Name: AI & HR Compliance in Canada: What to Know for 2026
Uploaded: 2026-02-18T19:01:35.588Z
Duration: 41 min 28 s
Description: AI & HR Compliance in Canada: What to Know for 2026

Transcript for "AI & HR Compliance in Canada: What to Know for 2026": Two one go. Hi, everyone. We'll get started shortly in just about a minute. Thanks. Welcome. We'll get started soon. Alright. So here we go. Before we dive in, let's quickly cover a few housekeeping items. This session is being recorded and will be shared with you afterwards. We'll also use the chat feature periodically to share links and key resources. And if you have questions, please submit them in the Q and A panel throughout the session. We reserve time at the end to address as many as possible live. Also, please stay to the end for a giveaway. With that covered, let's move into the introductions. So hi, I'll start first. My name is Kirk Taylor. I'm based in the GTA, and I'm the regional solution consultant leader for Canada at Rippling. And in this role, I work closely with customers who to understand their operational and compliance challenges and help ensure our product continues to evolve in ways that meet the specific needs of Canadian organizations. And with over twenty years in the SaaS space, I'm always delighted for the chance to share best practices that make our work lives better. And with that, I'll pass it over to Erin to introduce herself. Thanks, Kirk. Hi, everyone. I'm Erin Wood. I'm also based in the GTA and will soon be working out of Rippling's new Toronto office. I'm a global compliance lead at Rippling with over twenty years working both for and with the National Payroll Institute, and part of my role here is ensuring that Canadian compliance requirements are considered and embedded into our product features before they're built. I work closely with our product teams to translate regulatory requirements into system design, but I also really value opportunities like this to connect directly with current and prospective clients. To give you a sense of what we'll cover today, here's the agenda. We'll walk through real world examples of AI related compliance mistakes, the kinds that look minor at first, but can quickly turn into financial or regulatory exposure. Then we'll discuss how to structure prompts so they're both safe and operationally useful, kinda like learning how to Google better. Once we learn more about prompting, we'll review taking caution and mitigating risk. And lastly, we'll close with the live q and a, like Kirk said, so please submit those questions throughout, and we'll address as many as we can at the end. So before we move into the content, let's take a moment to reflect. So let's think about your own organization. Does your workplace actively encourage the use of AI? And how frequently are you currently using AI in your HR or payroll compliance work? Now keep those questions in mind as we go through today's discussion. They'll help frame how this applies to your environment. Okay. So AI gone wrong. I'm sure many of us are already positively incorporating AI into our daily work. At the same time, we've also seen situations where AI was used with good intentions but led to real compliance exposure. Let's look at a few examples where AI's confidence and our confidence in it created unintended risk. So our first example here that I'm calling the perfect job posting paradox, so this is around hiring where AI can produce something polished and professional in seconds yet still miss. For this scenario, an HR manager at a Calgary based tech company is launching hiring for a new Toronto office in early twenty twenty six. To move quickly, she asks an AI assistant to draft the job posting. Within seconds, she has a beautiful posting that's clear and professional, well structured, legally neutral in tone, and ready to publish. It's posted across all the major job boards the same day. So what went wrong? The posting did not include the mandatory salary range or the required AI disclosure statement under Ontario's working for Workers For Act that just came into effect January 1. The AI produced a strong posting based on historical norms, but it didn't reflect the most recent legislative changes. So the result, potential exposure to Ministry of Labor fines, which are quite high. They start at $250 and escalate significantly for corporations. There's a risk of human rights tribunal complaints if you're screening language, such as Canadian experience, if that's interpreted as discriminatory. Job listings can be flagged or removed by platforms that are enforcing the Ontario compliance standards. And when questioned internally, the explanation was simple. It was just we used AI to draft it. The real issue wasn't the quality of the writing. It was the assumption that well written means compliance. And AI systems are trained on historical data. They may not capture newly enacted jurisdiction specific requirements, especially those that are only months old. In regulated environments, speed and polish don't replace verification. And when compliance gaps surface, the organization, not the AI, carries the liability. Well, that's so important. pass it over to Kirk now. Yeah, thanks Aaron. That's so important to understand and appreciate. Now let's move to payroll, where small jurisdictional differences can quietly turn into significant compliance exposures when AI is treated as an authority instead of a tool. So here's the scenario. The overtime oversimplification. A national company operates in BC, Alberta, and Ontario. During a quarterly review, the CFO will ask an AI assistant, calculate overtime pay for our hourly employees across three provinces using standard provincial rules. The AI produces a polished table applying a