Four pieces of work, and what they cost to get right.
Each of these is a system rather than a deliverable. What follows is the decision behind each one, not a description of the finished thing.
One topic, three channels, one week
London School of Economics · 2026
Staff had a new set of options inside Microsoft Copilot and no idea which to use. The temptation was to build a course. A course would have reached the people who already sign up for courses, which was the group who needed it least.
What I built instead
A web guide
Ten minutes to read, with a decision flow, an interactive practice activity and a worked comparison. Hand-coded rather than authored in a tool.
An email
Three points and one link. Written to be skimmed in a crowded inbox, not read.
A 74-second video
Shot vertical at 1080×1920, because staff watch these on a phone inside Teams, not on a desktop.
The craft decisions
The guide supports dark mode by system preference and by manual override, because a page read in a dark office at 7pm should not glare.
Every diagram carries a written description for screen reader users — not alt text restating the label, but a sentence explaining what the diagram shows.
All animation is disabled under prefers-reduced-motion.
Every control is at least 44 pixels, which is the minimum reliable touch target.
The copy-to-clipboard button has two fallbacks, ending in a plain instruction to press Ctrl+C. It never silently fails.
None of that is visible to a reader who has no need of it. That is the point.
Screenshot of the guide — pending permission to publish
The guide's decision flow and practice activity.
What I'd do differently
Model names in the guide will date within months, and I said so inside the guide itself rather than quietly letting it rot. Next time I'd separate the volatile content into a block that can be updated without reopening the page.
A capability framework for a whole staff body
London School of Economics · 2026
The institution had plenty of AI training and no way of saying what any of it added up to. Courses existed because somebody had asked for one.
What I did
Defined eight role-based pathways across five proficiency levels, from knowing what responsible use means to leading a change across other people's work.
Mapped every existing course, live session and learning plan against it — forty items in total.
Designed the top two levels myself as scenario-led courses, assessed by work product.
What the mapping exposed
Provision stacked up at the bottom two levels and thinned to almost nothing at the top. The organisation had taught a lot of people to use a tool and almost nobody to supervise its use — which is where the actual risk sits.
That finding was only available because the mapping existed. It is the argument for building the framework before building any more courses.
Pathways and levels map — pending permission to publish
Eight pathways across five levels, with provision mapped against each cell.
Canvas across a Russell Group university
Imperial College London · 2024–2026
Two legacy virtual learning environments, one faculty with particularly complex requirements, and a decision to make about what the whole institution would teach on for the next decade.
My part in it
Represented the Faculty of Medicine through requirements gathering, requirements testing and launch — the faculty's voice in a decision affecting every department.
Ran proof-of-concept testing across Canvas, D2L Brightspace and Blackboard Ultra, and designed the structured evaluation sessions that informed the final choice.
Designed the Canvas course template now used university-wide.
Wrote the Canvas training guides for Medicine staff.
Managed migration of content and users from two legacy platforms into one.
Why the template mattered most
A template is a quality standard that nobody has to read. Every course built from it inherits a consistent structure, predictable navigation and accessible defaults, whether or not the person building it has thought about any of those things.
It is the same instinct as the design scale: decide once, centrally, so that the default outcome is a good one.
Canvas template structure — pending permission to publish
The course template adopted across the institution.
A published quality standard
London School of Economics · 2026
Five pass/fail gates, then 43 criteria worth 100 points across nine pillars, built from four independent expert reviews and eight staff focus groups — with a documented method for calibrating AI scorers against it.
Read how it works, including the limitation I'd want a reader to know about before using it.
Want to see more?
Most of my course builds are internal to the organisations they were made for, so they aren't published here. I'm happy to walk through them live — email dwaine.mclean92@gmail.com.