Summary:
)
For over a decade, Alberta’s Ministry of Infrastructure tried to replace two computer systems that were barely limping along.
The first tracks every government-owned building in the province. Roughly 4,000 properties, land titles, and physical assets worth about $12 billion.
The second tracks construction budgets and timelines for every new hospital, school, and government building being built in Alberta. Over 500 active projects at any given time.
Both systems were ancient. The underlying database had reached end-of-life. Staff were manually copying data between disconnected spreadsheets just to answer basic questions about how much a building was worth or whether a construction project was on budget.
To give you a sense of how bad it had gotten: there were employees in both Infrastructure and Technology and Innovation whose full-time job was to copy information from one part of the legacy system and paste it into another. A literal human bridge between two pieces of software that couldn’t talk to each other.
These are dedicated, loyal public servants doing heroic work to keep critical systems running. Some of them couldn’t even take a vacation without putting those systems at risk. They deserved better tools a long time ago.
Since 2016, there had been three separate attempts to replace these systems.
All three failed.
Everyone agreed they needed to be replaced. The questions were: How? When? And who was going to pay for it?
The Standard Playbook
Infrastructure and Technology and Innovation did what governments always do when they need new technology. Together, we launched a formal procurement process.
Thirteen companies submitted bids. The shortlist came down to four major consulting firms.
The upfront costs alone were estimated at $19 million. When you factored in the customization, migration, and ongoing implementation costs over a four-year timeline, the total bill would land around $54 million.
Of serious concern to me: the $54 million only covered one of the two systems. The construction project management tool wasn’t even included in the scope.
Four years. $54 million. And you’d still need to come back and do the other half.
And if you’ve spent any time around large government IT projects, you know that $54 million over four years rarely stays at $54 million over four years. In my experience, a project like this would more realistically land north of $100 million by the time it’s actually delivered.
And by then, the technology has evolved so much that the finished product is already outdated before you even turn it on.
Most governments would accept this. After all, this is how it’s always been done. Big system, big vendor, big price tag, long timeline. It’s the default setting for government IT across the country.
The Decision That Changed Everything
The Deputy Ministers of Infrastructure and Technology and Innovation looked at those numbers and did something unusual.
They killed the procurement.
Not because the vendors did anything wrong. The bids reflected the reality of how large-scale government IT projects typically get scoped and priced. Vague requirements lead to expensive proposals, because vendors have to account for all the uncertainty they’re inheriting.
That’s a system problem, not a people problem.
But knowing that doesn’t make $54 million for half a solution any easier to swallow. Especially when the legacy systems were failing and staff needed relief now, not in 2029.
So we tried something different. We asked: what if a small team of public servants, equipped with modern AI development tools, built the replacement systems themselves?
Meet the Team Behind PRISM
In June 2025, we launched the PRISM Initiative.
PRISM Core replaces the old building and asset tracking system. PRISM Project replaces the spreadsheets and SharePoint sites used to manage capital construction.
The project is led by Cohen McLeod, a director in the Ministry of Technology and Innovation, with support from Zoran Mijajlovic, Executive Director of AI Delivery and Enablement. Their group is part of our AI Maximalist program, where we identify public servants with strong technical aptitude across government, bring them into dedicated product teams, and equip them with advanced AI development tools like Claude, Copilot, and Gemini to build solutions for other ministries.
On the client side, this project would not have been possible without the exceptional support of Deputy Minister Mark Kleefeld and his team at the Ministry of Infrastructure. From the very beginning, Infrastructure’s leadership understood what we were trying to accomplish and backed it fully. That kind of top-down support from the client ministry is rare, and it made all the difference.
As Mark put it: “PRISM will have a lasting impact across the department, not only by modernizing complex legacy systems that touch every business area, but also by showing that we can drive innovation at a speed and cost that would have been hard to believe even a year ago.”
The team’s approach was straightforward. Build working software fast. Put it in front of real users early. Collect feedback. Fix things quickly. Release updates every two weeks.
No four-year implementation plan. No $19 million in upfront costs. Just a small team solving real problems at speed.
What 643 Public Servants Think About It
Remember when I said the original RFP called for a four-year implementation? Here’s the good news.
Both systems are live.
Today.
In production and being used every day by 643 government employees across the Ministry of Infrastructure.
Staff who spent years wrestling with legacy systems and manual spreadsheets are now searching properties, tracking construction budgets, and pulling reports through a modern interface. Information that used to require compiling data from multiple disconnected sources is now available in real time.
The employees who used to spend their days copying and pasting between broken systems? They’re now doing meaningful work.
Cohen put it simply: “Folks are legitimately thrilled to be using a new system.”
That’s not an exaggeration. More than 170 Infrastructure staff voluntarily attend biweekly progress meetings to see what’s coming next. New features, fixes, and improvements ship every two weeks.
When users flag a problem or request a change, the team typically addresses it within two weeks.
Think about that from the perspective of a government employee who waited a decade for a new system. You go from being told “we’ll get back to you in four years” to seeing your feedback reflected in a working product within days.
That builds something no procurement process can buy: trust.
“Staff feel like they are part of the process,” Cohen told us. “If something is broken or not working, they have the trust that we will fix it.”
The Numbers
We originally budgeted $5 million for the PRISM Initiative. That was already a fraction of the $54 million the vendors had proposed. We thought it was aggressive.
Turns out we were conservative.
Ten months in, the team has spent $858,000. The combined monthly cost is roughly $118,000. At that pace, the estimated total cost to fully deliver both systems is approximately $2.64 million.
Let me put that in context.
