Hello, everyone, and welcome to the Recruit Holdings Fireside Chat at the Goldman Sachs Communacopia + Technology Conference. It is great pleasure to have Hisayuki Idekoba, CEO of Recruit Holdings here today. My name is Minami Munakata. I cover Recruit Holdings and Internet, Entertainment here at Goldman Sachs. We have about 35 minutes for today's session, inclusive of audience Q&A at the end. With that, Idekoba-san, thank you very much for being here. Thank you for having me. Thank you. Let us get started. Idekoba-san, Indeed began as a job advertisement business, but pay more, get more, suggest what employers pay for today is much closer to an actual hiring outcome. Could you walk us through that evolution and how you personally define what Indeed is exactly selling? What Indeed is exact selling. We are selling basically hires. We are selling hires. That is the very important part. Basically, AI technology helped us to make sure that we are delivering qualified candidate. That is the most important part. I am going to give you a good example. In the past, Indeed was more like a search engine, meaning we were charging $1 for this click, $1 for this click, $1 for this click. It does not matter who is clicking it, right? Meaning, if this click was done by the person who just graduated high school or just graduated university, whatever, or three years experience, 20 years experience, we were charging $1, $1, $1. As a result, sometimes employers said, "Oh, we allocated a bunch of budget. We got 200 applications, but quality was not great. Now with AI, we have a great AI technology to detect, okay, how qualified this candidate is, who is clicking this? Who is clicking that? Who is clicking this? We are charging, let's say, the same JPY 3, we are charging JPY 2.7 for 20 years experienced person and probably JPY 0.1, JPY 0.1 from these two clicks. As a result, when employers are allocating more budget, they can get qualified candidates sooner. What is happening is there are so many positions employers are striving to fill. When they increase the budget, they can get qualified candidate, not just the quantity of candidate. That is the big evolution we have been doing for the last 18 months. Mm. Good. Is that answering your question? Yeah, you were selling better experience. I think you started to disclose U.S. ARPJ KPI. U.S. ARPJ grew 35% while Indeed job postings declined 4%. I get a constant question on whether that is sustainable at this level. Should we expect the U.S. ARPJ to slow down from Q2 onward? I am so focused on the long-term innovation. What I believe is right now, it does not matter inside recruiters, which employers have 50 people, 100 people in their talent acquisition team, or outsourced, which is staffing agencies or recruiting agencies, employers paying a lot to them. What they are doing is human being is checking thousands of resumes if they are qualified or not, and also, even if they are qualified, most of the time what they are doing is they are sometimes reaching out to them like, "Oh, do you really have driver's license?" That is the manual, repetitive tasks human being recruiters are doing every day. And 5 years later, 10 years later, I really, really want to automate these manual, repetitive tasks. And that is our goal. If we can do it. What we are doing is literally almost like AI company, right? We have lot of two-sided marketplace data, and we are training our model to detect which candidate is a great qualified candidate for this job, this specific employer. If we can do it, I think it does not matter. We should be able to monetize our value because that is the work, which is there is a cost associated with it, right? JPY 350 billion market, 65%-70% of this JPY 350 billion HR market is the manual cost, human labor cost. I think people are worried about AI is changing people's jobs or whatever, right? But I think we should be automate these things to boost actual productivity of all employers. That is our goal. Short term, what I have been focusing on is more like a. I have been very careful about employer satisfaction, client satisfaction, and the speed of the change. Because this AI change is so fast. The speed of the change is so quick. So we are delivering value. I have been very careful getting good feedback from employers, but I have been very careful. Is this too fast or not? Too quick or not? So that is the balance. I am very careful, but that is more my short-term focus. Is it. Great. So let me ask something about your comment about user satisfaction. Take rate is still under 1%, implies really headroom, but you also said you will slow monetization if the value is not there. How do you judge whether Indeed is expensive or cheap? So how much of the value created should Indeed capture? Of course, compared to other recruiting agencies, which are usually charging 15%-20% of first year salary, which is very expensive. So most of the time, employers, what they are doing is, this position has been open for 90 days, 120 days. Most of the time they reach out to recruiting agencies, staffing agencies, "Can you help to fill this position?" What is happening right now is, I think most of the time, what I am seeing the situation is more like a time to fill. Meaning, cost per hire is, of course, one measurement. Most of the time large enterprise customers, like monthly budget clients are thinking, "What is the cost per hire?" But think about it, the cheapest cost per hire, meaning if you can wait 180 days, that is the cheapest cost per hire. You can of course wait, the position to be open and no budget allocated, and if you can wait 180 days, you might get some candidates. You might be able to fill this position. That's the cheaper cost per hire. What's happening right now is if you can deliver qualified candidates in a day or in three days, they're happier. That's why they prefer to pay 50% more, 100% more to get the qualified candidate faster. That's why we're asking them, "Hey, you should try it." It's not only Premium Sponsored Jobs. When we think about AI innovation, which is happening everywhere, for example, AI sourcing, recruiters, human being recruiters are checking resumes online, LinkedIn, Indeed, whatever, reaching out to job seekers. 