Risk vs. reward: How to balance global hiring and compliance

[00:00:09] Mallory Hendry: Hello everyone and welcome. I’m Mallory Hendry, Specialist Content Manager with Key Media. Thank you for joining us for today’s webinar, Risk Versus Reward: How to Balance Global Hiring and Compliance. Today we’re joined by Adrian Drew, VP of Product Counsel and Head of Legal for GIA at Globalization Partners. With more than 15 years of experience advising entrepreneurs and global employers, Adrian is ready to delve into this timely issue, including how to access global talent while staying compliant with local laws, why AI document accuracy is a growing legal risk, and how to close the compliance gap by combining human expertise with compliant AI. At the end of the presentation, Adrian will participate in a question and answer period, so be sure to type any questions you have into the Q&A box within the webinar software. I’ll turn things over to Adrienne now to begin the presentation. Take it away.

[00:01:04] Adrienne Drew: Thanks, Mallory. Hi everybody, thank you for being here. As Mallory said, my name is Adrienne Drew. I’m an attorney based out of San Diego, California. I have been with Globalization Partners for nearly 10 years, and I’ve spent many of those years building the company’s entity footprint working on our employment practices, building tech products in the employment and compliance space, and really focusing on the global risks of hiring and employment around the world. I started my career as a lawyer in private practice as a litigator in California and in Massachusetts, and I am so thrilled to have been able to take my career international as I’ve been doing this global work. If you’re not familiar with Globalization Partners, GP is a global employment company that provides, among other things, AI-enabled employer of record technology. An employer of record helps companies expand their workforce across borders quickly and compliantly and without the friction of setting up legal entities. So we’re experts in global employment, and I’m really happy that you’re here with us today.

[00:02:07] Adrienne Drew: Mallory went over the agenda, and I just want to sort of— you’ve all probably noticed that AI has sort of moved from being a buzzword to kind of being everyone’s initiative. Everyone’s got a budget line for AI. And for workforce compliance, AI is a really big factor. But there’s a lot of legislation that’s starting to tighten around AI-driven hiring and using AI in employment practices. And so we’re here to talk about that too. So as Mallory mentioned, in this webinar, we’re going to cover accessing global talent while staying compliant with local labor laws. That’s a big topic. We could go on forever. We’re also going to talk about why 61% of corporate leaders feel slightly distressed over AI document accuracy, meaning in their employment documentation, the use of AI. And how you can protect your company from any liability or exposure when you’re using AI in your work. And then we’re also going to talk about solving common HR challenges by combining our human experience with the amazing technology that is AI.

[00:03:03] Adrienne Drew: Before we dive deeper, though, I do have a couple of quick poll questions here and then throughout the presentation. Question number 1, does your company currently have a global presence? Go ahead and vote. And we’ll see where we are globally. I’m gonna say yes for mine. We are in 180 countries, so we’re the 30+.

[00:03:40] Adrienne Drew: Couple more seconds to vote.

[00:03:44] Adrienne Drew: Okay, so we’ve got a good, a good split here. So, yes, we’re in 1 to 10. Yes, 11 to 20. Some people don’t have a global presence, but maybe you’re thinking about it. To be honest, even not having a global presence, if you have a multi-state presence in the US, that has similar, similar problems, right? Different states have different rules, particularly around compliance and AI. So, Even if you’re just in the US, if you’re in multiple states, the information that we’re going to talk about today is really relevant to you. So it’s good we’re kind of everywhere. I love the 30+ countries, not just me, but somebody else. That’s great. Okay. So GP does what we call an AI at Work study.

[00:04:24] Adrienne Drew: And when we looked back at workforce data in 2025, we discovered an interesting contradiction. So AI has made global hiring more seamless than ever. But it’s also become key to how we find talent and we onboard across borders. I think everyone sort of in the middle of maybe the end of 2025 coming into 2026 really started to embrace AI. We use it in our personal lives. And we’re starting to use it much more in our professional lives, particularly if our companies endorse it. And 41% of executives that we talked to when we built our AI at Work report felt that AI has made accessing global talent easier. On the other hand, though, everyone still has a little bit of unease. And those same leaders felt that AI might be creating new risks that they either weren’t aware of or maybe they weren’t prepared to handle. So today we’re going to talk about how to kind of release a little bit of that friction, starting with how the legal landscape has changed over the last 12 and 18 months as it relates to AI and how to use AI in your hiring.

[00:05:24] Adrienne Drew: So the legal landscape has really changed. If you feel like legal compliance and keeping up with it has become a full-time job for you or for a whole department, it’s definitely the case. Baseline regulations for managing a multi-jurisdictional team evolve constantly. And if we look broadly at labor laws, we see a lot of new and changing frameworks. And in the US in particular, we’ve seen the EEOC and the Wage and Hour Division really taking some strong positions, not just on AI and hiring, but in all sorts of classification and wage and hour related things across state laws in the US. So, There’s federal changes at the Wage and Hour Division. There’s state agencies that look at misclassification and do wage and hour tracking. If you’ve been on the receiving end of a wage and hour lawsuit in the US, you know that that’s a really tricky area of compliance and it takes a lot of energy.

[00:06:14] Adrienne Drew: Also in North America, if you look at Canada, there’s a newly implemented right to disconnect. And there are some stricter presumptions around employee status, and that really changes the workplace dynamics. Employees in Canada are, I think, getting more and more focused on enforcing their rights, particularly around the time of offboarding. And so we see a lot of HR teams sort of think, okay, what are we looking at when we try to onboard new tools? What is the AI regulations going to be like in a year so the tool that I onboard now to help me is going to still run on the right side of the law? Not just as it relates to recruiting and hiring, but maybe for training or for employee lifecycle issues. You know, the laws are changing quickly. Are my tools going to keep up? Are my processes going to keep up? These are questions that HR departments are having to look at from a new lens because of the speed with which we’re seeing these changes come out.

