116 The Impact of AI On Offshore & Entry Level Accountants

We're getting some really high quality studies for the first time benchmarking knowledge workers outputs when using AI versus not using AI, just in general on different types of tasks, but also on the technical skill spectrum. And I think the results are surprising and may challenge some of the assumptions we have about how work is going to change in the near future. So come on in, let's talk about it. Specifically, how does this stuff impact offshore workers and how does this impact new grads, those fresh baby cheeked new grads that we need to come in and help us out and ultimately buy us out of our firms. Right, that's really what we're concerned about here. So come on in, let's talk about it. I'm just kidding. I was kidding about, about new grads just buying us out. That's not why we want you here. We want you here because this is such a, a fun and challenging profession. That's why we want you here. Okay. I think the general, I see people saying oftentimes around AI, the same standard trope that we saw, we've always seen with automation, that oh, AI is going to be great to automate the menial tasks and get us, get us on to performing higher value work. And I don't think that's necessarily true. It will automate menial tasks, but it's actually going to automate tasks on the entire skill spectrum. This is really for the first time, it's interesting. It's for the first time that like white collar stuff is getting automated and the decision makers, the important people, their jobs are actually at risk too. It's not something that is being put upon like the lower skilled workers. It's impacting everybody. And I still don't know that people have their heads around this because I keep hearing things like, well, there goes offshore and what are all the interns going to do? And there's kind of two like inherent fallacies in that. One is that AI only impacts menial work, which is absolutely, absolutely incorrect. If anything, it's more disruptive to like mid level, mid, upper level folks for reasons that we're going to get into. But the second misconception here is just that there's a fixed amount of work to be done. And this is like the biggest fallacy in all the AI, ending anyone's jobs discussion is the idea that there's this fixed volume of work to be done. And in public accounting that could not be further from the truth. Riddle me this. If you run a firm and you have a team and you found a way to automate one of your team members jobs or half of their jobs, would you go out and fire that person tomorrow? No, you absolutely wouldn't. Like I like that was would have never been been the case in all of my firm running years. We were so hungry for more good people and constantly trying to upskill people and do everything we can to automate for them so that they can keep doing higher value stuff. And I don't know what better example we need than Covid to show us that there's not just a fixed amount of work in this profession. You get legislative changes, you get, I mean there's a hundred different examples of when the workload increases and contracts and for whatever reason we still go into work, we still work all day and then we come home. And that's probably going to happen until the end of time until you like give yourself permission to work less. But that's another podcast episode. So there's this notion that there's a fixed amount of work to be done in that that totally isn't the case. And so in the last couple of months we've actually got some interesting studies around that like try to as best as they can fairly benchmark giving a set of workers AI to work with and having a set of workers complete the same tasks without AI. And they've gone a long ways to try to like randomize this and ensure that it's not going to be biased towards people that maybe already had like experience using, you know, chat GPT or something like that, because there's definitely an element of learning the tool. In fact, one of these studies, one of the things they benchmarked was even like giving them a primer on this tool. And so they pulled in folks that had never used it before and it was just an AI chatbot. And the folks who they gave like a really high level primer vastly outperformed the folks who used it who didn't have that primer. A great reminder of the value of just knowing a little bit of prompt engineering and that sort of thing, AKA send your team my chatgpt videos. But interestingly, what at least the two main studies that I saw both found, I'll reference this one from Harvard Business School. It's actually an interesting read. Navigating the jagged technological frontier field Experimental evidence of the effects of AI on Knowledge, Worker productivity and quality. That is a mouthful. I'll put a link to this in the show. Notes if you're interested in general participants with access to the large language model GPT4 completed 12.2% more tasks on average and completed tasks 25% more quickly than those without access to AI. And the context here is it was a group of 385 participating consultants were given a set of 18 consulting tasks. And so this, what does this tell us? This tells us maybe how helpful it is on consulting tasks. Right. Which is not necessarily always the stuff that we do. I think a lot of this is actually like narrative type of things, which would completely make sense that it would be super helpful. There. It says the responses produced by the consultants with access to generative AI were more than 40% higher quality compared to a control group. I don't know how you measure that, but that's a big delta. Another interesting tidbit here. They ran a second experiment where they required study participants to analyze a company's brand performance using insights and financial data. The responses were evaluated based on correctness and there was a notable dip in performance among those with access to the AI compared to the control group. Subjects in the control group were correct about 85% of the time, while those in the two AI access categories saw an average