The vendor proposals would have delivered one system for $54 million over four years. We are on track to deliver both systems for under $2.64 million in a fraction of the time.
That’s a 95% cost reduction. Both systems instead of one. Delivered faster. With 643 users already on the platform.
What AI Actually Did Here
I want to be precise about AI’s role, because this story is easy to oversimplify.
AI did not build these systems by itself. Cohen’s team members are experienced practitioners who know how to design and deliver digital products. They understand the business processes they’re digitizing. They work closely with Infrastructure staff to make sure what they build actually matches how people work.
What AI did was make them dramatically faster. And it showed up in ways I didn’t expect.
Turning 50 Hours of Video Into Build-Ready Requirements
Here’s an example that still amazes me.
Infrastructure had over 50 hours of recorded walkthroughs of their legacy systems. Staff had filmed themselves navigating every screen, every workflow, every data entry process. In a traditional project, you’d pay consultants millions of dollars to watch those recordings, interview users, and produce a requirements document. It takes months.
Cohen’s team used Google Gemini’s vision capabilities to process the video frames directly. The AI analyzed the screen recordings and transformed them into structured requirements, user flows, and data models. At roughly one cent per image processed, with hundreds of images analyzed in parallel, the team converted those 50-plus hours of recordings into detailed, structured build specifications in minutes.
Not days. Not weeks. Minutes.
The output wasn’t just a summary. The AI preserved the structure of every screen, every table, every interface element. When it analyzed an image of a spreadsheet, it didn’t produce a jumble of numbers. It captured the full structure of the data so the development team could turn it directly into working software without losing any detail.
The team has built data processing pipelines that can take thousands of pages of source documents and transform them into detailed build specifications that are more precise than anything they’ve ever received from outside contractors. These pipelines run on Google’s enterprise platform using Gemini and Google-hosted Claude instances.
Faster Development Across the Board
Beyond the requirements work, the team uses AI to rapidly prototype new features, sometimes producing working interfaces within hours. They use it to research complex legacy code bases, generate initial designs, and handle routine development tasks.
On the design side, they’ve started building prototypes live on calls with their users, so staff can see their feedback take shape in real time.
As Cohen explained: “AI makes development a lot faster, but not on everything. We’ll be working faster than anticipated, then hit something that we just have to do by hand.”
That’s an honest assessment, and it matters. AI is a tool that amplifies skilled people. It doesn’t replace the need for expertise, judgment, or the hard work of understanding what users actually need.
But when you combine domain knowledge with AI capability, the results speak for themselves.
Something Even More Interesting Is Happening
Here’s something else I didn’t expect.
Some of the Infrastructure staff who use the PRISM system have started picking up AI tools on their own. A few users have gotten access to Loveable, a rapid prototyping tool, and begun creating prototype interfaces for features they’d like to see in the system.
To be clear: these prototypes don’t go straight into production. They’re a way for non-technical staff to quickly sketch out ideas and user interfaces so the development team can see exactly what users are looking for. The tech team then takes that feedback and builds the real thing inside our secure government environment.
But think about what that represents.
Government employees who are not software engineers are using AI tools to communicate their ideas visually, in real time, directly to the team building their systems.
That’s the kind of culture shift you can’t force. It happens when people see AI delivering real value in their daily work and realize they can do it too.
This is exactly why we launched the Alberta AI Academy, our free, open-access AI training platform. We designed it for public servants, but it’s available to anyone.
Over 1,700 government employees have already enrolled in or completed the training. The Academy gives every participant access to the tools and training to augment their daily work with AI in a secure, responsible way.
Not to turn accountants into software developers. But to give people across every ministry, every industry, and every community the ability to solve problems faster, automate tedious workflows, and stop accepting “that’s just how we’ve always done it” as an answer.
Cohen’s team didn’t have some rare, unreplicable skill set. They had domain expertise, good tools, and a willingness to try something new.
That combination exists in every ministry, every department, every office across the Government of Alberta. It exists in every business, every non-profit, every classroom.
If a small team can save taxpayers over $50 million on a single IT project, imagine what becomes possible when thousands of people are equipped with the same tools and training.
If we can do it, you can too. And we can show you how. Visit albertaaiacademy.com and get started. It’s free. Our gift to Albertans, Canadians, and anyone who wants to level up their productivity with AI.
What Comes Next
The PRISM team isn’t done. On the asset management side, two additional legacy systems have been identified for replacement. On the project management side, the next priority is a finance feature set that will finally retire the last remaining Excel-based workflows.
Cohen and his team will keep building, keep shipping, and keep putting working software in front of the people who use it every day.
And we’ll keep telling these stories. Because this is what’s actually happening inside the Government of Alberta.
Not in a pilot. Not in a strategy document.
In production, with real users, solving real problems.
If you want to stay up to date on the work we’re doing to make Alberta the most innovative government in the country, subscribe to this Substack and share this article with someone who thinks government can’t innovate.
The PRISM team from Technology and Innovation: Cohen McLeod, Zoran Mijajlovic, Maryam Beik, Irene Gao, Muhammad Muiz, Sina Vejdan, Yousef Sabir, and Aiman Naeem.
From Infrastructure: Kai So, Mohammed Sunba, Karolina Kowalski, Shannon Patterson, Natalie Jaber, Shaun Carreiro, Ivy Michael Bacud, Daryl Blouin, Bruno Chikeka, Tara Dreger, Sandra Marchuk, Lynn Benn, Jane Forbes, Craig Chupka, Husam Alimami, and Evan Friesenhan.
Outstanding work, all of you.
Nate Glubish is the MLA for Strathcona-Sherwood Park and Alberta’s Minister of Technology and Innovation.