[00:07:07] Adrienne Drew: But a big change is AI and employment decisions. And historically, it’s news to no one that, you know, artificial intelligence has been around for a long time. For a long time, there was what we call predictive AI, which was using logarithmic information to predict what was going to come next. And that was something that was on was in technology and it was great for sorting data. It was great for information like that. But, you know, when we saw generative AI come out, that means that the AI is generating a new answer based on math data mapping. That kind of was a whole new world. And people thought, well, this would be great for high-volume hiring. This might be great for looking at resumes, sorting through applicant tracking systems. That’s where the volume is. And that’s also where a lot of the early regulations sat. And a lot of company policies also focused on, okay, we’re going to use AI to sort resumes. Are we not? Are we going to use AI to do interview videos? Things like that. When we got to generative AI, moving just beyond the predictive nature of AI, people started to focus on the interviewing. But we’re seeing the trend is that the legislators, the regulation, and really focus inside these companies is moving away from just the point of hire into every step of the employment lifecycle. So regulators in the US and in the EU and in the UK, they’re looking at how predictive and generative AI tools are used at every step of the lifecycle. And it isn’t always clear which type of AI is the target, but it is clear that the crucial point the regulators are focused on is who’s making the decisions, right?

[00:08:42] Adrienne Drew: So automated decision-making or machine-led decision-making, I think, is something most of us have heard of at this point, which is something that regulators across countries are concerned about. The idea that a machine would make a choice about someone’s job kind of gives everyone the ick. And so regulators are really trying to ensure that humans remain in the loop and that AI remains in a non-decision-making role. So when you’re looking at your processes and your tools, if an AI tool is helping you make a significant decision that has an impact on an employee’s livelihood, could be positive or could be negative. That’s an area where you might want to be aware. You may see regulation coming into that decision-making process. This would be automated onboardings, promotion decisions, awards, locations, benefits, all the way down to automated decision-making about grooming or about time cards or staffing. AI-driven performance improvement plans could be problematic. Any sort of logarithmic inputs on employment-related outputs. So decision-making is the crux. What, regardless of the location and regardless of the type of AI, I think it’s really fair to be aware of. Am I making this decision? Is a human making this decision? Or is a machine making this decision?

[00:10:01] Adrienne Drew: And so just wanted to dive in a little bit into the way that, you know, the European Union and the US are kind of starting to regulate this. It’s new technology, which takes a while to regulate. But we’re seeing 2 different approaches, right? So on the one side, you have global frameworks, so like the EU AI Act. And that takes a very broad approach across all industries. We all know that the EU likes to regulate broadly. It did so with GDPR, and it’s kind of trying to do the same thing here with AI. And on the other side of the pond in the US, as is our custom, we don’t really have a unified federal law yet. There are attempts. There’s the AI Civil Rights Act, but it’s not likely to pass anytime soon. Our current administration has other priorities, and it’s really hard to get anything through any government in the US, to be honest, particularly at a federal level. And so what often happens in the US is we get state-by-state patchwork laws that kind of pop up in a grassroots fashion before we get a federal overlay. And what we’re seeing now is that states like California and Colorado and Illinois, they’re setting precedents and trying to feel out a regulatory landscape around AI compliance. Colorado, for example, just passed a framework that focuses heavily on automated decision-making technology, no surprise, and regulating that. And the theme there is that they want all of any localized AI use really has 3 main pillars: transparency, notice and auditing.

[00:11:34] Adrienne Drew: So that means that regulators want to ensure that the tools that are being audited by the companies that are using them, auditing them for bias or for discrimination or for unintended output. Bias is a big concern. I know of the EEOC, the commissioner has really talked about that, about wanting to make sure that any AI that is used in HR does not have bias. And bias can actually be unintentional. He told a story of— I saw the EEOC commissioner speak at a conference maybe a year and a half ago, and he told a story of a hedge fund that was using AI in the best of intentions. And they had gone through their hiring classes for the last 10 years and given the AI the resumes and said, all right, what is the most likely attributes of people that are going to succeed at our hedge fund? Because these are the people that have succeeded. What are the common traits between these people? And kind of humorously, the AI spit out certain courses of study and certain interests, but also the fact that they were likely to have played lacrosse in high school, which is not really something that is true, would have introduced bias if they had started looking for lacrosse players going forward. It was just a coincidence. So causation, correlation, things like that nuance is not picked up necessarily in the AI. So that was an unintended bias that they could have built in that they had to be on the lookout for. So AI tools that maybe target job ads or screen applicants based on recent graduation dates or majors that are named something that indicates a recent graduation date, that could exclude an older candidate and it could lead to indirect age discrimination. Examples like that are things that you want to make sure that your AI doesn’t have any bias built in.

[00:13:27] Adrienne Drew: So, what does it mean for you if you’re trying to expand overseas or across jurisdictions that there are different regulations in all these different areas, right? So, if you want to onboard an AI tool that is going to work really well if you have your headquarters in one state or in one country, but don’t realize that other countries that you want to hire into, or even other states in the US, might limit use of that AI tool. What it might not be regulated in one place, but it could be regulated in another. What are you supposed to do there? So, as we work through our roadmap today, we’re going to look at how to build kind of a firewall against that, ensuring that you can use AI while safeguarding against the legal risks that might pop up from state to state. So you don’t have to say, I can use this tool in one jurisdiction, but not in another. Having a global solution really is the answer here.

[00:14:20] Adrienne Drew: So, second poll question is, how is your company currently drafting employment documents? A little bit of a pivot here, but the lead is, I know that some people are looking to use AI to help with their drafting. And so, as we kind of pivot into using AI globally for your employment practices, I want to talk about what are we currently doing. Are we using AI tools right now? Do we have standard templates? What’s everyone doing with their current documents? And if you don’t have an international presence, you can think of a multi-state presence when answering this question.

[00:15:26] Mallory Hendry: All right.

[00:15:27] Adrienne Drew: Nobody is drafting using AI. Sort of nice to see. All right. So we have standard templates. I love that. Everything goes through country-specific legal counsel. That’s also a good strategy. A mix of AI tools and internal review. That’s what I expected. I think a lot of people are doing that. And not having a standardized process. That’s really honest, honestly. To— for a lot of companies, they don’t have a standardized process because they’re moving too quickly, because they have such skilled HR people that they just know what to do and they have that information in their heads. So none of this is really surprising, but it’s all great to see. And those of you that are using AI, combining it with human review, I think, is a really strong strategy. And we’ll talk about why that is.