90% of time or 95% of time, what they're hearing back from job seekers are, "Oh, thank you for reaching out to me, but I'm not actually looking for a job right now." Our AI sourcing knows which job seekers are active checking jobs within 30 days, which job seekers are opening emails, which job seekers are checking the notifications. That's why it's so productive. As a result, employers can feel that they can get the qualified candidate faster. I think most of the time, "Oh, okay, I'm going to pay this much, but I can get qualified candidate much, much faster." That's why so far, last week we had a client event, we called Indeed FutureWorks. I literally had more than 100 conversations with the employers, but I got so many good feedback so far. I'm a little bit worried we're charging too much, but actually what's happening is we're delivering good value and they're happier. That's what we're seeing right now. You got a positive feedback from both small and mid-size enterprise and also large enterprise as well? That is a good question. Big, large enterprise have very complex tech stack most of the time. Also, it is depending on what level of people we are talking to, like head of HR or CFO, COO or CEO, because this time this is more like AI change. So that is why we are having more meetings with CFO, COO, CEO, because most of the time CHRO is thinking more like HR way, but I know CFO or CEO is thinking, okay, how to use AI to improve productivity dramatically. So sometimes, this top down is happening. So we are trying to have more and more meetings directly with CEOs. You also mentioned for large enterprise, it takes some time to Yes make some final decision, right? They need to test something first, right? Yeah. Actually, at the very beginning of this year, we couldn't have a good growth from the enterprise clients. But finally, we're getting good growth from enterprise clients. It took time, but I think, of course, these are AI products. Most of the time we need to go through their legal and compliance processes. It took time, but we're getting it. That's great. Let me switch gear a little bit. Indeed Hiring Lab's view is that demographics, not AI, will be the bigger force on the size of the labor market, leading to structural shortages through 2040. That's likely not the consensus in this room. It is the consensus in this room. What informs it, and what does that world do to your monetization model? A bunch of people ask me, "Hey, do you think AI will have a 20% unemployment rate? AI is replacing human beings' jobs." Our research showed that for the next five years, the impact, especially I'm talking about the U.S. market, but the impact of the aging workforce is much bigger than the AI impact. We are losing experienced, skilled baby boomers every day. But the impact is not the same for all industries. Think about a percentage of engineers who are 60 years old, or a plumber who's 60 years old is different. What's happening is, for example, electricians. You can find electrician jobs that pay more than JPY 200,000. It's so important for everybody to build a data center to hire electricians. That's why also the construction workers, if you want to hire skilled workers, I think it's happening even today. We're going to lose 6 million workers in the next five years. 6 million workers. That's almost 4% of the U.S. labor force, and our current unemployment rate is 4%. I would say it sounds weird, but the AI change, AI replacement, is too slow to catch up this aging workforce change. Aging workforce is for sure happening. That's why employers are struggling to fill positions. I think the competition is getting worse every day, and that's why they're struggling to fill skilled worker positions. That's why they're happier to pay more. I think actually the vacancy cost, everybody is underestimating vacancy cost. Actually, I've been talking with some trucking companies. Some CEO told me they have more trucks than drivers. That's the challenge. That's why most big trucking companies are running CDL drivers license schools. They're trying to have the new drivers. Anyway, I understand that people are worried about AI and job relationships, but the reality is. I'm going to give you one more example. Software engineering jobs. Everybody said software engineering jobs will be automated by AI, and everybody will lose jobs. But what we're seeing is in the last 12 months, an increase of hiring demand for software engineering jobs. Why? When we dig deeper into the details, you can see a lot of new job titles, which is AI ethics managers, AI conversation designer, AI prompt designers. What I'm saying is the AI change is not a one-to-one change, meaning, okay, this AI can replace this person's job. No, it's not happening. A human being is doing many tasks, and some of the tasks were automated by AI. It requires a lot of change or a lot of redesigning for the leadership team or CEOs. That's what's happening. That's why I'm very positive that AI is not a job killer. It's more like a job redesigner. That's how I'm thinking. Maybe I'm totally wrong, but from our data point of view, that's how I feel. Hmm. Interesting. So if we face a more severe shortage of the labor force, the monetization will be accelerated, or you will change the