forty four hour weekly overtime threshold to all three provinces. Finance accepts it, payroll processes it. Problem? The rules aren't uniform. BC requires overtime after eight hours in a day or forty hours in a week. Daily overtime was ignored entirely. And in Alberta, well, it requires overtime after eight hours per day or forty four hours per week. The AI only applied the weekly threshold. In Ontario, we used forty four hours per week, so that column appeared correct, reinforcing false confidence. So employees in BC and Alberta were underpaid. No one noticed immediately because the output looked clean, authoritative, and even complete, where weeks later a complaint in BC triggered a broader review. The company faced back pay, penalties, interest, and reputational damage. When asked how the calculations were determined, the answer was we used AI. The real issue wasn't the math. It was the assumption that the standard rules exist across jurisdictions and the failure to validate AI generated legal interpretations before acting on them. And in payroll compliance clean formatting isn't the same as legal accuracy and AI doesn't carry the liability. Starting to see some themes here, so let's go at our third example, third and last example. Let's look at the contractor classification conundrum, where relying on contractual language alone can create significant tax and employment law exposure. So in this situation, a rapidly scaling technology startup wanted to onboard specialized consultants across Canada. To save time and legal costs, the company used an AI tool to draft a master independent contractor agreement. The agreement was comprehensive. It contained clear scope of services, payment terms defined, explicit independent contractor language, no entitlement to benefits. We all know that's why we want contractors. So it was confident in the wording or confident in the wording the company used the AI generated agreement as its primary safeguard against employee classification risk. Months later, CRA conducted a review. Applying its two step approach, first examining the intent of the parties, then assessing the actual working relationship using the common law factors, CRA concluded the workers were actually employees. Why? The company controlled how and when work was performed. The consultants worked exclusively for the start up. They were integrated into core operations. They bore little financial risk and had limited opportunity for profit beyond their fees. All of those items are part of determining that this is an employee, not a contractor. So the contractual language said contractor, but the operational reality said employee. And the result, liability for unlimited income tax withholdings, employer and employee, CPP and EI, interest and penalties. The issue wasn't that the contract was poorly written. It was that classification is determined by facts and substance, not labels. AI can draft a document, but it cannot change the legal reality of a working relationship. And side note, this is one of the top items listed, on the audit by CRA, so we definitely wanna make sure that we're classifying our contractors correctly. Yeah. That's really important. So after seeing how AI can create confident but incomplete outcomes across hiring, payroll, and worker classification, the question becomes how do we reduce the risk while still benefiting from AI? And so one practical safeguard is structured prompt design. So we're probably familiar with this. You've googled it at all or checked out any YouTube videos, but the five c plus model provides a a disciplined way to frame requests so responses are clearer, more accurate, and better aligned to professional use cases. That's the structure we'll focus on today. It builds on the familiar four c framework, context, clarity, constraints, and continuity, and adds a critical fifth element, core intent. At its core, this model mirrors how professionals give strong instructions. We define the situation, specify the task, set parameters, explain the purpose, and clarify how refinement should occur. So here's how it works. Context, it anchors the request. Who are you? Why what environment are you operating in? Clarity, defines the exact deliverable or question. Constraints, it shapes the output, like the format, the tone, or length. And core intent explains why the output matters and how it will be used, allowing the AI to calibrate depth and risk sensitivity. And then the continuity directs iteration. So refine, ask clarifying questions, or flag uncertainty instead of defaulting to confident assumptions. So the plus represents adaptive prompting, explicitly allowing the model to surface uncertainty rather than manufacture confidence. When used together, the five c plus model addresses the most common business failure patterns with AI responses that are polished and plausible yet contextually incorrect. It shifts the interaction from asking a question to briefing a collaborator. Let's apply that framework. In our first scenario, the job posting looked polished, professional, and ready to to publish, but it quietly missed Ontario's new salary transparency and AI disclosure requirements. That's exactly where the structured prompting makes a difference. If you simply ask, what do I need to know about posting a job in Toronto? You'll likely receive a long generic summary. It won't be tailored to your role, your organization, or the specific regulatory risks at play, and it probably won't flag areas that require validation. So using the five c plus approach, the prop becomes far more deliberate. So here we go. As an HR manager drafting a job posting for a senior software engineer role in Toronto, outlined the key legal requirements for an Ontario based posting to ensure