[00:16:10] Adrienne Drew: And so I think, I think the main kind of trick or main thing to avoid is leaning too heavily on AI as an expert too early, particularly when it comes to things that have a legal consequence, right? So you don’t want to have a false sense of security that comes out of using a tool that is a general off-the-shelf LLM. The output can look really good. It can be beautifully formatted. It can be quite convincing. But it can generate a little bit of false confidence because you don’t know what you don’t know. And so if the LLM is missing something, maybe it’s not in there, but you wouldn’t know to look for it because it’s in a country you’re not familiar with or something like that. So you might think, hey, I got a great contract from Brazil from my off-the-shelf LLM. So like, I’m going to go ahead and use this in Brazil. But it is possible that that contract doesn’t have the right answers in it. So maybe the prompt was too narrow, or the LLM didn’t have the right behavior to elaborate on the question. So no matter how advanced the technology gets, I think every multi-jurisdictional company, there’s still hurdles that, as we all know, AI can’t solve on their own.

[00:17:16] Adrienne Drew: So I’m just going to talk about 3 areas where we’ve seen people still get into trouble when they’re trying to expand overseas and where AI can help, but having a general practice is not as strong as having a country-specific practice. So one of those is worker classification. I see a lot of companies that are looking to hire internationally or in a jurisdiction where they don’t have a registered office opting to onboard people as contractors. And this happened on a flight 2 weeks ago. I was sitting next to someone who worked in solar and we were flying from San Diego to Germany and I told him what I did and he said, oh, We hire all of our international people as contractors. And I said, yeah, I get that a lot. I know that that’s something that a lot of companies do. You know, I’ve heard companies that have 300 independent contractors in foreign jurisdictions because they just haven’t gotten their arms around how to hire them directly. So it happens very often, and depending on the company’s risk profile, they choose to do that for a while. But a lot of countries, because of, I think, the rise of the gig economy, because of the rise of platform-style workers are trying to sort out who is an employee and who is really an independent contractor.

[00:18:29] Adrienne Drew: I know in the EU they had the platform worker initiative, which really tried to focus on that. And we’re even seeing it in the US. Different tests are used in different states to determine if you’re an independent contractor or an employee. And that’s also the case in Canada, I think in response, like I said, to the rise of the gig workers. And so If you’re using an AI tool to help draft an independent contractor agreement or an employment agreement, you want to make sure that you have properly classified your worker up front, and maybe you’re not relying on an AI tool to help you classify that worker, but that you also have a sense of what the worker’s day-to-day reality looks like, not just what it looks like on paper. So a contract either created by, you know, your organization or someone working for your organization might say this person’s an independent contractor, they have full autonomy, and they use their own tools and set their own hours, but then in reality, that person shows up to either their home office or a WeWork space with a laptop that the company gave them, and they work from 9 to 5, and they take a lunch, and they’re called to account on company hours. Despite what it says on paper, there might be a finding that that person really should have been an employee, and that would come with all of the sort of the back pay and the consequences that would come with that. So Because the law in all of these jurisdictions looks in one way or another at the actual behavior and the financial relationship, what the worker does and how they spend their day. They look more to that than the title or the form of the contract when they’re deciding if a worker is properly classified. So you want to make sure that if you’re using AI to either help classify workers or to help write an employment contract or an independent contractor agreement, that you’re not relying on the AI’s interpretation of what that test might be.

[00:20:20] Adrienne Drew: And so a second common area of complexity in a multi-jurisdictional workforce is statutory benefits and mandatory entitlements. This is maybe more of a challenge when you’re going international from the US, but even within the US from state to state, statutory entitlements actually are starting to change much more than they ever were before. So there are legal expectations regarding what an employer has to provide, what leave is available, There are some right-to-work jurisdictions in the US. There are guarantees for parental leave in certain states, but not in others. There’s mandatory sick leave in certain states, but not in others. As well as in international jurisdictions, most of them have different benefits and different entitlements across borders. And still yet, we see companies using one employment contract for every employee around the world. Sometimes they’ll just fire up an LLM and say, can you give me an employment contract that I can hire someone with? In this country, they’re going to get an output, then they’ll then just change the country name and use that output in a different country. So trying to implement one set of documents globally, and that almost always misses the local nuances, right? So in LATAM and Asia, there’s like a 13-month salary, there’s guaranteed vacation pay, there’s notice, there’s different restrictions on rolling over leave. All of these things can get quite granular, as you probably know if you have a multi-jurisdictional workforce. And these are the things that are likely to get missed if you’re using either a standardized template or generating a template or even trying to reproduce a template using AI without a human review.

[00:21:56] Adrienne Drew: In Europe, there are also a lot of benefits that are provided by the— that are nationalized, but also you can add value on top of that. So you can offer supplementary benefits. There are paid leave minimums and there’s also severance rules and things like this in different countries that you want to make sure are built into your employment practices from jurisdiction to jurisdiction and that you pay attention when you move across the border. But trying to run a global team on sort of one version of a contract or one version of a policy can be really hard. And I have a lot of sympathy for HR who’s trying to explain that to management, particularly in a US-based company, to say we can’t implement a global non-compete policy because we’re not, you know, global, it’s not enforceable globally, or we can’t implement a rule that you can’t roll over your vacation because in every country we’re operating in, that is different. I’ve seen a lot of American-based companies struggle with that nuance because America is an at-will employment jurisdiction and we don’t have paid leave at the federal level. So things like that can be a hurdle for HR teams when they are informing their business colleagues what the limitations they’re working on. So it’s really important, I think, for the HR teams to be leaders with that discussion and say we need to be localized because there are different practices everywhere and we can’t have a standard global approach.

