an explanation of a framework that is accurate but too generic to apply to the case.

In Canvas, set this up as a short discussion or an ungraded assignment. Ask each student to post one verdict, accept, reject or modify, with two sentences of reasoning and one course source. Add a line asking them to say where they used a tool, which gives them early practice at disclosure.

Notice that the task does not ask students to use AI. It asks them to evaluate an output, which holds up whatever your assessment rules say.

Part three: a low-stakes practice activity

Many students meet AI in assessment conditions before they have practised anything. Low-stakes practice gives them a safe place to try, and it works whether AI is permitted in the assessment or not.

A simple version uses a Canvas practice quiz or an ungraded short task. Students attempt a question first. They then compare their answer with an AI output where AI is permitted, or with a model answer where it is not. Finally they write a short note on what they changed and why.

Three design choices keep it useful:

• No marks. Say plainly that completion is all that counts, so students try things without worrying.
• Quick feedback. Use automatic feedback on each answer, so students learn something while the attempt is fresh.
• A one-sentence disclosure. Ask students to practise stating what they used and how, in the form your assessments will require.

Keep a human in the loop for anything that carries marks. My earlier post on the dangers of using AI for summative assessments explains why this practice stays low-stakes.

Putting the week together

You do not need three new modules. You need three small pieces, spread across one week:

1. Early in the week: publish the access page and mention it in an announcement.
2. Mid-week: open the verification task.
3. End of the week: run the practice activity, then post a short summary of what the group found.

If you are unsure where these pieces sit in your course, my post on choosing a structure for your Canvas week compares three ways of laying them out.

What to look for afterwards

Treat the week as a small experiment, not a guarantee of capability. Read the verdicts and look at the reasons. Which students rejected the flawed answer, and which accepted it? Which questions about access did you receive? Those patterns show where students need more support, and they give you real evidence to bring to any institutional conversation.

Check your understanding

Your next step