compliance with the ESA and human rights code. Use bullet points, keep it under a 120 words, and include a short note on required salary transparency or pay equity statements. This summary will be used to brief the recruiting team before posting. If any legal assumptions are unclear, ask a clarifying question before answering. So now let's break that down. Here we go with the context. We started off with as an HR manager drafting a job posting. Right? The clarity, outline the key legal requirements, and you specify what those requirements are. Constraints, use bullet points, keep it under a 120 words and so on. And then core intent, this summary will be used to brief the recruiting team before posting. And then the continuity factor, if any legal assumptions are unclear, ask a clarifying question before answering. So this structure guides the model toward a focused compliance aware output instead of a broad overview. But even a strong prompt is only one control. AI is a step within a process, not the final authority. Verification, review, and accountability still sit with you and your organization. Okay. So now we wanna talk about where to be cautious, and how to mitigate risk. So as a payroll compliance person, I'm always concerned about risk, and you're obviously trying to avoid any penalties and fines. So strong prompting and using that five c plus improves output quality, but it doesn't eliminate a compliance risk. So where specifically should we be cautious? First, jurisdictional nuance. As we saw in the overtime and the hiring examples, Canadian employment law is layered. It's whether it's federal, provincial, jurisdictional, you know, there are differences everywhere, and sometimes AI groups things together. It doesn't understand the nuances of each of the different jurisdictions. Second, outdated or nonspecific sources. AI models may default to broad principles like general employment rights rather than the precise province specific requirements. Recent legislative updates are especially vulnerable to being missed. I know for years, Ontario EHT had increased their limit during the pandemic to a million dollars as an exemption, and it took about three years for CHAT GPT to catch on to that. That was my test. It finally now knows that it's a million dollars, but it took a number of years to get that nuance. Third, legal liability. If an AI generated summary is followed without validation, the exposure's real. Wage claims, misclassification penalties, fines, any complaints from employees. The output may read like guidance, but the organization remains responsible for the outcome. So how do we mitigate this? Start with human validation. So, you know, treat AI as a briefing draft, but not a final decision tool. For anything jurisdiction specific, you know, ensure it's reviewed by internal HR, payroll compliance, or legal counsel. Make sure someone is doing due diligence to ensure that this the information received that you're gonna put out there is correct. Next, ask for time stamps and sources. Prompt the AI to indicate when the information was last verified. If it was verified ten years ago, you know payroll changes yearly. We wanna make sure that we have, you know, everything updated. And finally, prompt defensively. Be explicit. Instruct the model to reference only the applicable jurisdiction, identify assumptions, clearly distinguish between any confirmed requirements and inferred guidance, and invite it to flag ambiguity rather than resolving it confidently. I like to call it the devil's advocate test, you know, put it in there, you know, what if someone was going to go against this, what would they say? In short, prompting improves precision, but governance review and accountability are what prevent the kinds of failures we've just walked through in our three examples. Perfect. Thanks for sharing that, Aaron. So we've just discussed mitigation strategies, human validation, defensive prompting, and requiring jurisdiction specific sources. Let's apply that thinking to our contractor misclassification example. In that scenario, the issue wasn't the wording of the agreement, it was the assumption that a well drafted contract determines status. In reality, classification depends on how the relationship functions in practice, and provinces assesses that using slightly different lenses. So how do we reduce the risk of AI oversimplifying that analysis? Again, the five C plus model provides structure for us. Rather than asking what's the difference between a contractor and an employee in Canada, which would produce a broad overview, we frame their request deliberately. So here's the context again. Ask HR hiring independent contractors across Ontario, BC and Quebec and from a clarity, summarize how to distinguish between employee and contractor under each province's common law and relevant employment standards. Constraint, present the answer in a three column comparison table, so maybe province, criteria, risk, and keeping that tone professional and advisory. And then for core intent, the goal is to prepare the internal hiring guidelines for the finance team. And then for our continuity, if tax and employment standards loss differ significantly between provinces, note where local legal advice may be needed. That final element is particularly important. By directing the model to highlight divergence and flag where legal advice may be required, you encourage transparency around uncertainty instead of overconfidence. So the objective isn't a definitive legal ruling, it's structured insight with visible risk indicators. So AI can organize a framework and cannot replace the legal and factual assessment required for worker classification. And with that in mind Erin will now expand on key considerations