[00:23:18] Adrienne Drew: A third— these are just 3, there are myriad, but these are kind of 3 big ones that we see people struggling with. So data privacy and cross-border data flows. Hiring across international borders means you’re moving sensitive data across borders. That could be payroll information, it could be It could be birth dates, it could be health information, it could be political affiliations, or if you’re subject to a collective bargaining agreement overseas, it could be bargaining information, banking information, addresses, performance records. So opening an office or hiring overseas means you have to start to think about, all right, are we compliant with GDPR? Are we compliant with the equivalents in Singapore and Australia and Brazil? And there’s sort of additional privacy laws. Are we in China? Because there’s new privacy laws in China as well. And what do we have to do to get employee consent to move their data to the US, to work with our vendors? Do we have processors and subprocessors? Do we need additional contractual safeguards? Often when opening up international offices, you might have to get like a data supervisor based in Europe, and have data localization practices and make sure that your vendors also have those. At this point, you know, we’re almost 8 years past, more than 8 years past GDPR coming online. And I think a lot of these things have become common practice and things like data sharing agreements and processor agreements, you probably are all familiar with. But if you’re expanding into a new country, you just want to make sure that you double-check those requirements and make sure that you’re still up to speed with what’s going on locally.

[00:24:55] Adrienne Drew: It was an issue for some companies after Brexit because GDPR was no longer applicable in the UK and they came up with their own regulations. Similarly, when the EU-US Privacy Shield was invalidated, everyone sort of had to scramble to come up with a new regime under which we could transfer data between the EU and the US. So every once in a while there’s a little wrinkle that comes along that it bears good practice to kind of revisit that just to make sure that everyone’s in compliance with their data movement processes. But you don’t want to slow down business growth, right? Compliance is great, but you can’t be steering the ship. We’re all working to grow our businesses and having these commercial profiles. So how do we balance pushing this compliance agenda without slowing down business?

[00:25:46] Adrienne Drew: And I think, I think having a compliance mindset built in rather than in peaks and valleys can help do that because it’s kind of humming in the background. As opposed to cresting and then grinding the business to a halt, right? So speed versus compliance. When we talk about with leadership teams at companies that are growing quickly, speed is imperative, right? Speed is always something that the business leaders are really critical on, but they care about compliance. They don’t want to move so fast that they aren’t mindful about the quality of their business and the quality of their product. They want to capture market share. They want top talent, but they also know they have to do it right. So there’s a little bit of a tug of war, but rather than it be a tug of war, what you want is you want to have some coexisting. You want to have the right infrastructure, the right local expertise, know that you’re looking for the right things, and know that your systems can be flexible if you need to tack on a new country or if you need to hire someone quickly in a jurisdiction that you’re not familiar with.

[00:26:46] Adrienne Drew: In all of this, think about, all right, well, these laws are kind of changing and some of the international labor laws are tightening. I was in Brussels about a month ago talking to some of the members of the European Parliament, including their employment members, and they are really focused on protecting workers’ rights. And so we’re going to continue to see European laws that put to the front the rights of labor, that are protective of labor. And I don’t expect to see that changing, even though the EU is simultaneously pushing the common market. They’re pushing the EU Inc. initiative, which is trying to allow startups and scale-ups to really take root in the EU. So there’s still a tension, but it’s very still employer protective. So those of us that work in Europe are going to continue to see pressure on employees being protected by their employers.

[00:27:41] Adrienne Drew: And also that’s happening here in North America, right? So we can talk about how it’s manifesting right now in Canada. A lot of US companies expanded to Canada early in their expansion strategy because of proximity and because they— the time zones line up and the language lines up. And we presume maybe that the Canadian employment practices are similar to what they are in the US, but it can be quite a shock to learn that they are not, right? So Canada operates more similar to Europe really than to the US, including their strict changes at the front end of the hiring cycle that can slow down onboarding. And that can be hard if you’re really trying to move quickly, right? So worker classification is one of those things. If you are— everyone, there’s a presumption of being an employee if you are paid in Canada. And so hiring people in Canada as an independent contractor is harder to do because there’s a presumption working against you. So you have to kind of prove that they are not in order to classify them as an independent contractor. And they have some built-in techniques to share tax data to let the regulators know when someone might be improperly classified. So I think enforcement is actually getting easier there as well. And then also, there are— to make a change to an existing employment agreement in Canada, I’ll talk about this a little bit later as well, requires the employee to agree. And if any change to the contract sometimes requires consideration, which means you have to pay them a bonus if you want to change anything good or bad in their employment contract. That’s not something that US employers are really used to doing. And so when it comes to managing expectations of your business stakeholders so that you can move quickly without compliance and regulatory friction, these are the types of things that you really want to be on top of.

[00:29:31] Adrienne Drew: Same in LATAM, which is also highly desirable. Lots of great talent coming out of LATAM. The time zones align. The people are— remote work has really opened up certain markets, and I’ve seen a lot of hiring into Colombia, but also in Brazil. And it can be hard to hire in Brazil because there’s a local entity requirement, meaning you can’t hire an employee in Brazil without having a corporate tax ID, which is super hard to do. And if you want to bypass this to save time and onboard someone, either by treating them as a contractor or using sort of like an NRE or some other strategy, there’s some tax liabilities and it can get kind of messy kind of quickly. There’s also a pay transparency law that requires providing salary and diversity data twice a year. So Knowing these things, there’s not a reason not to employ in Brazil, but it’s a reason to think really carefully about how you’re going to do it rather than just saying, okay, we’re going to do an independent contractor and take on the liability about that. You want to look at what do I have to do before I hire this person?

[00:30:37] Adrienne Drew: We already talked a little bit about Europe, and there are a lot of regulations that are coming that are quite employee protective. So there are supply chain requirements about warehouses. There are vendor requirements. So they’re starting to try to make end-user companies responsible for downstream employment conditions in their supply chain. So there’s a ban on any products that can be tied to forced labor starting in 2027, which regardless of whether you’re employing those people, if they’re in your supply chain, they’re sort of imputing responsibility for those workers to the end consumer. So those are things which, if you work in manufacturing, if you work in retail goods, that’s something you want to definitely think about. And if you don’t have the ability to look down your supply chain or to hire directly, you kind of want to think about those things that you want to— that you need to onboard, that you need to recommend to your organization.

