Alberta’s Ministry of Infrastructure abandoned a multi‑year, $54 M procurement for legacy system replacement and instead let a small, AI‑enabled team build the new tools themselves. The PRISM Initiative delivered both asset‑tracking and construction‑management platforms in under a year for roughly $2.6 M, a 95 % cost reduction. The article examines the technical approach, the cultural shift it sparked, and the arguments that caution against treating this success as a universal formula.
A surprising turn in government IT
When the Ministry of Infrastructure announced a four‑year, $54 million contract to replace two decades‑old systems, most observers assumed the price tag was inevitable. The bid covered only one of the two required platforms, and history suggested the final bill would balloon well beyond the estimate. Instead of signing the contract, Deputy Ministers Mark Kleefeld and his counterpart pulled the plug and asked a different question: What if a small team of public servants, equipped with modern AI tools, built the replacements themselves?
The PRISM Initiative in practice
In June 2025 the Ministry of Technology and Innovation launched PRISM (Public‑sector Real‑time Integrated Systems Modernisation). Two parallel products emerged:
- PRISM Core – a searchable database for every government‑owned building, land title and physical asset (≈ 4,000 items, $12 B in value).
- PRISM Project – a dashboard for budgeting and scheduling of all capital construction projects (over 500 active projects at any time).
The team, led by director Cohen McLeod and AI Delivery lead Zoran Mijajlovic, combined domain expertise with AI assistants such as Anthropic’s Claude, Microsoft Copilot and Google Gemini. Their workflow was deliberately lean:
- Rapid prototyping – build a usable screen in hours, not months.
- User‑first feedback loops – bi‑weekly demos for 643 staff members, with changes shipped every two weeks.
- AI‑augmented requirements – ingest 50 hours of legacy system walkthrough videos, extract UI elements and data models via Gemini’s vision API, and generate structured specifications in minutes.
The result? Both systems went live within ten months, serving 643 employees who previously spent days copying data between spreadsheets. The total spend to date is $858 k, and at the current burn rate the projected cost to finish both platforms is about $2.64 million – a 95 % reduction compared with the original vendor proposal.
What AI actually contributed
It’s easy to attribute the whole achievement to AI, but the article makes a clear distinction:
- Human expertise – the team understood the business processes, performed user research, and made architectural decisions.
- AI as accelerator – Gemini’s vision model turned video frames into data schemas; Claude helped draft code snippets; Copilot suggested UI components. These tools cut weeks of manual work to hours.
- Limitations – the team still faced tasks that required manual effort, such as integrating with legacy databases and handling security compliance.
The takeaway is that AI amplified skilled staff rather than replacing them.
Community reaction and cultural ripple effects
The most striking side‑effect is the emergence of a bottom‑up innovation culture. Some infrastructure staff began experimenting with a low‑code prototyping tool called Loveable, creating mock‑ups that the development team could refine. Over 1,700 government employees have enrolled in the free Alberta AI Academy, a training platform that teaches public servants how to apply AI responsibly in their daily work.
Counter‑perspectives
While the PRISM story is compelling, several cautionary points deserve attention:
- Scalability of the model – The initiative succeeded because it focused on two well‑defined domains with clear data structures. More complex, cross‑departmental systems (e.g., health records) may involve tighter regulatory constraints and legacy integrations that AI alone cannot untangle.
- Vendor ecosystem impact – Large consulting firms argue that their experience in risk management, contract governance and long‑term support remains valuable, especially for projects where failure could have public‑safety implications.
- Talent concentration – The PRISM team relied on a handful of technically adept civil servants. Replicating that skill set across all ministries would require sustained training investments and may encounter recruitment bottlenecks.
- Security and compliance – Using cloud‑based AI services raises questions about data sovereignty and auditability. The article notes that pipelines run on Google’s enterprise platform, but broader adoption would need rigorous policy frameworks.
- Long‑term maintenance – Rapid delivery can produce technical debt if not paired with robust documentation and hand‑over processes. Ongoing costs for updates, security patches, and user support must be factored into any cost comparison.
What this means for the broader public sector
The PRISM Initiative demonstrates a viable alternative to the traditional, multi‑year procurement cycle:
- Cost efficiency – Even a conservative estimate shows a ten‑fold reduction in spend.
- Speed to value – Users see functional features within weeks, not years.
- Empowerment – Front‑line staff gain a voice in shaping the tools they use.
If other jurisdictions can replicate the combination of domain knowledge, AI‑enhanced tooling, and strong executive sponsorship, the potential savings could be significant. However, success will hinge on thoughtful governance, clear security policies, and realistic expectations about what AI can and cannot do.
Looking ahead
The PRISM team plans to retire two additional legacy asset‑management systems and add a finance module that will eliminate the last Excel‑based workflows. Their iterative, AI‑augmented approach will continue to be a live case study for how governments can modernize without the traditional price tags.
For a deeper dive into the technical stack and to explore the AI training resources mentioned, visit the Alberta AI Academy.
Comments
Please log in or register to join the discussion