Alberta's Best

Alberta Let AI Read Every Line of Its Code. Then It Gave the Playbook Away

Alberta Let AI Read Every Line of Its Code. Then It Gave the Playbook Away
AB Best July 8, 2026

Alberta Let AI Read Every Line of Its Code. Then It Gave the Playbook Away

Alberta used AI to scan 466 million lines of government code in 20 hours — then published its entire playbook free for other governments. The inside story.

A small team inside Alberta’s government spent 18 months using AI to secure and rebuild decades of aging systems — then published everything it learned, free, for any government to copy. The story behind the quietest big announcement of the year.

On July 6, Alberta’s government did something governments almost never do.

It published its homework.

Twenty-one technical papers — the Velocity White Papers — landed online, free and open-source, documenting exactly how a team inside the Ministry of Technology and Innovation spent the past 18 months using AI to review, secure, and rebuild the province’s digital plumbing. Not a press release describing the work. The actual methods, step by step, for any government on the continent to take and use.

The announcement got wire coverage and a few national tech briefs. What it didn’t get is the treatment it deserves as an Alberta story — because underneath the technical language is something more interesting than a procurement win. It’s a small team, a very old problem, and a decision at the end that says a lot about how this province thinks.

The $2 Billion Legacy Code Problem Inside Alberta’s Government

Every government in Canada is sitting on the same thing: decades of accumulated software that still runs essential services and that nobody fully understands anymore.

In Alberta, the Ministry of Technology and Innovation maintains the systems behind all 27 provincial ministries — roughly 1,280 applications and 3,400 collections of code covering everything from social services to public safety to wildfire response. These are the systems that process an AISH payment, hold a family’s case file, store tax records and procurement data.

Much of that code had never undergone a systematic security review, and by the province’s own accounting, the technical debt buried in it — outdated software, unaddressed bugs, insecure code — runs into the billions of dollars.

The conventional fix carries a conventional price. Modernizing it all the traditional way, the government says, would cost roughly $2 billion and take more than a century.

A century is not a plan. It’s a way of saying never.

466 Million Lines of Government Code Scanned in 20 Hours

In 2025, the ministry stood up an internal team with a mandate to make those systems more secure and easier to maintain, working with AI tools from Anthropic and Google.

The number that travelled furthest from last week’s announcement: the team put roughly 50 AI agents to work in parallel and scanned 466 million lines of government code in about 20 hours, checking it against the province’s security controls. By the team’s own estimate, the same review done conventionally would have taken about six and a half years.

The scan was the loud part. The quieter parts matter more.

Where the review found a vulnerability, the AI could often generate a fix, test it, and build it. Where a system was missing the automated tests needed to confirm a patch was safe, it wrote the tests first.

And where code was too old or too tangled to patch efficiently, it rebuilt the thing in a modern language — including a subsidy program portal originally hand-coded in Java about 25 years ago. Five months to build the first time. Four to five days to rebuild.

Every patch went through human engineers before it shipped. The team also built specialized review agents that now run continuously through the development process, checking new code as it’s written. Nothing about this was AI running loose in government systems — which is precisely why it’s a story other governments can learn from rather than a cautionary tale.

The Alberta AI Academy Started With 65 Volunteers

The part of this story that reads most like Alberta is buried in the white papers themselves.

The whole effort didn’t begin as a mega-project. It began with about 65 public servants who volunteered for an early cohort the team called AI Maximalists — ordinary staff willing to figure out what these tools were actually good for. That experiment revealed enough demand that the team pitched a training program to Minister Nate Glubish in June 2025, and the Alberta AI Academy launched that September.

Since then, thousands of government employees have gone through it, along with more than 10,000 members of the public — the course materials are published openly too, at albertaaiacademy.com. The Academy runs in levels, from the foundations of working with AI up to building full applications, with cohorts drawn from every level of seniority.

That sequence is worth sitting with. The security scan gets the headline, but the province spent its first year on people — a few dozen volunteers, then a training pipeline — before the big technical swings. The tools came second.

The Velocity White Papers: Alberta’s Free AI Playbook for Governments

Which brings it back to July 6.

Alberta could have kept all of this as a quiet internal advantage. Instead, it published the full methodology — the 21 Velocity papers, free at thevelocitywhitepapers.com — and is hosting an industry day in Edmonton this month to walk others through it.

This fall, the province begins scaling the approach across the rest of the government, with plans to replace nearly two hundred separate legacy systems with 16 new applications that Alberta will own outright, rather than license.

Glubish’s framing was characteristically direct: “Alberta spent decades building technology that worked for government.” Now, he says, the province is rebuilding it to work better for Albertans, faster and for far less — and sharing the tools because every government is stuck with the same aging systems Alberta was.

The province claims the AI-agent approach can speed up work by as much as 20 times and cut modernization timelines by as much as 95 per cent. Those are big numbers, and honesty requires saying plainly that they’re self-reported — the province’s figures, echoed by its vendor, not yet audited by anyone independent.

Publishing the full methodology is, if nothing else, an invitation for that scrutiny. You don’t open-source your homework if you’re worried about someone checking it.

The Other Half of the Alberta AI Story

Alberta’s AI moment has mostly been told as an infrastructure story — the data centers, the power, the pitch to Silicon Valley, the $100-billion target. That story is about Alberta hosting the AI boom.

This one is different. This is the provincial government actually running on it — the same ministry, the same minister making those Silicon Valley trips, quietly becoming one of the first governments anywhere to put AI to work at scale on its own systems and then documenting how. Digital pipelines were always Glubish’s metaphor for the data centers. What his ministry just published is closer to a drilling manual, handed to the neighbours for free.

Somewhere this month, in an Edmonton conference room, public servants from other governments will sit down to be walked through how a province of five million did what their own IT departments have been calling impossible for twenty years.

They’ll have come to the right place. Albertans could have told them that.

Sources

Tags:

Leave a Reply