b'Forum: Evolving Use of Artificial Intelligence in Government using automated decision-making in support of serviceinspection, examination, enforcement action, or judicial delivery, it is required to assess the associated risks. Theproceeding.government developed four levels of impact an AI tool might have on society and government, ranging fromPeople: Train public servants on how to use AI tools. little to no impact that could be reversible and brief toIn January 2019, to address a skills gap and ensure very high impact, which might lead to irreversible andgovernment programs use AI tools responsibly, the perpetual changes.Canada School of Public Service launched a pilot cohort For use of AI with little or no impact on serviceof its public sector Digital Academy. It is seeking to programs, the directive allows for the possibility ofimprove the digital acumen of public servants at all automated end-to-end decision-makingin otherlevels and eventually expanding training to all public words, making decisions without human involvement.employees. Elevating the digital literacy of employees can However, it states that program officials must be able tohelp them get more comfortable with new technologies. explain how conclusions were reached. RequirementsAside from providing digital, data, and AI skills, the for AI used by high-impact programs, on the othergovernment hopes the training eases concerns by raising hand, include a peer review by government academics,awareness among public servants about the current state nongovernment organizations or other advisory boards;of AI and other digital technologies, and how they could repeated training for employees using the AI tool; andaffect their jobs and even private lives.documentation posted on relevant websites describingConclusionhow the tool works. In addition, a person must make anyAI tools also are expected to impact the federal government final decisions based on an AI tools recommendation.substantially, with implications for federal systems and Depending on the impact level, programs also muststructures. To capture the benefits of AI, federal agencies disclose to the citizen whether a decision affecting themmust be prepared to address related risks. The Office is made partly or wholly by an AI tool. The directiveof Management and Budget and Office of Science and also addresses AI transparency and the CanadianTechnology Policy should continue to lead efforts to manage governments right to access and test proprietary AIthose risks, given the technologys potential to transform work systems if necessary for a specific audit, investigation,governmentwide.64 www.businessofgovernment.org The Business of Government'