The Government of Canada just published an AI Readiness Scorecard. Most departments should read it twice — and be honest the second time.

The Canada School of Public Service recently released a job aid to help public servants assess whether a problem is a strong candidate for an AI-based solution. It's well structured, practical, and frankly, one of the more grounded tools I've seen come out of the GC on this topic.

There are five sections. The first two ask whether the task is a good fit for AI and whether the problem is actually worth solving. Most teams will breeze through these feeling confident. Yes, the task is repetitive. Yes, the problem is painful. Someone book a vendor demo.

Then comes Section C. Is the current process well defined and stable? Has it already been simplified as much as possible?

Then Section D. Is the data accessible? Is it clean? Is it structured with appropriate governance?

This is where the scorecard quietly becomes a mirror. An uncomfortable one. The kind nobody puts in the boardroom.

I spent years working inside the Government of Canada. I watched departments spin up AI strategy working groups while their SharePoint environments were a living archaeological dig of folders named "FINAL_v3_REVISED_USE THIS ONE." I watched teams debate machine learning use cases while their data sat in siloed spreadsheets maintained by someone who was eighteen months from retirement and had never once been asked to document their logic. I watched metadata tagging exercises get celebrated as governance wins — actual applause, in actual meetings - while the RBACs and group policies remained completely unmanaged.

And then Copilot showed up. And suddenly all of that exposure was one prompt away from anyone with a Microsoft 365 licence and a vague sense of curiosity.

The scorecard's Section D asks three questions: Do you have the right data? Is it of sufficient quality? Is it structured and governed? In most departments I worked in, the honest answer to all three was no. Not close. Not almost. Not "we're working on it." No.

The tool is smart enough to tell you what that means. A combined score of 0–2 on process and data quality means one thing: pivot. Significant foundational work is required before any AI project can succeed.

That's not a criticism of ambition. The GC needs to modernize and AI is part of that future. But ambition deployed on a broken foundation doesn't accelerate progress — it accelerates the mess. Faster. At scale. With an AI co-pilot.

The departments that will actually succeed are the ones willing to score themselves honestly on sections C and D, accept the result, and do the unglamorous work first. Clean the data. Define the processes. Lock down the access controls. Build the foundation nobody photographs for the deputy minister's year-end deck.

The scorecard exists precisely because someone in the GC understands that most AI projects fail not from lack of technology, but from lack of honesty about readiness.

Use the tool. But use it with a straight face.

And maybe - just maybe - resist the urge to name the working group before you've answered section D.

ai-readiness-scorecard-eng.pdf

Originally published by Ross Norrie, founder of SkyeConnex, on LinkedIn.

Published March 19, 2026 · More from the SkyeConnex blog