Alberta’s Ministry of Technology and Innovation built two critical government asset and construction tracking systems in-house using AI tools for $858K so far, versus a $54M vendor quote for one system.
Key Takeaways
The PRISM Initiative replaced two legacy systems – one tracking ~4,000 properties worth $12B, one managing 500+ capital construction projects – both now live with 643 users.
Vendor bids for a single system reached $54M over four years; PRISM is on track to deliver both for an estimated $2.64M total, a 95% cost reduction.
Team used Google Gemini’s vision capabilities to process 50+ hours of legacy system screen recordings into structured requirements, data models, and user flows at roughly one cent per image.
AI-assisted prototyping produces working interfaces within hours; two-week release cycles let users see feedback reflected in production quickly.
Non-technical Infrastructure staff began using Loveable to sketch prototype UIs directly, closing the gap between user intent and development input.
Hacker News Comment Review
Commenters were put off by the writing style and an AI-generated header image, with skepticism that fluffy presentation undercuts credibility of otherwise interesting technical claims.
A significant concern: no mention of testing anywhere in the article. Commenters questioned whether government-critical systems built this fast are adequately validated before production use.
The counterpoint acknowledged by one commenter: even a marginally better in-house approach beats a procurement process this broken, making cost and speed wins real regardless of style.
Notable Comments
@chrisjj: Searched the article for “test” – zero results. Flags the risk of applying the same move to higher-stakes infrastructure.