for contractor engagement. Let's review contractor engagement again because this is you know, even with structured prompting, it's an area where caution is essential. And I know in this day and age in 2026, lots of companies are trying to go the contractor route. So we just wanna make sure that we are well versed in what we need to look at. First is cross provincial variability. We've already mentioned this, but that worker classification is not harmonized across provinces. Now federally, is for CRA, but then Quebec has their own Quebec has their own. So as a civil law jurisdiction, they introduce concepts such as economic subordination. There's a few like, even on CRA's website, it's, like, within Quebec and outside Quebec as far as their as far as how they have it set up on CRA. So it is highly fact specific tests. AI can unintentionally flatten those distinct or blend legal frameworks that should remain separate. Second, overconfidence bias. AI systems tend to respond in a definitive tone even when there's legal interpretation that should be nuanced or evolving, and that confidence can obscure ambiguity and influence decision making more than it should. That's what I say about my husband. He speaks with confidence, and then everyone thinks he knows what he's talking about. So he's, That's good. you know, the precursor to AI. Third, the misclassification risk. If there's generic criteria being applied without jurisdictional validation, the exposure is significant. CRA reassessments, provincial tax authority penalties, retroactive source deductions, reputational impact. You know? So how do how do we mitigate this? We wanna start with the local legal validation. We've already mentioned this, but use the AI generated comparisons as a structured starting point, not as a determination. Right? So human first or sorry. AI first, then move to a human. Then force transparency through disclaimers. Prompt the model to explicitly note where criteria may vary by province or where the law is unsettled or unclear. Then require source traceability. Ask for those references to official materials. I really like asking it for links to CRA's website. Finally, use AI for harmonization, not determination. So help identify patterns and differences across provinces. It should not be that the tool declares status or drives payroll treatment without a specialist review. Structured prompting reduces risk, governance, and validation prevent this. So the bottom line, as far as a mitigation strategy, is that AI is a strong starting point for compliance work. It can help you frame better questions, structure information clearly, accelerate initial research. It's such a great place to start. But it's not the final authority, especially when it comes to HR or payroll compliance. The most defensible workflow is very simple. AI for synthesis. Let AI write it all up. Have a human for validation, go through it, and make sure that it is, you know, properly defined and that it makes sense, and then go to HR or legal for approval that this is what how your what's gonna stand for your company. And that sequence preserves efficiency without transferring risk to a tool that doesn't carry liability. So passing it on to Kirk for how Rippling can help. So yeah, absolutely, my team focuses on this all the time. So if you do book a demo with us, you'll get one person from my team or even myself talking about how Ripley can help. But let's go back to those cautionary examples where we walked through. Non compliant job postings over time, sorry, over time misclassifications and contractor misclassifications all share a common theme. Fragmented processes and over reliance on assumptions. The good news is that structured systems reduce that risk. So Rippling helps by embedding compliance into your workflow, not leaving it to memory, manual checks, or AI interpretation alone. For example, in the job posting scenario, Rippling recruiting centralizes the hiring workflows, supports pay transparency requirements, and enforces approvals before roles go live. In the new overtime scenario, Rippling payroll applies province specific rules within the system, helping ensure calculations reflect local thresholds rather than generalized assumptions. And in the contractor classification scenario, Rippling provides structured onboarding and worker type management. And where needed, Rippling Global can actually engage workers compliantly on your behalf. So if your goal is to reduce compliance exposure, not just draft faster, Rippling really brings controls, auditability, jurisdiction specific logic directly into your operational systems. So that's really key. So something else to point out, when we look at the whole rippling solution as a fan here, when you think about all the different offerings that we've got, you can see wrapped around the top of Rippling AI. What's we've been talking about today is AI. Well, next month, Rippling is launching a powerful new AI tool. And it can't right now, I can't share every detail, but here's what it means for you. Instant access to customer customer reports, smarter team management and the ability to spot gaps before they become expensive problems. Think of it as having the right data exactly when you need it. If you want an early look, book a demo with our team using the button in the top right corner. Alright. So I think we got through that pretty quickly. This is our first time doing. this, and I'm so excited get through that pretty quick. this. And so Aaron and I are opening up for live q and a, But even before we do that, I'm just thinking, you know, Aaron, maybe you wanna share, do you wanna think about maybe some examples? You brought up your husband already. Do you have an example of maybe. I was I did I was