[00:31:39] Adrienne Drew: Asia is a fascinating, I think, jurisdiction right now. In China, their social insurance compliance is now highly digitized and they have online filings and automated compliance and payroll monitoring, which makes it a lot, I think, easier for regulators there to determine if there’s improperly classified workers or if there are ambiguities or errors in payroll records. And it makes auditing a lot easier. So we’re seeing a lot more audits coming out of Asia. But again, Asia’s really desirable because it has a huge market. And it has a lot of labor that is accessible and I think eager to support. But making your filings late or doing them inaccurately can come with penalties. And there’s also caps and quotas. So only X percent of your workforce can be subject to a visa, or only X percent of your workforce in China can be outsourced, things like this. So engaging with compliance early while moving into these new countries, I think, is a really important thing to do.

[00:32:43] Adrienne Drew: India is a country that a lot of companies have been outsourcing to for a long time. India is similar to the US in which there are states and each of them has their own laws. And for a long time, India has had more than, I guess, 25— no, 29 labor codes. And they’re endeavoring now to consolidate those into 4 main labor codes. But that can be hard to do. But what they’re doing in order to make everyone happy is they’re really kind of, I think, increasing penalties for noncompliance. And so there is a real focus on, okay, we’re going to standardize what we can do because we want to make ourselves appealing to the global marketplace. But we also want to continue to recognize our local labor practices and protect our workers. And so you see those 2 themes, common market, open market. We want to bring in jobs, but we also want to make sure that our traditional values persist across this new global market, and that you’re going to see that coming out of Europe, and you’re going to see it coming out of Asia, and you’re going to see it coming out of India.

[00:33:44] Adrienne Drew: And Singapore is another good example. Singapore is about to approve the Workplace Fairness Act, which is an anti-discrimination law. And so there’s a lot more enforcement around local discrimination in hiring. And I think all of these laws are something that when you’re Looking at hiring internationally, you have to think, okay, we just can’t export our US employment dynamic into these new countries. We really want to get ahead of where are we going? What are we subject to? How is it going to benefit our organization? How do we respond to our leadership’s need for speed with a compliant solution here?

[00:34:25] Adrienne Drew: Okay, I think this is our last poll question. How confident are you That you could handle a sudden legal stress in an international market. Those of you that have international markets, and those of you that don’t, we’ll try a different state.

[00:35:17] Adrienne Drew: Couple more seconds for voting. All right. Very confident. One of you, congratulations.

[00:35:29] Mallory Hendry: I love that.

[00:35:29] Adrienne Drew: Moderately confident. I think that’s maybe half the group is moderately confident and have blind spots. You know, I think that’s pretty standard. It’s almost impossible not to have blind spots. I’m not very confident and unprepared. Also pretty confident. I think that that speaks to that tension that we’re talking about where the business wants to move quickly. And sometimes they do so maybe in service to one goal rather than to another. And so the ability to say, you know, if I have a legal issue pop up in France, I’m 100% sure that we can handle that, you know, I’m not so sure that everyone can say that. And I think that that’s perfectly fine. But it’s always something to just work for, to work towards, and know that you’re not alone in that journey. But I think, I think that that’s not uncommon, right?

[00:36:20] Adrienne Drew: So the data point that I mentioned briefly from briefly ago was from the AI at Work report. 41% of executives say that AI is making global talent easier. It also creates risks they aren’t ready for. So that kind of aligns with our poll results, right? Like nearly half of everyone says there’s risks out there that I’m not sure that I’m ready for.

[00:36:38] Mallory Hendry: Yeah.

[00:36:40] Adrienne Drew: 61% of leaders also admit to feeling some moral distress over the accuracy of AI-generated documents and AI outputs and work. And what does this unpreparedness actually look like in practice when you’re scaling a global workforce? I think, I think it can sort of feel like it’s really exciting and exhilarating. Everything feels really great at first. You have your candidate, you’re able to expand, you have great talent, and you’re thinking, gosh, the The horizons of the world have expanded so quickly, and there’s so much great labor out there. The talent wants to move fast. You want to move fast. You can get contracts and documents and things like that from online or from a databank. But then you sort of think, oh God, what are we doing? I don’t know what I don’t know in this circumstance, right? So that could be something that maybe some of us lose sleep over, like those of you that maybe said, I’m not really sure that I can handle a hiccup in In a certain country. I do see a question. I’m going to get questions at the end, so keep asking and we’ll tack them on at the very end.

[00:37:41] Adrienne Drew: So, back to using kind of AI decision-making to help you scale. I think it does serve one of those needs, which is the speed, right? So drafting your contracts, learning about overseas models, figuring out, you know, does this country have paid leave or do they not, right? But there’s a limit to that ability of those sources to help you. And I think that knowing where that limit is, is when you can really kind of move forward with wisdom and with professional expertise. So an example would be obviously an offboarding. Those are some of the hardest things that we ever do. And if you’re in a really protective labor market like France, or if you’re in Canada and the working relationship deteriorates, The strength of your documentation is really going to be critical in those moments. And so you want to make sure that you go and you, if you’ve expanded quickly, if you’ve got independent contractors that you think might be misclassified, if you’ve got people in place with a contract that you’re thinking, I don’t know if this is really going to hold up if they push it at the time of offboarding, go back and take a look at those and see, does this need to be revised, right? There might be a notice period that’s left out.

[00:38:54] Adrienne Drew: In Canada in particular, and we’ll talk about this in a minute, There is some, some strict language that can— it’s kind of a strange dynamic, but if you put some offboarding terms in your employment contract in Canada and say, this is your notice and this is how much you’re going to get paid and blah, blah, blah, if you’re compliant with that, you can adhere to those terms. But if one of those is wrong, the employee at the time of offboarding can seek to invalidate the entire contract and then revert to the baseline statutory provisions, which tend to be higher than what they would have negotiated for in the contract itself. So we’ll talk about that in a second. But that’s an area where you want to go back and make sure your contracts, regardless of how you got them, if you got them through ChatGPT, or you inherited them from the person that worked your job before you, or you brought them from somewhere else, or, you know, translated from another jurisdiction, just make sure that they’re really compliant in the jurisdictions where you are.

