One summary per publication, written so that you can tell in twenty seconds whether the full thing is worth your time. No jargon, no statistics without a plain explanation of what they mean.
All 20 publications. Free to read, free to reuse under Creative Commons BY 4.0.
Canada spent twenty years getting people online. A new gap is opening on top of that one, and it is about who uses AI rather than who has internet.
The gap between the age groups who use AI most and least is wider than any digital gap measured before — wider than gaps by income or education. Fixing internet access did not fix this, and will not.
Schools cannot train people for jobs that do not exist yet. They can teach skills that transfer.
About 39% of the skills people use at work are expected to change by 2030. Teaching a fixed job description is a bet against that. Teaching skills that move between jobs is not.
For the first time in decades, recent graduates are more likely to be unemployed than workers overall.
Entry-level work is where people used to learn a trade or profession. If AI absorbs those tasks, the rung people climbed is missing — and nobody has replaced it.
Women now use AI tools at close to the same rate as men. They are still not the ones deciding how it is used.
Access is no longer the barrier. Leadership is. The gap shows up in who runs AI projects, who founds AI companies, and who is in the room when the decisions get made.
The AI gap across Northern, rural and remote Canada is following the same map the broadband gap did.
Connectivity came late to these communities, and adoption support has not arrived at all. Without it, the same regions get left behind twice.
Older workers are not refusing to use AI. They are not being trained on it.
The fastest-growing part of the workforce receives the least employer training. Early evidence suggests AI has been kinder to experienced workers than to beginners — which makes the training gap harder to justify.
Almost every organization now uses AI somewhere. Two-thirds never get past the pilot.
The blocker is rarely the technology. It is data quality, workflows that will not bend, and no agreed way to measure whether it worked. More pilots will not fix any of those.
Treating AI as a pile of separate experiments wastes most of the value.
The organizations getting returns manage AI like a portfolio: a few governed bets, reviewed together, with someone accountable for the whole rather than each piece.
The distance between organizations that are ready for AI and those that are not is growing, not closing.
Leaders report several times the revenue growth of laggards. Because the advantage compounds through better data and better habits, waiting makes catching up harder rather than easier.
When every company can buy the same models, the model stops being the advantage.
The market is sorting into three groups. What separates them is not technology but how fast an organization can decide, redesign work, and act on what it learns.
Training budgets went up. Guidance from leaders did not.
Only about a quarter of workers say their leaders give enough direction on AI. Sending people on a course does not answer the question of what they are supposed to do differently on Monday.
AI is the biggest hiring boom in a decade, and women are getting a shrinking share of it as the roles get more senior.
Women hold roughly a quarter of new AI hires, a fifth of Head of AI roles, and an eighth of AI seats in the C-suite. The pattern narrows at every step up.
Training decides who gets promoted, not just who gets hired. Disabled and racialized workers face barriers inside it.
If the training that leads to advancement is harder to access, the barrier compounds: people are not just excluded from a course, they are excluded from the progression it unlocks.
Everything a small organization needs to evaluate its own work, without hiring an evaluator.
Theory of change templates, an indicator library, consent and interview guides, and a reporting outline. Adapt them; they are meant to be edited.
A checklist so that accessibility is designed into a study rather than added when someone asks.
Covers recruitment, consent, session design and how findings are shared. Each item is a decision to make early, not an accommodation to react to.
Short answers to questions practitioners keep asking, with an honest confidence rating attached.
Each note says how much the evidence actually supports, so a thin answer is never mistaken for a settled one.
How to write findings that a busy person will finish reading.
Rules with worked before-and-after examples. The test is not whether it sounds rigorous but whether the reader can act on it.
Large print, easy read, audio and tagged documents are all produced on request, free, and without asking you to explain why. Use the request form or email contact@cometx.ca.