giving Kirk a breakdown of this, but just to show that AI, although we kinda think of it as everywhere, there are still organizations that are limiting how it can be used, and so that might be some of you guys. Some of you guys might not be allowed to use AI, or you're only allowed to use AI in certain ways. Maybe you have Copilot, or maybe you maybe you only have access to chat, or, you know, you'll so that's one thing. So in his I was telling Kirk in his situation, he's hiring a new employee that is going to be, like, a coordinator, and they're going to need to take meeting notes. And he works in a government adjacent position, and no AI is allowed to record the meetings. So I'm not sure about you guys, but I know in the last few companies I've been at in here at Rippling, we have AI recorders in every meeting, because that's what takes great notes, and it allows you to respond and and and, you know, not miss any nuances. So as he's now hiring, he can see that people are applying, and they're talking about using these type of tools. So it'll only be in the interview process when he's talking to them that he'll say, and now what would you do when you can't use AI? Which kinda seems seems like they're going backwards, but there are still organizations and and and industries that don't allow AI for various reasons within their note taking. And his he's got, you know, people representing each of the governments, all the different provinces. So it's just it's interesting to see. I was wondering if anyone had any, comments or questions, like, similar, that their companies don't allow them to use it. I know it's been the big push in you know, I've been working for payroll software now for a number of years, and it's obviously a big push for us to be using it both internally and then providing AI solutions for customers. Is there Kirk, do you have any have do have any situations where AI has gone wrong for you? Yeah, I can think of a number of times when I was asking something of AI and I was like, no, that's not right. I'm trying to think of a specific example where I asked for something and I double checked it and it was like, no, you were right. That was incorrect. I'm trying to think of something off the top of my head. I don't want something that's coming to me right away. I'm just I'm kinda drawing a blank. we do we have a good we have a good question here, Yeah. from Reese for can or should companies use AI to create uh-oh, I just lost it. To create HR policies and procedures. And that is the perfect situation, Reese, for what we are talking about. 100%. Have it drafted. Right? Draft policies. Draft procedures. Be specific in those prompts. Use that five c plus. Be very specific. HR, like payroll though, does have, legislation backing it for a lot of things, especially like employment standards, labor standards under jurisdiction. And so you need to make sure that you have it drafted, but that is not a final you know, that's not the the thing that's gonna get sent out. You then need a human. Hopefully, you have an HR professional, someone who understands. But if not, then that's where you you take the policy you've already created, and you maybe hire an HR person. It's gonna cost you a lot less. They're not gonna draft it from scratch. They're just gonna, you know, look at that little that policy that you've created to make sure that you're not going against anything that would be, with HR, there's also a lot of, like, best practice that's important to know about as well as, like, legislative backgrounds. Yeah. So definitely the the, you know, AI to draft it, human to to look at it, and validate it, and then pass it on to, like, HR legal to make sure that it can be approved. Yeah, I'm seeing some good questions coming in too. This person actually said it's from Ryan, not a question, but I love using AI for drafts as it comes up with the aspects that I would not have thought of. That's a good point. I think that, Ryan, what you're bringing up there is exactly what we're saying here is if you start with AI and you validate, there's some things that by just the way we're using the 5C, it actually makes you think from a structured way when you prompt it always in that certain way, you think that way, then AI now becomes your sparring partner. Right? It literally becomes the thing that, hey. I didn't think of that. And so that factor that you may have to go and fact check, of course, outside and double check to make sure the facts are correct, But it is really to spark ideas and, I think, get the ball rolling. There's another great question here, from Amir. As someone working in a payroll role, which skills do you think I should learn to avoid losing my job in the future or being replaced by AI? I think that is a great question because we have you know, there's so much talk now about the types of jobs that are gonna be replaced by AI. But if you think about it, computers already replaced what payroll was. So payroll used to be 98% of a data entry job. So when I started in payroll, I'm aging myself, I had a client that was still using, like, green sheet, like, written ledgers to run payroll. It was and using tax tables, like, downloaded from CRA. And then I got them onto a computer, which was great. But and at the time, they were very outdated for doing it that way. So already computers have taken it from that data entry position to more of an analyst position, and that's where payroll is going in the future. You need to be able to look at what AI or a computer system, including Rippling, has generated and be able to spot the the inconsistencies or things that look different. And that's there are systems like Rippling that can help you with those type of things, with great reporting. We have a great