[00:39:50] Mallory Hendry: Okay.

[00:39:51] Adrienne Drew: And then we’re kind of gonna circle back around to using AI. For those of us that are using AI plus human review to write our documents, that are interested in using AI to help speed up our work processes, what is kind of the challenge with these tools that AI— that are coming into the market, particularly in the HR space? So if you’re implementing AI tools in your department, you want to make sure you look at transparency and auditability. So, so a black box tool that can’t be audited is less likely to be deemed compliant with all these incoming regulations, right? So that could be AI sorting tools for high-value resume screening or for performance evaluations, but not doing those impact assessments on those tools, not looking at those tools for bias, not reserving the final decision-making or even early decision-making into humans and documenting how that decision-making is integrated into the tool. Those are things that you really want to do.

[00:40:48] Adrienne Drew: In the US, failing to automate your tools for discriminating out discriminatory outcomes can really catch the attention of the EEOC and maybe even bring some fines your way. So we talked a little bit about data privacy risks. We also talked throughout about the human oversight piece, right? So you want to make sure that you’re moving data around in conjunction with the applicable privacy laws, and that you have human oversight in the decisions that you’re making. It’s never that you shouldn’t use AI. In fact, I think we all should be using AI. It’s just that you want to make sure that you and your teams are trained to use it in the right way, and then to know how to look at the output in a way that you can sort of run it through the lens of a compliance program.

[00:41:36] Mallory Hendry: Okay.

[00:41:39] Adrienne Drew: We sort of talked about this a little bit, which is the myth of the global standard. This is something that American employers, I think, try to export the American employment relationship out. And I’m not going to spend a ton of time here because I want to make sure we have time for questions. But this is definitely that challenge, right? We cannot use one employment model around the world. So the bigger your footprint, the bigger your compliance program is going to have to be.

[00:42:08] Adrienne Drew: Okay, terminations. I talked about this a little bit, but I told you I would come back and talk about Canada, right? So that the Wexdale trap— so Wexdale was the case in Canada that established that a termination provision must be evaluated as like a single scheme in the contract, right? So if you write a 4-clause termination clause in a contract that defines misconduct a little too broadly or in a way that is different than the local statutory standards, the court can declare the entire section of that contract to be unlawful. So if you have written your employment contract in Canada like just a little bit too advantageous to the company, then the entire termination provision will be illegal, and then you can’t refer to that to terminate the employee, which means you have to fall back on the Statutory. And that can be kind of tricky. Like I said, statutory tends to give— you have to pay statutory severance, whereas maybe you wouldn’t have had to if you had built in some notice and pay into your employment contract. So instead of owing a terminated worker like a few weeks of statutory minimum pay like you had negotiated for, you’re on the hook for providing common law notice, which can be full salary, benefits, bonuses for, you know, 6, 12, or 18 months depending on the type of worker. So you really want to look at those clauses in your employment contracts. They’re going to have a big impact.

[00:43:36] Mallory Hendry: Okay.

[00:43:41] Adrienne Drew: So how do you choose and vet the AI tools? And I think that a lot of us are in the process of looking at AI tools. I’ve actually looked at 5 or 6 in the last couple of months. And The technology is a little bit mind-blowing. So you kind of think, how do we know which one is going to work for our organization? I’ve had a benefit of working with a lot of product developers, but I’m still a lawyer. And so I’m trying to think, like, how should I look at these tools?

[00:44:10] Mallory Hendry: Right?

[00:44:11] Adrienne Drew: And I look at them and think, where does my team fit in? Right? How can I make sure that this tool supplements my group of lawyers versus making decisions for them. Also, how can I make sure that the information that I’m putting into this AI is not used to train a model, won’t be surfaced in answers that come across to other clients, right? So I want a siloed environment for my AI that I know I can work in in an enterprise fashion so that my information is secure. I want to be able to download my documents back out if I’ve put them in. And I want to see if the AI is purpose-built for my industry or my organization or even for HR. And I also want to make sure that they’re— they are amenable to working with maybe my legal team, maybe my HR team, maybe the behavior of that AI works particularly in my field. I’ve seen some tools where the output looks very much like what a software developer would expect, and it uses words I don’t understand, and it presents data in ways that I can’t use it as well. And so you just kind of want to make sure, is this purpose-built for your field?

[00:45:25] Adrienne Drew: And so there is some— there are some red flags when you’re looking at vendors to do these tools for you, right? So like a one-size-fits-all output is not a good example of a purpose-built tool. You don’t want— if there’s a platform that says, here, you can use this employment contract for any of your hiring, or I can make you an employment handbook for the US without having state-specific output, those are not really the tools that are going to be suitable for the type of complexity required in a global employment situation. We talked about lack of transparency, no audit trail, the ability to run reports about usage, the ability to see sources. Citations are really important if you can see those. Transparency about where the data is coming from and where it’s going. And then anything that eliminates human review. So you want to see a tool that really lets you put yourself into the process. Can I edit this feedback before I accept it? Can I edit this output? Can I change my prompt? Can— is the AI asking me to clarify things? You know, is the AI suggesting that I seek legal counsel if there’s a legal issue? Is, you know, what, what is the, the impact of the AI versus the impact of the human? And you really want the human to be really front and center when you’re making employment decisions.

[00:46:44] Mallory Hendry: Yeah.

[00:46:47] Adrienne Drew: And so really, the, the gold standard would be like evidence-based compliance, even if you’re using AI. So you want to look for comprehensive jurisdictional coverage, make sure that your AI is specific to the jurisdictions that you’re working in, not sort of general purpose-built system. You want to make sure that there’s references to the local laws, that sources are updated, and that the information that you’re getting is relevant and accurate and helping you make a compliant program across jurisdictions.