decrease of 19 percentage points by comparison. So the folks that used AI for financial analysis scored 20% worse. And actually part of the idea of this study, they call it the. What is it? The jagged edge of innovation, the jagged technological frontier. It's just like the difficulty of knowing what it will do well and what it won't do well and how to ultimately trust it when you don't exactly know what it's going to nail and what it's not going to nail. Because when it's magic, it's amazing until it does something really dumb. And you don't always know when that's going to come. But the most interesting thing in this, the that the last two studies I've seen both said, was that those participants who were categorized as bottom half skill performers saw a larger boost in performance from having access to AI 43% when completing experimental tasks compared to the bump experienced by top half skill performers, just 17%. So the folks in the lower 50th percentile in terms of their performance levels coming into the study, increased performance with AI by 43%. Folks who are in the top half only increased by 17%. And this makes sense in general, if you try to use this stuff for your area of domain expertise, it's not always that impressive. Right? But if I use this for something I don't know anything about, like cooking or coming up with a cocktail recipe or applying a tourniquet in battle, like these aren't things that I'm an expert in. So it's gonna be really helpful. There's. And I suppose by extension then it makes sense that the people who maybe within your field are not as accomplished as you have more to gain by using that AI than you do. And so to bring this back around to grads and offshore staff, what this tells me is that this could be the golden age of noobs, the golden age of, of entry level skill people, of upskilling folks at a level that like was never before possible. If we can get our underperformers to improve to a level of parity with the folks who were the over performers, it's like this now, this standardization of capability. If everybody has AI now, there's a much smaller delta between those new folks and the folks who are high performers previously. And so to me, if anyone is disadvantaged by this, it's the high performers, I think, right? Because it's easier than ever before for the underperformers to reach parity with them. This episode is sponsored in part by the fine folks at Cloud Accountant Staffing. Do you hire accountants? Bless your little heart. Not the best part of the job, in my opinion. Not something I ever enjoyed. Well, listen, you can build your accounting dream team with talented offshore accountants in the Philippines that work 100% full time for your firm. Their accountants aren't freelancing or contracting for multiple firms. They're all yours. They work exclusively for you and are incentivized to stay with you and your team long term. And I can get swiped? Cloud account staffing is 100% dedicated to the accounting industry and founded by a former accounting firm owner that understands your business, knows your pain points. They had to hire some accountants and they said, you know what, we're gonna build our own pipeline in the Philippines, gonna pull in some super talented people and then open that up to other firms. Basically, that's the story I've been talking about a lot about staffing, building more resilient staffing pipelines for your firms. I had staff in the Philippines like totally red pilled me to like, oh geez, like we need to globalize the way that we get our work done. Check these folks out link in the show description cloudaccountantstaffing.com this episode. Oh, this episode is sponsored in part by countests. You know what? I don't enjoy spinning the old wheel of does this person I just hired know what the hey they're doing? Because I can tell you that more times than not. The answer was no. Because how in the world do you actually figure out if this person can be productive based on, I don't know, a resume based on the firmness of their handshake? You can't but account tests, it's gonna give you a little more information than you had otherwise. So Count tests is a super duper simple way of doing skills testing for the people that you're going to hire. Basically generate a link, you, you can run them through these skills tests. It's nothing super sweaty, but you know what it is is information. Information that you wouldn't otherwise have unless you sent them this little quiz. They got tests for accounting stuff, for Gap stuff, for tax stuff. Whatever you're hiring for, they probably got a test for you. Super cost effective. Because let me tell you, it is a whole heck of a lot cheaper than hiring the wrong person and then training them for months and then realizing I have made a huge mistake and we just wasted months of our time. We with the wrong person. Boy, do I have a whole closet full of those T shirts. So that's something that keeps you up at night. Check out account test. We'll put a link in the show notes. Now specifically looking at the, the offshore discussion because I've seen a lot of people talking about this, that this ultimately will, will hurt offshore workers. The adoption of AI. I think there's a number of things like loaded into that line of reasoning. The main one being the super broken mental model that says offshore workers are just going to do low entry level menial work. I'll let you in on a secret. Offshore workers are capable of absolutely everything from a technical standpoint that onshore workers are. And it's like it sounds dumb to say aloud but it's like these people are humans. Like they can learn every bit as, as quickly as the people that we have onshore. The main thing that most companies aren't going to do is they're not going to put them in a client facing capacity. At least in professional services you won't do that. But because the first work that went offshore culturally was like low level, entry level, menial stuff that nobody else wanted to do, there's this really unfortunate projection onto offshore workers that that's