The Ontario government is dipping its toe into the rapidly growing world of artificial intelligence, telling civil servants they can use Microsoft Copilot in their day-to-day jobs and offering some more advanced tools.
The move appears to have been embraced, according to internal metrics, which show the Ontario Public Service has the highest Copilot chat use in the country.
Government presentations, obtained using freedom of information laws, show more than 15,000 civil servants use the artificial intelligence assistant every week, with more than 120,000 pageviews for the “Copilot Chat InsideOPS” page.
That, government presentations claim, is the highest usage in Canada.
Security concerns around public data and sensitive information mean government employees are banned from other chatbots like Google Gemini or ChatGPT in most cases, but are being encouraged to find ways to integrate generative artificial intelligence into their roles.
Ontario’s Minister of Public and Business Service Delivery and Procurement, Stephen Crawford, who is in charge of the program, said the early signs were that the new system was working.
He said, in addition to Copilot across government, some artificial intelligence leaders and pioneers had been chosen for more complex tasks and integrations.
“We’ve initiated a rollout to several hundred people of a more advanced program, which we’re under review with right now,” he told Global News. “From what I understand, at current levels, the average public servant is saving almost three hours per week.”
The documents seen by Global News show the government is currently in its first phase of rolling out artificial intelligence, where it is looking to “establish AI foundations” and identify “use cases.”
Get breaking National news
For news impacting Canada and around the world, sign up for breaking news alerts delivered directly to you when they happen.
Once it wraps up its pilot phase, it will look to develop purpose-built artificial intelligence, something one document suggests it hopes will increase productivity by 20 per cent.
A third phase labelled “industrialize AI” would see the government apply artificial intelligence “at scale.”
Crawford said the widespread introduction of artificial intelligence across government wouldn’t necessarily mean job losses among the civil servants, arguing it would give people time back.
“There will be job transformation, people are going to move into more productive roles,” he said.
“So no, I don’t see job loss, I see job enhancements and productivity improvement, which companies and organizations and governments will reinvest to get further productivity gains.”
While the long-term implications could be major, the current examples are relatively straightforward.
Examples of good use cases for artificial intelligence referenced in government documents included media monitoring and drafting news releases.
Cam Vidler — the co-founder of the artificial intelligence company Authentica and former chief of staff to the finance minister — said the potential of agentic artificial intelligence meant the government could think bigger.
Agentica AI is a developing area, where AI agents can be programmed with a set of criteria to perform certain tasks — combining the decision-making ability of AI with strict rules to avoid hallucinations seen in broader chatbots.
“On the back end of government, you’re talking about automating work that is around the operations of government — sourcing and procurement, making payments that the government’s obligated to make,” he said. “There’s a lot of efficiencies you could find.”
Vidler said training agents to handle tasks at ServiceOntario or the administration of questions with programs like government rebates could save manhours and offer faster answers for members of the public.
Last year, for example, the government offered $200 rebate cheques to people across the province, ahead of an early election.
In some cases, the program faced issues with out-of-date addresses, delays or cheques being sent to people who had passed away. Those issues couldn’t generally be solved through existing IT systems, a situation Vidler suggested AI could be used in the future.
“Exceptions go straight into somebody’s inbox and it’s something somebody has to deal with,” he said. “So when you look at how many call centre agents you need to deploy to roll out a program like that, the main driver of that is there’s just things we can’t reliably automate.”
With Copilot the only artificial intelligence approved for most in the civil service, the government is likely still far from those situations.
Crawford said it was “hard to say exactly” when he would ramp up the use of artificial intelligence.
“But if I had to speculate,” he said, “I would say within a few years.”
Vidler said it was important to select a few situations to use the systems rather than to try and come up with an entire overarching strategy for a technology that is ever-evolving.
“ You don’t have to go and decide to do AI on everything, right? Or come up with a huge AI policy that thinks of all the different circumstances,” he said.
“Pick a couple low-hanging fruit examples, define processes that you know you could automate better, that you know are prone to errors right now, and that you don’t want your people spending time on.”
© 2025 Global News, a division of Corus Entertainment Inc.