monetization model of Indeed? It's going to be definitely a tailwind for us. Yeah. But I feel bad. You said that, oh, competition is tough. You can make more money. I feel bad, but that's the market and competition, what's happening today. Hmm. Great. Also, a growing number of companies are reporting rising AI token costs, so cutting into margins. How are you managing your overall cost to keep investing in AI, and how does that factor into the midterm ambition you said last quarter of greater than 50% of margins? Yeah. We're managing the AI cost separated into two parts, which is external AI costs, meaning external AI cost is job seeker is using AI, employers are using AI. And internal AI costs, which our employers are using AI. External AI costs, we're lucky. We're investing a lot into AI, but we are able to monetize these AI costs. We can leverage the AI cost to monetize it. So I'm not so worried about this part. Internal AI costs, I need to make sure all the AI investment costs into our employers has to be justified really well. So I need to keep checking almost every month if the productivity is getting better. The typical numbers are revenue per one employee. Is it going up? So that's how we're checking. As you guys know, the open-weight model is getting really good, and I feel like of course, we're spending a lot of money for AI, but most of them are more like electricity and computing power, not a model, I think. We're very flexible to have based on the latency, we can have open-weight model or old model. I'm not so worried about AI investment cost. That's great. So we cover a lot of interesting topics, so I'd like to open it up to the audience. If you have any questions, please feel free to raise your hands, please. Yeah, go ahead. I hate to be the first question. It's always like never very interesting. Please go first. I know. It's going to be a letdown. Let me ask you, can you talk to us a little bit about AI in Japan, and to the extent that I feel like maybe this will be a little bit different than cloud and some of the other technologies that business there has embraced. I'd just love to hear your thoughts about that, and as you think about the Japanese businesses, how you think about bringing AI domestically. Thank you. Great question. I think I'll talk about two parts. One, in general, in Japan, people are welcoming AI a lot, including even human and robots. I think the demographic change is so big. Aging workforce is they're losing labor every year, 1%. They're losing 1%. That's a huge thing. So all the employers are struggling to hire people, so they're welcoming AI. Compared to U.S., I still feel a little bit hostile and some clients or person is saying, "I'm not going to have any AI." Okay. I saw that situation. I think Japan is different. That's one thing. One more thing is regulation. AI regulation is there are two things, which is typical old school HR related regulations. In the U.S., which is Fair Credit Reporting Act. In Japan, you need a license to recommend a job. In European countries, there are many different old non-discrimination or you have to be fair type of typical HR related regulations. There are so many. That's one thing. One more thing is AI specific regulation, right? EU AI Act, and Japan has AI regulation. So for us, we have to be very careful to clear these two regulations because most of the time, old school HR regulations didn't care about the automation. That's why you have to be sure that you're very compliant for the old regulation, and also we need to very take care that the AI specific regulations, AI, EU Act, whatever. So U.S. is relatively open for the new regulation, and Japan is actually open for the new AI regulation. As you know, EU is more gray. But what we're doing is we're trying to be very compliant for both. But relatively speaking, Japan is relatively open, so we're testing a lot of new things. You need to have licenses to do the job recommendation or these type of things. Is that answering your question? Thank you for taking my question. When we look at the average revenue per job performance, curious if there's any difference between customer cohort sizes. Does the growth look similar with large enterprises, medium-sized businesses, and small businesses? Maybe could you talk about the price elasticity of each if it exists? Because I would think if you're adding the requisite value, there should be no elasticity, but sometimes small businesses don't fully appreciate things rationally, and large businesses sometimes expect discounts just because they're large. Maybe just talk about the customer reception and elasticity. Yeah. Great question. I think what I'm seeing is very clear gradation of the change of the speed. The small businesses started to use these new AI products much faster, and when it's getting bigger, slower. That's pretty much it. That's what I'm seeing. A good example is there are so many enterprise, big brand companies which have a headquarter and regional managers and store managers. Most of the time, probably I should not say it, but the store managers are so frustrated against the headquarter HR team. Headquarter HR team is saying, "Oh, every month, this is our budget. We allocated it for the next five, six months." But the store manager is like, "Hey, I need to hire logistics manager today. He left today, so I need to do the business with somebody." That's why, to be honest, majority of enterprise customers are actually posting jobs directly to Indeed, not using their tech stack because they're frustrated. This is typical frontline folks and headquarter type of situation, so they're trying to reimburse the cost. That's why, again, the enterprise clients, from my point of view, there are two different types of