Richard Gascoigne: On the AI Proposal Worth Funding

Richard Gascoigne: On the AI Proposal Worth Funding

What 71 delegates at CanvasCon Europe told us about AI in the VLE.

What 71 delegates at CanvasCon Europe told us about AI in the VLE.

Written by

Richard Gascoigne

Published on

Copy link

1. The university AI proposal worth funding 

We had a very successful trip to CanvasCon Europe, the annual event for users of the Canvas (Instructure) VLE. Will Moindrot and Laura Blundell from the University of Liverpool, who ran a Plato pilot with us, presented their analysis of the project. They closed the session by polling the room.

Picture a Tuesday morning in week six. A student opens Canvas to check what they need to submit. The module page points them to a handbook; the handbook is vague, and the person who can settle it is teaching until noon. The student sends an email. By lunchtime, the lecturer has three versions of the same question to answer. 

Before anyone buys an AI tool, I’d sort out the handbook. Then I’d ask whether students can find and use the corrected guidance when they need it. That’s where a case for investment might begin. 

That Mentimeter poll drew 71 participants. The first question asked where their institution is with AI embedded in the VLE. 

Only 22% described active, secure AI pilots embedded in Canvas courses. The rest split almost evenly three ways: a defensive focus on assessment and integrity, grass-roots experiments with no central coordination, or working groups and governance reviews. Three quite different kinds of institution, and none of them has an embedded pilot yet. 

Poll result, barriers: cost, budget and funding 42%; trust, fear and resistance to change 28%; security, data and governance 25%; staff time and training 19%; impact on learning and integrity 12%. 64 respondents.

When the session asked about barriers to introducing a tool like Plato, money came first. “Cost” was the most frequent single word in the responses, appearing 17 times. Count “budget”, “money” and “funding” with it and about four in ten respondents named money. Trust and security came up too, as did staff time and training. 

Poll result, blueprint: training and support for teaching staff; student involvement and consent; a clear educational purpose. 49 respondents.

Asked what they’d put in a blueprint, respondents mentioned support for teachers most often, then student involvement and consent, then a clear educational purpose. I read those answers as a set of conditions for spending. Who’ll make sure the answers are right? What will staff have to do to keep them right? What will students gain from it? 

There’s a reason the money question is sharp. In its May 2026 report, the Office for Students said 35.8% of the English higher education providers it analysed reported a deficit in 2024–25. If your university has no spare money for a standalone AI project, start with a problem it already pays to manage. In our example, that’s the lecturer answering the same question three times. (Office for Students.) 


2. Spend to save 

Start with one module and the routine questions that keep reaching its staff. Fix the source material first, then test whether a course assistant adds enough value over the clearer guidance alone. Count the repeat enquiries and the time staff spend on them. Include the time it takes to check answers, update the material and pick up the questions the tool passes back. Put those costs beside the licence and implementation costs. 

If you free up a few hours each week, agree what the team will use them for. More time for feedback is a real gain. To put a cash saving in the business case, name the expense that would actually come out of the budget. 


3. Spend to improve 

Go back to the student trying to submit their work. Can they find the right answer sooner? Can they tell where it came from? Can they get to a person when their situation needs judgement? Test that with students, including students who use assistive technology. If the tool gives a confident answer that sends someone down the wrong path, count that too. 

I’d take a one-term test to finance with a spending cap, a named owner and a decision date. Agree in advance what the tool can access, who checks its answers and what improvement would justify paying to continue. If the tool can’t clear that bar, keep the better guidance and end the trial. 

That’s a proposal a university can make a decision on. It starts with a student trying to get on with their work and ends with a result the university can check.


Richard Gascoigne is a non-executive director of Plato Academic.