What we are working on, and what we still do not know.
Most research centres publish findings and keep the agenda private. We would rather say what we are looking at, so that people who know more than us can say so before we publish rather than after.
Questions we are actively working on.
Each of these will become a report, a brief series or an open resource. If you have data, a case, or a reason we are wrong, we would like to hear it.
What happens to the training budget when AI arrives
Budgets are reported as rising while workers report less guidance. We want to know where the money is actually going, and whether any of it reaches the people most exposed.
Whether AI adoption is inheriting the shape of the broadband gap
Our northern report suggests it is. The next step is testing that against sector and firm-size data rather than geography alone.
Does accessible-by-default training cost more?
The assumption that it does drives a great deal of exclusion. We have not found good Canadian evidence either way, which is itself a finding.
Where women leave the AI pipeline, and why
We can see the drop-off between graduation and the C-suite. We cannot yet say which transitions do the damage.
What community research ethics review should look like without a REB
Small organizations run real studies with no review mechanism at all. We are working out what a proportionate one looks like.
Whether hand-over actually happens
Every consultancy promises capability transfer. We want to follow up on engagements a year later and find out how often it survives contact with reality.
What we know we do not know.
Stating this is part of the method. Each of these is a limit on how far our published work should be pushed.
- Small firms. Most enterprise AI evidence comes from large organizations. What any of it means for a twelve-person company is largely unstudied.
- Quebec and francophone Canada. Under-represented in our own work, which is a limitation we own rather than a gap in the field.
- Longitudinal evidence. Almost everything in this space is a snapshot. We have no cohort we can follow.
- Disability and AI at work. Thin evidence, strong opinions, and the group most likely to be affected by automation decisions.
- What training actually changes. Attendance and satisfaction are measured everywhere; behaviour change almost nowhere.
Read a draft before we publish it.
We are building a small panel of reviewers who will read a draft honestly and tell us where it overreaches. Reviewers are credited unless they prefer not to be, and paid an honorarium.
This is the cheapest way we know to keep a small centre's research defensible.
Tell us we are wrong
If something we have published does not match what you see in your sector, that is worth more to us than agreement. Write to contact@cometx.ca and we will take it seriously, including issuing a correction where one is warranted.