reporting tool. But, again, this isn't a sales pitch. So there's lots of systems that have these things, like your payroll system, whatever you're currently using, hopefully, is providing some form of reporting, like exception reporting. That's what, you know, validating your peer reports for your CPP and EI before, you know, you submit your t fours. Those type of things are all analyst jobs, and that is where the future of payroll is going. So hopefully that I I mean, that's my my educated guess, but that is where where payroll is going, you know, as far as what I've seen over the last twenty plus years. That makes sense. There's another one here. I don't think it's really a question, maybe a comment from Margaret. I think it's more of a fear issue and no policies in place for the use of AI. That is hindering these companies to avoid using AI. So you're right. Sometimes there's a policy, sometimes there's not a formal policy, but it's a bit of a fear factor. And what's interesting is I heard Erin mention it at the beginning, like, it's sort of like Googling. Right? I know we use Google, of course, we got so used to Googling stuff and those returning pages, and those pages would sometimes be outdated. So I think of it as like an extension of that, But you're right when it comes to you can't just take the assumptions and apply them as if they're completely true. So I can totally appreciate the fear, but I think where where fear goes away is with education, which is really what we're doing here. And I'm sure you're all educating yourself in different ways. That's why you showed up here. Gathering insight, and I think knowing what to do with your own human judgment is the key to, I think, solving for the fear factor and knowing that you're using caution, of course, so that your judgment is is properly aligned with what your risk factor is. Right? So you're not taking undue risk, but nothing's ever perfect. Right? There's always gonna be something that needs to be validated. And, of course, AI, as we know, needs that extra validation step. What I need AI to do is prompt me to do things like stop sharing so you can actually see our faces instead of just seeing a screen that says live q and a. But we're not there yet. But eventually, we will be for speakers as well, have some some better tools. There is one thing about doing this virtually and we can't see you, I don't know the audience of what we're listening to. I, of course, come at everything from a payroll background. But one thing to point out, I'm hoping there's a number of people here are the who are the decision makers in the room, whether that's for payroll being, you know, payroll manager or CEO, CFO, you know, top level positions, it is important to make sure that you understand what the HR department and what the payroll department, and really legal, are going through as they deal with all of the different nuances, you know, in Canada. Because we are very similar as far as we only have two, reporting bodies with with CRA and, Quebec, but we do also have 14 jurisdictions, federal's its own jurisdiction, that need to be taken into consideration. And if you are a, multi provincial or multi jurisdictional employer, you need to make sure that you have someone that understands that they don't just use AI and they think it's right. And that that is that is definitely something that, you know, we wanna make sure that whether or not you are able to again, not a sales pitch, but book a demo. We do have one comment here, and I just wanna shout out Trish for making a comment because I do know her. And it says, you know, that nice to see me and that she's taken part in some of my stuff, which I remember, and that, she was in one of my discussions at the Niagara Falls conference, which reminds me that Rippling will be at the National Payroll Institute's conference in Montreal this June. And so if any of you are payroll people or if you are someone who has a payroll person reporting to them that's going to go to that conference, please come see Rippling. We will have our booth there, and we'd love to show you a demo. Alright. We got about eleven minutes left. Are we gonna use the whole time, Aaron? I'm wondering if there's any more comments. on well, so we've answered all of the questions we've had. Looks like we have. We still have a number of people here, though. So before before we just kick everyone out, I would like to see if any more questions come in. We'll give it like maybe another two minutes, and then if not, we can thank everyone for joining us. Sure. Looks like Ryan just added another one here. You can ask for AI to provide publicly available references or sources as well to confirm the information it pulls is correct. That's a good call out, Ryan, because really when you think about the source, that's really where AI should be pulling from and if you go to that actual source, then you can actually verify there. It's important to do that extra double check. I also found that, yeah, there's dates on things. So just like when you Google something, like, if that source was referred to was 2022 or something, like, completely, you know, probably dated now to the point where it's inaccurate. So it's important that it tells us the source, but I think it's also us making sure that it's an up to date reflection of what could be accurate because things do change all the time. There is a request here to show the slide with the five Cs, and so I will do that. I think this is the one you're looking for, and this, for sure, is the takeaway from here. And the thing to think about when you're doing five c plus is this is also how when Google became again, aging myself. But when Google became a thing and we all started googling everything, that learning how