Ottawa funds AI literacy; the part employers actually need is the audit habit

Canada's government has launched a national artificial intelligence strategy, AI for All, which links AI to public trust, economic opportunity and Canadian sovereignty, and which includes a National AI Literacy Initiative aimed at postsecondary students and educators ^[1][2]. Whether that initiative produces tool fluency or work an employer can stand behind is a curriculum question with a payroll consequence.

The case for teaching prompting is real and narrow. Prompt engineering means devising and revising instructions to get a useful response, and a strong prompt names the task, audience, context, relevant evidence and desired format ^[4]. A systematic review of prompt engineering in higher education found that structured prompting can support learning when students deliberately formulate, test and refine their interactions with the system ^[5]; other researchers treat prompt design as a rhetorical process shaped by purpose, context and audience rather than a set of technical tricks ^[6]. Asking students to compare several prompts, explain their revisions and evaluate the effect makes part of their reasoning visible ^[15].

The failure mode is that a polished output can still be false, biased or unsupported. Generative systems produce falsehoods with unwarranted confidence, fabricate sources, obscure uncertainty and generate claims students lack the subject knowledge to assess ^[7]. They also produce confident answers without bearing any responsibility for whether those answers are true, which leaves the responsibility with the person who uses the material ^[8]. The author of the argument, who teaches professional communication and researches prompt engineering, concludes that teaching students to communicate with AI is valuable only when it develops rather than displaces human judgment ^[16].

That is where the accountability requirement belongs, and it is specific rather than aspirational. On this account, students are answerable to instructors, classmates, employers, clients and the public, and in professional settings AI-assisted communication can affect patients, employees, customers and communities that had no say in whether AI was used at all ^[9]. The operational test is that a student can explain how AI was used in a piece of work, how its claims were checked, whether confidential information was entered into the tool, and why the final product meets the standards of the course or the profession ^[10]. That requires enough subject knowledge to recognise weak output, and verification of consequential claims rather than acceptance because they sound authoritative ^[12]. It also requires knowing when not to use the tool, including not entering private information into a commercial platform and not outsourcing high-stakes decisions ^[14]. Where disclosure is expected, the student names the AI's contribution and stays responsible for the whole ^[13]. UNESCO's AI competency framework for students takes a similar shape, pairing practical knowledge with ethics, human agency and responsible citizenship ^[11].

Read that list as a job description and the transfer of cost becomes obvious. A graduate who can prompt well but cannot produce a use log, a verification trail or a defensible reason for using the tool at all does not arrive without those artefacts; someone downstream builds them, or nobody does and the unverified claim reaches a patient, a customer or a regulator ^[9][10][7].

What to watch: whether the funded programs under the literacy initiative are specified beyond their stated audience of students and educators ^[2], and whether their assessments demand disclosure and verification records rather than task completion ^[10]. Employers hiring the first cohorts can test this cheaply by asking a candidate how they checked an AI-assisted claim and what they refused to put into the tool ^[10][14].