Alberta Presses Vendors to Adopt AI

Glubish cited examples of internal work completed with AI that he said was built "for 95 percent cheaper and in 95 percent less time." BetaKit reports that he described one contract whose lowest bid was $54 million over three years; according to Glubish, an in-house AI alternative met the requirement for $2.5 million in less than a year.

He also criticized traditional time-and-materials billing, telling the symposium audience that the model was "out of date, and not here to stay," BetaKit reports. These are claims made by the minister, rather than independently audited procurement comparisons in the available reporting.

Alberta has publicized earlier examples of the same broader program. In a March Substack post, Glubish wrote that a public servant used AI to migrate 1,519 contracts into the province's 1GX procurement platform, describing the result as two minutes of work that otherwise represented a 15-month backlog. That post put the AI compute cost at $1.12 and claimed $90,000 in savings.

The province has released white papers describing its methodology for bringing AI into the public service over the past 18 months, BetaKit reports. The documents are presented as open-source technical material intended to enable collaboration with governments, companies, and other organizations.

The Logic reported on July 10 that Alberta had built in-house tools on top of Anthropic models to assess government technology systems and develop applications for departments and agencies. According to that report, staff have used the tools to scan software code for security vulnerabilities, identify applications needing upgrades, and develop some replacements. The Logic also reported that Alberta open-sourced the full stack.

For ML and platform teams serving public-sector customers, the reported examples put attention on delivery metrics beyond model capability: migration throughput, system-maintenance scope, security review, and comparable cost estimates. Organizations undergoing similar AI-enabled modernization efforts often reassess whether external engagements should be priced around hours worked, defined outcomes, reusable tooling, or a combination of those models.

The available reports do not provide technical evaluation details for the cited in-house projects, such as model versions, validation procedures, error rates, data-governance controls, or lifecycle operating costs. They also do not establish whether the $54 million proposal and the $2.5 million internal alternative had identical requirements, risk allocations, or support obligations.

Those gaps matter when comparing an AI-assisted internal build with a multi-year vendor contract. Still, Alberta's public statements make clear that the province is connecting its AI adoption program to how it evaluates digital-service suppliers, with Glubish inviting vendors to follow its published roadmap, according to BetaKit.