to properly Google and now Google has AI. So, you know, over, like, like, right within it, that if you write out your Google prompt correctly, you're gonna get a much more accurate response than if you don't. So in my world where I'm mostly googling payroll legislation, if I do a very generic prompt about, you know, workers' compensation, let's say, I'm gonna get a whole bunch of information that isn't specific to the province I'm looking for or the reason I'm googling. Or, like, if like, I've I've just been working on, you know, holiday pay and and, vacation pay and and those type of things. If I type those in, it's interesting. If just type in, like, holiday pay in Canada or even in Ontario, you get everything from other providers. You get what all of the service providers are posting. And then you get you don't get necessarily what the government is saying. So it's really important in your prompt to be direct. Right? You need to define who and why. You wanna define what you want. So, like, who you are, like, as a payroll person who's creating a policy, you know, and then do your clarity, like, define what you want, give some constraints. One of my constraints is, often use use a government website. Like, I ask it to to give me a government website, so that I don't get all sorts of tax blogs. Yeah. It's a good one. I saw one here that, Elizabeth asked about. You talk a lot about what to avoid, but what most excites you about the AI development in the last year? And, of course, other than those AI profile takes that that that still scare me. That's what she says. I yeah. I've seen a lot more of those lately too. You know, I think what it is for me that excites me is just I feel like being a prompt engineer, which is kind of what we're doing when you have the five c's here, it helps me structure the way I see problems. So I'm not just only relying on AI to think we'll call it in a structured way. I'm excited about how it's helping me rethink things and how I solve for problems. And and then so I guess really what it turns out to me is, like, being able to do things quicker. Because if I know how to use a drill, for example, I can definitely drill and get things done way faster than if I was just doing it all by hand. I think of it like a power tool. So whatever you put in your hand that you now know how to use, you can build that particular thing that much faster. So whether it's, of course, building your resume, if you're applying for a job, or if you are for example, in my work, there's a lot of calls that are recorded, and I can now listen to calls without having to listen to the whole call. Alright? I take the transcript. I can actually digest what's there. I can get the highlights for it. I think of it what excites me is that I can be me supercharged. Right? So it becomes not a substitute. It just makes me better. That's what I'm excited about. How about yourself, Aaron? Yeah. One of the things that you just pointed out with that get the transcript, I am not an auditory learner. I am a whatever it's called when you read it. Like, you can tell me as much as you want. I'm not hearing it, but I need to read it in order to learn it. And so having transcripting available now has really helped speed up my learning process when new things are on the table. And so that is like, that part of AI has been very beneficial. I'm easily distracted if there's a noise. You know? Squirrel. That type of so for me, that's like, I the accessibility of AI from, like like, the the broad scope of accessibility, that it is definitely helpful for, you know, anyone like, think think that if you can't if you are blind and can't read, but now you can have it transcribed or you can have it, you know, transcribed and and then read out to you. Those type of things, the accessibility of it really does, impress me, and I think it it overall will make, things better for people. Now I do have children, and so, you know, we understand, like, in school, I know it's a bit of a problem. But I have, one son did a college course where they were forced to use AI in their communications course so that they could better write, like, their communications. It was business communication, and they had to use it, but the same thing. They had AI generated, and then they, as the human, had to go and validate it. So AI was helping them, but they were like, the the college course was teaching them how to properly use AI. Because I saw one of the other questions here, you know, what do we recommend in terms of adding AI legal stipulations and incorporating those into company policies? There's a lot of, like, legal stuff surrounding AI that is in Canada. We haven't fully come out with, like, policies that say what you can and can't do. And so that is something we're talking about, like, specifically for your organization, having that AI legal stipulation. You need to be I would you know, again, I would write it, but then I would go to a legal person and have them vet it. Agreed. Because yeah. Because and and we definitely wanna have some kind like, your companies should come up with what is acceptable in your organizations and that what what can and can't be done so that people are not taking things too far or not or not getting the benefit of actually using it. Absolutely. Absolutely. So we've got four minutes left. I don't see any more questions coming in, but thank you so much for all that stayed to the end and fully engaged. As promised, if you are the lucky winner, look out from our marketing team mid next week to see if, you're one of the lucky winners for staying with us. And thank you so much for attending our first AI and compliance session today. Thanks, Erin. Thanks, Kirk. Thanks, everyone, for joining us. Great. Take care. Bye bye.