[00:47:21] Adrienne Drew: As I mentioned at the beginning, GP is an employer of record, so we, we do a lot of cross-border employment. We have our internal teams around the world, but we also hire for our clients in 180 different countries. So we have a really strong compliance-minded company, and one of the things that we have done on the AI front is that we built an HR compliance AI to help us internally. And then we took that product out to our customers and to the general market. So an example of an AI that is purpose-built for HR would be the one that GP built, which is called GP GIA. It’s an AI-powered HR agent that answers global employment questions. It generates documents. It also has compliance monitoring. So it surfaces legislative changes from around the world so that teams can stay on top of these local laws that we’re talking about. It also allows you to collaborate across departments. So you can work from HR into legal and maybe turn a compliance update into an update of your handbook. You can make these things programmatic and you can operationalize these types of changes. So Gia is an example of an AI that we built that is out there that can service an HR need. There are others, but I want to talk a little bit about what human in the loop means because we made sure that our AI has humans in the loop.

[00:48:41] Adrienne Drew: We’ve talked about that a lot today. What does a human-in-the-loop solution look like in real-world AI? Is it just I’m the human and I am prompting the AI, or is it something different? And how do you implement a human-in-the-loop framework with your team without losing the speed and efficiency that AI can provide? We use AI a lot internally. We have an enterprise LLM. We have developer tools that are based on AI. We use Jia internally. We also use AI tools that are in all of our sort of tech stack. I use AI every day, but we treat it as a starting point, a reference point, but it’s never the final decision maker. And I think documenting processes that have AI as a resource but not a decision maker is a great place to start. I never copy and paste the output of my AI. I always revise it. I ask for the underlying sources and then I actually look at them. And if they’re legal sources, I’ll confirm them. If they’re not legal sources, I’ll read them and, you know, gut-check them to make sure that I agree that it aligns with my expectations. And that’s not even decision-making, that’s just general output.

[00:49:52] Adrienne Drew: So human in the loop simply means that your technology handles the heavy lifting. It can— AI is great at going through a handful of documents and finding common information. It’s great at summarizing. It’s great at providing options. You could say, what are the, you know, what are the risks and rewards of doing A, B, or C? Or what are my varying options that I should consider? But not which one should I do, right? Humans should retain accountability for the ultimate outcome.

[00:50:20] Adrienne Drew: Okay, so one example of a compliance workflow that integrates AI could be who reviews what and what stages of the lifecycle. So this is just an example of maybe something you could convert into an SOP. where you’ve got a recruiter, you’ve got— and then you have your TA team that’s working on bringing someone on board, right? So you maybe have AI at certain points in this process. Your AI platform could draft a job description. You could use AI to draft a job posting and then confirm that it complies with the job posting requirements of a local jurisdiction around things like equal opportunity and including salary information if that jurisdiction requires it, right? But your recruiters need to review it. They need to make sure it agrees, it is compliant with all of those laws, right? So then you’ve used AI to help you, but it’s not making the decisions. You have eyes on it before it goes out. And then stage 2 is shortlisting. So you’ve got 100 candidates for your role, and you might have a system that ranks candidates based on some criteria that you have provided. But you shouldn’t just accept that ranking. That should be something that humans definitely review as well. Review the criteria every now and again that you’ve given the system, and then review the top candidates and maybe even some of the other candidates to do a spot audit and make sure that there are good candidates everywhere.

[00:51:44] Adrienne Drew: And then where the highest stakes are, when you’re making an offer letter, you’re sending out your employment contract, this is where you want to be fast because those great candidates can get snapped up really quickly, particularly in jurisdictions like India. You know, I know that we’ve seen people offering Really high sign-on bonuses to get someone to keep them from taking another job. And if you have an employment contract template or if you need one, and this is an area where I think AI is probably not something you should use as much, you should have your team really customizing, I think, those contracts. Using an AI-generated contract at this point, one, your target, your candidate might notice. Everyone’s getting a little bit better at being able to identify, I think, when things are generated by AI. But also you could miss some of those nuances that we’ve kind of talked about. And then stage 4 is your ongoing monitoring and your employee lifecycle. So your HR leadership and your compliance teams might need to do audits of your documentation and of your decisions, especially when it comes to end-of-year decisions around promotions and things like that.

[00:52:50] Adrienne Drew: AI is really great for doing bulk reviews. So an example, maybe 6 years ago, we had to do an assessment of non-compete clauses across jurisdictions. And so Painstakingly opened up every employment contract and had a spreadsheet and said, this person has this type of non-compete and this person has this type of non-compete. I would have loved for AI to have been able to take that off my desk and give me a report. So that type of audit work is great. Just like brass tacks, no decisions needed. How many in column A? How many in column B? AI all day long. I think we should use it for that. But you also want to take that output and then use Human judgment to decide what to do with it. Never let AI say, these are the people that you should change their employment contract. You have these people that should get this and these people that should get this. You really don’t, don’t want to be doing that.

[00:53:39] Adrienne Drew: So, all right, so how could you execute this type of strategy in your own company? Maybe you found your tool, maybe you’ve, you’ve onboarded it, maybe you’re using a standard off-the-shelf LLM, but you have an enterprise instance of it. So it’s— you’ve dealt with the privacy and security concerns. So you want— you can use an AI draft, you get your expert review, and the AI and expert combined is really the gold standard. And I remember that most of you actually had that format in your company. So I think that that’s a great starting point for using AI in a compliant HR program, is to say, all right, I’m gonna use AI here, but I know when to bring the human in the loop. And that way you don’t get sort of like the false sense of security by having a bunch of words on a page because they look good. But maybe you’re missing something around the edges, right? That’s always what you want to avoid.

[00:54:31] Adrienne Drew: Okay. Also, the, the show your work collaboration model. This is really important, right? Because you want to have legal building the baseline, and you can use AI to help you with that, but you have a legal baseline. You don’t want to stall. Sorry, you don’t want to stall out because you’re waiting for feedback. And then you want to immediately reach out to legal if you have any other questions, right? So get AI to get you your speed. Use legal, use your expertise and your other tools to get you your compliance. So it’s a collaboration model. Here, I’ve used this tool. Can you help me confirm that it’s right? We want to move forward. It could be, I want to calculate a commission structure in Korea. Can you help me? I think that it looks like this. Or I have a termination in Canada. What do we do? I think it looks like this. You don’t want to take a problem like that to legal. You want to sort of say, I have gone to AI and I’ve got this proposal. I love taking a proposal to legal. And then you say, this is where I think we should go. Can you confirm that this is correct? Let’s talk about this, and then you can move on. So that’s really an example of using AI. I think, to get you further down the field, to speed you up, and to make sure that you are compliant and fast and meeting your company’s goals.

