Quebec Cancels Public-Sector AI Projects After Technical Failures

Quebec abandons AI pilots amid implementation challenges

Quebec's Ministry of Cybersecurity and Digital Affairs has terminated three artificial intelligence and automation pilot projects in the provincial government after they encountered what officials described as "significant challenges," according to documents obtained by The Canadian Press.

The scrapped initiatives included a chatbot for the Quebec.ca government portal, automated insurance form processing for civil servants, and streamlined billing workflows. All three projects, announced in February 2024 by then-minister Éric Caire, failed to progress beyond the pilot stage.

Technical obstacles derailed chatbot development

The conversational AI project faced multiple technical barriers. A November 2024 briefing note to the ministry's deputy minister cited problems with conversation fluidity, handling complex queries, error management, and system latency.

The memo also highlighted market readiness concerns. Available AI solutions "are not yet mature and are evolving rapidly," according to the document. While these tools enable rapid development of basic chatbots, "all complex and specific operations require additional development" that becomes more difficult with open-source technologies.

Despite initial plans to deploy a chatbot on Quebec.ca by December 2026, the ministry confirmed the project has been permanently shelved with no new timeline established. The feasibility study and business case development planned for 2025 never materialized because the initiative "was not prioritized."

Automation efforts exposed infrastructure gaps

The two workflow automation pilots using Microsoft tools showed initial promise but ultimately proved unworkable. The ministry identified several critical barriers: restricted data access, security-limited features, and insufficient internal technical expertise to identify workarounds.

More fundamentally, at least one targeted process was "not sufficiently optimized for robotic automation," and the ministry "had not reached the maturity required to deploy automated workflows in its administrative processes," the briefing note stated.

Why it matters

These failures reveal a common trap for government AI adoption: attempting to layer advanced technology onto unstructured processes and data. The Quebec experience demonstrates that successful public-sector automation requires foundational work in data architecture and process optimization before deploying AI tools—a lesson relevant to government agencies worldwide racing to implement generative AI without adequate preparation.

Expert calls decision sound management

Stéfanie Vallée, a researcher in sustainable digital transformation at Université du Québec à Chicoutimi, characterized the ministry's decision to halt the projects as "wise" and evidence of responsible stewardship of public funds.

"In digital transformation, you can't just digitize chaos," Vallée said. "It all starts with structuring the data. When our data isn't structured, it's hard to achieve anything."

Vallée noted that Quebec lacks the data infrastructure necessary for optimal AI deployment and hasn't fully conceptualized its digital government strategy. She emphasized that while the pilots were "entirely legitimate" learning exercises, they revealed resource and capability gaps that required acknowledgment.

The ministry maintains the experiments "yielded tangible results, such as the development of promising prototypes and the establishment of a solid foundation for future initiatives."

Quebec announced a separate partnership last June with AI companies Cohere and Mistral to explore public service integration opportunities, suggesting the province hasn't abandoned AI ambitions entirely—just the specific implementations that proved premature.

These details were first reported by The Canadian Press.