A few weeks ago, Quebec’s professional regulatory body for HR practitioners, the Ordre des conseillers en ressources humaines agréés, published a list of recommendations ahead of the next Quebec provincial election.
Some of these recommendations address the use of artificial intelligence in the workplace and echo the guidelines issued by the professional body last June.
Here are the key takeaways for a “responsible, thoughtful and compliant” use of AI in the workplace.
“The integration of AI creates a paradox: it opens the door to new possibilities while also introducing risks when it is used without sufficient knowledge and careful consideration.”
This is how the roughly 50-page document begins. It outlines 10 key principles related to the professional and ethical obligations of Quebec’s chartered HR and industrial relations professionals.
Here are the main points to remember.
1) AI never removes professional accountability
This is one of the central themes of the guidelines. Using AI “cannot, under any circumstances, be used to justify errors or excuse their consequences for individuals or organizations.”
Whether it is a communication drafted with ChatGPT, a recommendation based on predictive analytics, or a decision informed by automated screening, the HR professional remains fully accountable for the outcome.
In practice, this means professionals must always be able to explain and justify a recommendation on its own merits, rather than simply relying on the fact that it was generated by AI.
2) Learn how the tool works before using it
The professional body recommends developing a sufficient understanding of an AI tool’s capabilities, limitations and risks before using it.
Professionals must also have enough expertise in the relevant HR field to properly assess the quality of the results produced.
“AI cannot, under any circumstances, compensate for a lack of professional expertise.”
In other words, AI should remain a support tool, not a substitute for professional competence.
The guidelines therefore encourage HR professionals to continuously develop their digital and AI literacy and to consult specialists when their own knowledge is insufficient.
3) Confidentiality is a line that should not be crossed
The message is clear:
“Never enter personal information, confidential data or strategic information into an AI tool unless you are certain that the data is adequately protected.”
The guidelines distinguish between three main scenarios:
- Free public AI tools: there may be no specific contractual agreement protecting the organization, and data may potentially be used to improve or train the service. Sensitive information should therefore never be entered into these tools.
- Paid AI services: contractual terms may exclude the use of customer data for model training, but the information may still leave the organization’s own systems.
- Privately hosted AI models: hosting the model within the organization’s own controlled environment provides the greatest level of control over data.
4) Always verify sources
This addresses the well-known issue of AI hallucinations.
AI systems can produce plausible answers even when they do not have sufficient or reliable information, which can result in inaccurate or even entirely fabricated content.
In short, information generated by AI must be corroborated using reliable sources.
A confident-sounding answer is not the same thing as a verified answer.
5) Professional judgment cannot be delegated
The guidelines are explicit: no professional recommendation should rely entirely on an analysis generated by AI.
Instead, AI should be treated as an assistant that can produce a draft, suggest a direction or identify a possible trend.
The professional must then verify, adapt, qualify, complete and contextualize the result.
The guidelines apply this principle differently depending on the type of AI use:
- For text generation, review the output and ensure compliance with applicable laws, policies and collective bargaining agreements;
- For analytics, cross-check findings against other sources;
- For conversational AI, clearly define which topics and situations the tool is allowed to handle;
- For agentic AI, establish operational limits and continuously monitor its actions and performance.
6) Assess risk before using AI, not after
The level of risk depends heavily on the context.
Using AI to improve the wording of a document is very different from using it to recommend promotions, disciplinary action or terminations.
The larger the volume of data being processed, as with automated resume screening, the more difficult human validation becomes and the greater the need for oversight.
The conclusion is important: after assessing the risks, deciding not to use AI at all can be a perfectly legitimate outcome.
7) Bias requires continuous vigilance, not a one-time audit
AI systems can reproduce biases present in their training data and, in some cases, amplify them.
The guidelines specifically highlight the risk of proxy discrimination.
Even when characteristics protected under Quebec’s Charter of Human Rights and Freedoms are excluded from an algorithm, discrimination can still occur through seemingly neutral variables that correlate with those characteristics, such as postal codes or educational background.
This means that bias cannot be addressed through a single compliance review. Organizations must continue monitoring AI systems over time to detect unintended discriminatory effects.
Explore our trainings