all they're capable of. When I know firms whose most senior technical people, most senior tax technical people are offshore, oftentimes it means building that investment yourself. Like investing in people over decades and getting them to where you want them to go. There are talented people. I mean all the big firms are offshore. So Those people, like, you could hire a big four person in the US that's super talented, you can do the same offshore, but they're going to be really hard to get, just like they are here in the US but. But if you believe that you can develop somebody onshore to a level of expertise, promise you you can develop an offshore person to that same level of expertise. So what I get really excited about when I read this is the notion that I could pull in offshore folks who are not like the mega high level, technical big four types, but I can pull people in and actually upskill them faster than ever before by pairing them with AI, which for me makes them more valuable than ever to bring this back to college grads. Same thing like that makes those people more valuable to me than ever. The notion that I could pair a college grad with an AI and they're capable of more productive output than they could be otherwise. I mean, we all know when we were juniors, just the astronomical time you spent floundering and trying to figure something out because you were afraid to ask another question. And you just burned like entire afternoons in the summer trying to figure out something really dumb because you just don't want to be the numpty that has to ask. Imagine having a little friend you could ask who wouldn't judge you, who, like, there was no fear of asking stupid questions. My gosh, like, chat assistants are going to be so, so valuable for juniors. So if we all of a sudden have like junior technical people that are capable of way, way more, I have a really hard time imagining there's not going to be something for these people to do. Right. If anything, if an entry level person will get me 80% of the capability of a person with a decade's experience, like, I would think that person ought to be concerned. Now, ultimately, in our space, we are so, so far from anybody needing to be concerned because there are so many things to do and so many more people that we can help and more things that we can do for all the people that we already have that, I mean, I genuinely believe. Nobody is in a position where they're like, oh, we found a way to automate what you do. See you later. It's just not realistic, I don't think. And so I heard this, this came up last week at an, at an executive roundtable. And there was kind of just this implicit assumption that it was almost kind of a bygone conclusion that, oh, we won't, we won't really need our offshore teams anymore. And I just, I think I think what that's actually founded in is the past realities around automation. The stuff that you would automate is like data extraction, data entry, stuff like that. Are we, I think we have some assumptions that maybe we're not aware of that are founded in past types of technology where the stuff that was most ripe for automation were the tasks that the entry level people were doing. But that's absolutely not the case with AI. Everything on the skill spectrum is subject to AI being able to help you get that stuff done more efficiently, do it completely for you. Right now, I genuinely don't think it is concentrated at one level of expertise spectrum or the other. If you think about stuff like the most technical research, oh my gosh, the power of AI to go through mountains of documentation, super, super nuanced documentation, and make connections instantly in a way that would take you so, so, so long to do, doesn't mean it's a replacement for your verification and for you going in and doing a lot of that digging too. But some of those hairiest things that require the connection of a bunch of stuff that a junior person may not be aware of, AI is going to be tremendously helpful for that stuff. So another couple interesting technical developments that I think lean into this even further a little bit around kind of the architecture of how AI is continuing to get better and what the solutions that ultimately like can prepare a tax return or do these really complex tasks will look like. It was maybe a month or two ago, there was a project put together that was really hot for a while called Meta GPT. And what was novel about Meta GPT is Meta GPT, the multi agent framework was it basically created these avatars simultaneously that would work with each other and bounce things off of each other to get to a higher quality final product. And this was specifically created around the notion of software development. And so if you have a software agency that's not like a single person with all of the expertise doing all of this stuff, right, you have different roles within that agency, people with different types of expertise that add their own, make their own contributions to a project. And as a result, kind of the rubbing together of those different skill sets creates a higher quality output. And that was kind of the premise of Meta GPT. And so you would task it with creating an application and within that you had a bunch of different agents. You had an engineer agent and its jobs were to write code, review code, debug code. You had a QA agent, its job was to write tests and run tests. These are like tests for the system to see if it breaks, that sort of thing. Quality assurance. Third, you had a project manager. It would write tasks, assign tasks to different agents, review. I don't know what that is. Review design and review code. You have an architect, it would write the design, review the design and review the code. You had a product manager and then you had a boss that was managing the overall requirements of the project. And all of those different agents were worked together on that same task. And they would critique each other and iterate on exactly what they were doing through all of those different perspectives. And this was really novel and