[00:55:51] Adrienne Drew: Alright, so a couple of key takeaways, and then I know there’s some, some couple questions. So AI is a starting point, not the final, final answer. I’ve harped on that a lot. The illusion of completion, the comfort of maybe having an answer, that’s not necessarily something I wouldn’t think any of us would do that, but we don’t want to do that. Global standard is not really the way to go when you’re across jurisdictional lines. And then a show-your-work collaboration model of saying, here, this is what I did. This is where I used AI. This is what I have. Can you confirm it? The human-in-the-loop model is really, really what you want to use when you’re using AI to expand. Okay, couple minutes for questions.

[00:56:30] Mallory Hendry: Okay, yes, we have a few here, Adrienne, for you. Someone asked about hiring in Africa. Do you have any perspective to share on that?

[00:56:40] Adrienne Drew: I’m trying to remember the— so, yes, we do a lot of hiring in Africa. I think that we’re seeing more and more hiring into Africa, and actually African countries are really great to collaborate with right now because they are looking at generating a lot of outside investment. They’re looking at bringing companies in. They’re looking at exporting their workers in a remote work environment. We see a lot into Kenya. We see a lot of hiring into South Africa. And I think that the same thing holds. The business models are not as clearly defined. So in Europe, there’s— it’s very employee protective. I think African countries are still trying to collaborate with the needs of the market. And they are flexible on the one hand, but kind of hard to work with on the other because some of their systems I think are not what we would consider easy to manage. So when you’re setting up an entity in an African country, it can be very expensive to capitalize. For example, I think when we were trying to capitalize an entity, it was like half a million dollars just to open a bank account. So certain things in expanding into African countries can be really hard, but the labor market is fantastic. And I think it’s a great choice. To do again, using that careful compliance structure. Watch out for gig work. Make sure that you’re aware that these countries are more protective than the US, perhaps less protective than some of the European countries, but assume it’s about the same, the same level.

[00:58:16] Mallory Hendry: It’s great, Adrienne. Thank you. Someone else has asked, so they say you mentioned the general purpose AI like ChatGPT shouldn’t be used for contracts. But their HR team uses it to draft quick policies or localized emails. So where is the line on that?

[00:58:34] Adrienne Drew: I think it’s a real judgment call, right? If, if it’s a— and it also depends on if you have an enterprise version or you’re using just an off-the-shelf on your phone. If, if it’s just my own personal account on ChatGPT, I don’t use it for any of the above, right? I might ask it for a good recipe for what I can make for dinner that night, but I won’t ask anything work-wise. If you have an enterprise version of ChatGPT at your organization, then you can maybe expand a little bit. I think using it to help draft emails is something that I know our team does in our enterprise LLM. Again, that’s specific to our organization. It’s locked down. Our administrators don’t look at our chats, and the LLM company can’t look at our chats either. And then also, you know, documentation isn’t shared or used to train the model. That to me is the critical point. To draft a quick policy, I wouldn’t use it unless I had significant inputs. If I wanted to say, here’s our old policy or here’s our parameters, we know that the following policy bullet points are accurate. Can you use my facts to write the policy based on my facts? That’s an area where I think that depending on the risk profile of your organization, you might want to take that approach saying, here, I’m going to give you the data you need. Can you give me an output based on this data? I, in my organization, I would not advocate for saying, you know, just add the LLM suggest to you what it thinks you should say.

[00:59:59] Mallory Hendry: Yeah, that makes sense.

[01:00:01] Mallory Hendry: Someone else is asking about the Canadian Wakesdale ruling. They’re calling it concerning for US employers. So if our termination clause is invalidated due to an overly broad for-cause section, what is the process for fixing existing employee contracts in Canada?

[01:00:17] Adrienne Drew: Oh boy, yeah. So this is— I sort of alluded to this a little bit, which is that anytime you change an employment contract in Canada, you have to get the employee’s explicit written consent to the change. And if it’s something that changes the terms and conditions of their employment, and that could be like for the— even for the benefit, like they get a raise or anything, you have to give fresh consideration, meaning you have to get their approval and you have to pay them for it. So if you want to say, I’m going to change your termination clause, then you’re going to say we’re going to give you a bonus and you have to re-sign your employment contract. Without a re-signed employment contract and consideration, you risk that new contract not being valid. So it’s a labor-intensive process and it’s going to be a little bit money-intensive depending on what the value of that change is. And there’s some guidance in Canada that talks about how much should you— how much does each change sort of What is sufficient consideration for that change? So you can seek guidance on that point as well, but that’s an area where you’re going to want to talk to your lawyers, to be honest.

[01:01:19] Mallory Hendry: Okay, that’s great. Okay, we have one more question here. I know we’re running short on time, but how do we evaluate if our global ATS’s new AI screening tools risk violating the EU AI Act or Colorado law? What specific questions should we be asking the vendor?

[01:01:37] Adrienne Drew: Yeah, I think it’s fair to ask the vendor, you know, what’s their take on those laws, and can they, can they give you a one-pager on how they think about those laws and how their, their tools work within them? You know, in Colorado, you can ask, is there a bias audit? Are there appeal processes? And are there notifications? If there aren’t built in, can we, can we do some configurations so we can get those tools? A configurable tool will help you probably be able to configure certain settings so that you can be compliant, but you want your vendor to make sure that they’ve thought about that and they can give you a point of view on that.

[01:02:12] Mallory Hendry: Okay, that’s excellent. Well, thank you so much, Adrienne, for sharing your expertise with us today. And everyone in the audience, thank you so much for joining us. Keep an eye out for more insightful Human Resource Director webinars and enjoy the rest of your day, everybody. Thank you.

[01:02:27] Adrienne Drew: Thanks. Bye.