Janak Alford, Alberta’s deputy minister of technology and innovation, was sitting on stage The upper limit in Edmonton in May, while his laptop directed an agent across the room.
The agent had been working 18 of the previous 24 hours on a question a colleague had called Alford, Alberta’s deputy minister of technology and innovation, around the previous morning. She wanted to know how Alberta’s approach to reducing red tape compared to any other province, without understanding what the underlying laws were designed to do.
Six months ago, that kind of benchmarking took six to nine months of consulting work. Alford’s agent was still building the proof of concept on his laptop as he answered questions on stage.
Last week Alberta made the method public. On July 6, the province released 21 subpoenaed technical documents Velocity White Paper that walk through how they ran AI agents across their codebase, what came back, and how the province is using the results to find vulnerabilities and rebuild legacy systems.
Anthropic, the artificial intelligence company whose Claude models enabled the audit, published a case study on the project on the same dayshowing scan totals, timelines and rebuild figures under notification.
The audit that produced the documents scanned 466 million lines of code and found that a system of 1,280 applications actually did about 700 things. The rest were doing work that other ministries were already doing.
This discovery is where rebuilding begins, and it’s what most large organizations sitting on decades of code have yet to do for themselves.
Alberta directed the agents to herself first
The audit that produced the letters took place earlier in 2026, before Alford came to the Upper Bound.
The Alberta Ministry of Technology and Innovation leads the digital infrastructure behind all 27 provincial ministries, including case management tools, benefits portals, registries and cyber security systems. Everything from fire response to social services to public safety sits at her helm.
This works across approximately 1,280 applications and 3,400 code repositories, according to the Anthropic case study. Many had never undergone a systematic safety review.
“I think nowhere are tools needed more urgently than in public service,” Alford said.
Audition was a separate project from the overnight demo Alford was running from his laptop on stage in Edmonton. About 50 agents ran in parallel across the ministry’s infrastructure using Claude Code with the Opus and Sonnet models. They scanned 466 million lines of code in about 20 hours. Anthropic’s own estimate for a human-led equivalent is 6.5 years.
About 50 agents ran in parallel using Claude Code with the Opus and Sonnet models. They scanned 466 million lines of code in about 20 hours. Anthropic’s own estimate for a human-led equivalent is 6.5 years.
Alberta says that applying AI agents to legacy systems can cut modernization time by up to 95% and speed up deployment by up to 20 times. He estimates that a conventional modernization of the same systems would cost $2 billion and take more than a century.
Alberta’s Minister of Technology and Innovation, Nate Glubishtalked about the security issue for the Anthropic case study audit.
“Albertans trust their government with some of the most sensitive information in their lives, and it’s our responsibility to protect it,” Glubish said. “By using AI to find and fix vulnerabilities across our systems, we accomplished in hours what a traditional approach would have taken years to complete.”
What Alford talked about in Upper Bound was the audit that came before reconstruction.
He told the room AI can extract code at 10,000 lines per hour from anyone, coder or not. Doing so, on top of decades of code no one had a complete picture of, produces what he called “the potential slope of what AI could produce.”
First, his team ran agents across Alberta’s code base and broke down the results into “something in the region of 700, give or take, government business functions that are served by code,” Alford said.
That count of 700 functions exposed just how much duplication was hiding underneath.
Different ministries had built their own entrances, case tools and ticketing systems for the same basic jobs. Because why have one logical login screen when you can allow 27 different ministries to build 27 different login screens to do the same job?
“If you don’t have those pre-built parts, you can’t build your car,” Alford said.
What Alberta is building now is a set of common components that go into new applications so that no one ministry has to rebuild what already exists.
One ministry has 185 legacy applications that Alberta plans to consolidate into 16 modern ones, built on components that surfaced.
A rebuild of the province’s subsidy program portal took four to five days. The original Java version took about five months to build roughly 25 years ago.
Of course, tech debt isn’t just an Alberta problem. Gartner places about 40% of infrastructure systems across all asset classes in the technology debt category.
McKinsey estimates that technological debt amounts to 20% to 40% of the value of a company’s technological assets. The same report found that roughly 30% of CIOs said more than 20% of the budget earmarked for new products was being diverted to deal with it.
Other large organizations are trying similar approaches. Morgan Stanley is leading its own version and the bank’s global head of technology and operations, Mike Pizzi told Bloomberg in October, AI coding has had “a pretty profound impact” on how software is built within the bank.
What is different about Alberta is that it has published their method. The most difficult question (and probably one that every CEO will raise) is what happens to the time that AI gives back.
Time saved does not stay saved
The Alford response is Jevons’ Paradoxnamed after a British economist who observed in 1865 that as steam engines became more efficient, England burned more coal. Cheaper coal meant more people used it, for more things, in more places. Efficiency unlocked demand rather than shrinking it.
Alford went through more contemporary versions. Cars were more fuel efficient and Canadians drove more. Dial-up gave way to fiber and internet consumption exploded.
“The more efficient you make a system, the more demand there is for that system,” Alford said. “I haven’t found myself personally working less.”
of 21 for free Velocity White Paper walk through the entire show, from scanning the code to rebuilding it to training the people who work with the agents.
One of those documents documents Alberta AI Academythe ministry’s own training program. More than 2,000 public servants in Alberta have been trained through the Academy since its launch in September 2025, and more than 15,000 people across Canada have used the platform.
The agent, working on Alford’s laptop while he sat on stage, produced a working proof-of-concept for his colleague’s problem in less than 24 hours. The ministry-wide scan did the same for 466 million lines of code.
Alford’s reading of where that all leaves us came at the end of his Jevons argument.
“I haven’t seen an upper limit to humanity’s appetite to invent or accelerate or develop new and creative things,” Alford said.
The last shots
- Everyone wants to talk about agents. Alberta spent more time on what she found. In 1280 applications, different ministries had built their version of the same basic works, over and over again.
- 21 Velocity White Papers show the work. Most organizations that do agent-driven modernization keep it behind closed doors.
- Any enterprise that maintains decades of software will eventually have to answer the same question that Alberta asked. Before you upgrade, do you know what you’ve built?