yielded a much, much higher, higher quality output than if you simply told a language model how to do a thing and then asked it to do it. It turned out that all of those different perspectives and how they interacted with each other yielded a higher quality output. This episode is sponsored in part by the fine folks at Tech Guru. Because you got better stuff to do than worry about your computer problems. Tech Guru is an IT firm that just works with accountants, accounting firms so they understand you understand the annual cycle, the oftentimes awful software you're forced to use. Not always. And they do it via their three S's approach. That's right. There's three of them. I'm now going to give them to you one at a time. One, strategy. Industry focused tech strategery. Sessions with accounting technology experts. Like people that do this stuff for a whole bunch of accounting firms. Number two, security. Ensure nobody's gonna steal your lucky charms. My copy, not theirs. And three, support so that you got somebody by your side when bleep hits the proverbial fan. Am I right? Spend less time stressing about computer stuff. More time stressing about client stuff. That's what you should be doing. Learn more about Tech Guru at the link in the show notes. This episode is sponsored in part by copilot. Okay, everybody, gather round. Take, take one another's hands. Now we can all agree the way that we exchange information with clients. Very important. Can be a massive time waster if you don't nail it. But I get it. The fear of training clients on a portal system, right? What if you end up not liking it? What if you train a whole pile of people on this giant system and then you're like, turns out this actually isn't the one for me. Or this new super cool AI rocket ship platform just took off and now I gotta get all my clients to move. Here's the thing that is the best argument for why your client portal experience ought to be separate from your Workflow management tool. You getting what I'm saying? Copilot man, all they want to touch is that client experience. Just the portal giving you a mega flexible platform for how you want to work with your your clients. So that if the workflow stuff changes, if you want to pull in a different tool for that, you can without changing the client experience. Pretty smart. Especially in these scary changing times of AI. Right? Right. Actually got a, a demo day coming up on the main YouTube channel from Copilot in the next week or two where we actually get hands on with it. You can see even, even more about it. If you're looking for a cool modern client portal experience, check out Copilot link in the show notes. And this is kind of building on, you know, something we've talked about in the past, chain of thought prompting where when you're. If you're talking to chat GPT and you want it to solve a complex problem, you'll get a much higher quality output if you ask it to solve it step by step because it kind of as it outlines that stuff step by step, it is changing the probabilities related to like what is subsequently generated. And in that study where GPT4 passed the CPA exam, the this step by step approach actually increased the score by 20 points. So it makes a big difference in the quality of the output. Meta GPT, the notion of having all these roles and they interact with each other is kind of a further development of that concept. And this got picked up again recently because Microsoft just launched their own version of this framework that's a little more flexible. It's not specific to developing software. It's really interesting. It's called autogen. I'll put links to all this stuff in the show notes. Auto Gen enabling next generation large language model applications and it lets you in a more modular way build basically what Meta GPT was explaining. And you can even design kind of different ways for those agents to interact. Like you can have a hierarchical version where you have kind of a boss and everyone reports back up to that boss. You can have like a more circular sort of thing and then what we do, what this ultimately looks like. I was talking with Ashley Francis about this the other day. When it comes to tax and accounting, you're probably going to have these agent roles that really represent like human roles on the entire expertise spectrum. So take a tax return for example. Like an AI that ultimately will prepare a tax return behind the scenes. There's probably all of these different agents attacking that problem from different angles. So you've probably got something like an admin that is like ensuring that the project is moving through the steps and the documentation is like in line with the company standards, stuff like that. Second, you probably have like a detail reviewer, the person that's making sure that you know this number got keyed in correctly over here. Just making sure there wasn't transpositions, that sort of thing. Third, you probably have a technical reviewer, the person who is reviewing the output more through the lens of the technical rules. Fourth, you probably have the client who is considering does this outcome align with past communications I've had and the expectations we had for, for this outcome? You probably got this whole cast of characters that is approaching the problem like from their own different perspective. And you'll notice it's not all entry level stuff like the agents are speaking into this project at every single level of expertise. They're whittling away at that problem in the context of a month end close. In an accounting firm it probably looks similar. You've probably got an administrative sort of thing where it is managing the movement of the entire project itself. You may have roles that look like a traditional finance department. So maybe you've got like a bookkeeper, an accountant, a controller, a cfo, maybe even a CEO that's looking more at the kind of over the top results. But some of the bigger, more complex AI applications that will come together to perform, perform this really complex novel task. Right now it looks like the way that stuff's going to get done is by collections of agents acting through the lenses of different roles. And the fact that this research has floated up to the top and you have people like Microsoft investing in these frameworks to help people develop things this way. To me this is really compelling proof that what LLMs are ultimately great at is not the menial entry level work, is all of the work is the entire vertical. It's simply a matter of the knowledge that you put into that language model that is what it's going to do really well. The perspective that you give it, the things that you want it to prioritize, ultimately that's what it's going to do. And that can be C suite level stuff or it could be entry level stuff. And so as we were kind of having these conversations last week, I kind of initially took it for granted like yeah, we're talking about automation. So it's going to be the entry level stuff and it's going to be, you know, the entry level people ultimately that are at risk. But maybe that's not a problem because we can't find new people. Right. But the more I thought about it as I was looking more into Autogen recently and thinking about Meta GPT as I'm reading some of these new studies that are coming out, all of these things are kind of corroborating that notion that it's not the entry level stuff that's at risk. Everything is at risk, really. Probably. And maybe at risk isn't the right way to put it, but everything can be improved. We can improve upon the way that we're completing virtually all of these tasks, which is great news. Like CEO wants to get home for dinner just as much as the entry level dude. Right. And so I like, I don't wade into the whole pipeline discussion and CPA exam and all that stuff, but this is one area where like I'm. All of the evidence that we have right now I think points to the fact that this is great news for people coming into the profession. It's going to be an accelerant for them to make meaningful contributions. You remember what your first couple years in the profession were? Out in a stinky warehouse somewhere counting potatoes. Like, man, nobody's going to miss that stuff. So the notion that new people can come in and be much more productive than ever before, that's great news. That is a big old win for all of us in my book. And I would also extend that as far as offshore staffers staffing offshore workers. Where maybe that goes sideways is in a situation where the firm just doesn't have the right mindset and isn't treating those people as like peer members of the team, but rather this like lower level team that's only allowed to do menial things. I think that's a massive waste. I mean that's just like, like, I guess I don't get why people, I guess, see offshore teams through such a different lens. It'd be like going out and hiring a bunch of people from a different college and saying like, oh, they're not capable of this other work. Like they're capable of all the same things strategically. When you plug these people into your business, like yes, you're probably not going to put offshore team members in client facing roles, but we can't even manage to get the work done. Like it is all hands on deck in this profession right now. Could you use more people to help you do all the technical stuff? Absolutely. Now one thing that is the same, I think one thing that does carry over from the old automation tropes is just the continued march towards there being a shrinking amount of space for folks who can't be human, for people who can't build relationships with other human beings. We've been saying that about automation for a long time. There's less and less room for that person who's gonna sit back in the corner and just crunch numbers all day and not manage relationships or manage people. AI is gonna keep pushing us in that direction, I think. It sure seems that way right now. If you go out and you hire 10 entry level people and they're capable of really high level, powerful stuff, that's awesome. But then what does their jobs become? It skews then more heavily towards relationship management rather than doing the work. Right now, if your job is 5050 relationship management and doing the work, I see AI knocking off more of the doing of the work tasks than the relationship management tasks, which just take time. Sure, there's stuff like writing emails and all that that it can help with, but ultimately AI automating a bunch of work is probably impacting the technical stuff more than the human stuff. So I would say that remains unchanged. We're continuing to go down the path of it being more important to be able to be human, to build relationships, to be, you know, have high emotional intelligence, to be able to make people feel like you care about them, to be able to like listen and be compassionate and have empathy and all that stuff that is becoming more important by the day. And I think especially as we have more digital experiences and more things get productized and we talk to chatbots more instead of actual people, we're going to get more and more of a premium on human in person experiences and service. That is human, not a chatbot. So you still got to figure out how to be human, not just for work, but because that's also just a good thing to do, you know, okay, if you are offshore working for like a US company, if you are a new grad, if you're considering coming to another profession, in my mind this is all good news. If you have people who are afraid of this, or like fellow students who have been spooked by this, please share this. Everything here seems to point to AI actually being killer for you. And this is a win for everybody. Like the boss, they want you to be as, as productive as possible too, right? So this all seems like really good news to me. What do you think? Anything you've seen that would, would make you think the opposite, or any compelling evidence that we've got to the counter to this, would love to hear it. We're still kind of figuring this stuff out day by day, right? So thanks for coming and hanging today, and I'll see you in the next one, Sa.