enterprise. It's more like enterprise-enterprise, which doesn't have any division or stores type of companies, just one headquarter can manage everything. But most of the time, if it has, let's say, 1,000 stores, they have a different decision-making process. They're acting more like small businesses. So, from my point of view, that's why I do really care about the product, is it working or not? But the good test is, when we deliver these products for small businesses, if they are okay, it is just a matter of time. Because if it is working, they will realize, "Oh, it is working." But big enterprise customers, it takes time to talk to the head of HR or headquarter people. They do not want to increase their monthly budget. But the frontline folks are different, so it takes time to convince people, and as I said, also, this time it is more like an AI change. Sometimes we need to include CFO or CEO type of C-level people, and that is why we are having World Cup sponsor, Formula One, we are going to have sponsor, and we are going to invite, and we are going to have party. Typical enterprise sales, which we have not done before. But I know how to do it. My background is more like enterprise sales, but we have just started this year, and we are seeing a good uptick, so I am positive about it. It is just a matter of time. As you said, probably, from the number of jobs point of view, of course, big enterprise customers are, we still have probably much lower penetration compared to small businesses. There should be an upside. It is a good sign. We are seeing the good uptick of the revenue growth. I think it is going to be better next year. When you are looking at the revenue per job growth, do you have some underlying sort of metrics, whether it is cost per hire or some sort of metric that you are ensuring it is sort of within certain bands, so the enterprise are seeing good results? This is a sort of a way of asking the potential pricing upside. Again, probably one tricky thing about the pricing increase, which I How can I explain it? When we were talking about cost per hire all the time, we need to talk about the, okay, you have 100 people in your talent acquisition team. AI sourcing is a good example. Okay. Think about it. You have this much recruiters, and your recruiters are spending this much time to reach out to each resumes, and AI recruiters can do this. So right now, our result is showing that we can save 11 hours per week. So we need to discuss, this is more like enterprise type of talk, like, "Hey, you can save 11 hours per week per person, and blah, blah. Okay, how much you can pay?" This is typical. And also, screening is like the big enterprise employers are having thousands of applications, and your team is spending, let's say, five minutes for each application to check resume. That's hours. You can automate it, maybe then you can reduce 100 people to 80 people, and then you can replace these automation with this investment. Whatever. That's like a cost per hire type of argument. But what's happening today is more like a time to fill, which is, I do not know if it's a good example, but I love Uber Eats. I do not want to spend time to buy ingredients and cook something, but I just want to get a nice food very quickly. So what's happening is, it's more like a time to fill. I want to fill this position as soon as possible. Are you okay to pay more money? Employers are paying. That's why, again, as I said, it's happening from the small businesses or a store manager level, regional manager level. It's going up now. That's why I know it's very tricky. Is it really cost per hire? But when we think about the value, it's actually not. The time to fill seems to be a great value for employers. That's why we can charge a lot better than last year. Is it answering your question? Sorry, maybe not. Well, I guess cost per but time to fill, is there some sort of measurement certain days will- Yeah Yeah. Our internal product KPI is number of qualified applications, and because the hire is ultimately accumulated successful speed of qualified candidates. So again, now we have kind of score for each application. If it's good, it should be accumulated qualified applications is hitting certain level, they should be hiring somebody. But that's the metrics we're trying to have, and we need to train everybody that not a quantity, quality is the real value for you. Thanks very much for taking the questions. I have three, if it is not too many. First of all, the company has the partnership with Claude previously. I was wondering, what is your experience in terms of protecting your data moat, at the same time trying to, I guess, using Claude as maybe your distribution channel or any other strategic thinking on that? The second is your premium ads basically launched in Canada and the U.S. very successfully. What is the near-term plan to launch in the other markets? The third one is, I know previously you sort of cut off your free ads, majority in the U.S. market. What is the next step in terms of the free ads? Okay. What was the first question? The first one was, how is your experience when you partner with Claude? Claude? We are using everything, Claude and Gemini. Gemini is good at extracting PDF file, whatever, and ChatGPT is good at this and that, and Claude is good at this. Also, we have Salesforce and Salesforce Claude, we are using it. The second question is? When you plan to launch the premium ads outside of Outside the U.S. North American market. Okay. Yeah. We're expanding it. As I said, the EU have different regulation where we have to be compliant 100%, and each countries have different regulations. Again, it's more like old HR regulations. So we have to be very prepared really well to expand the expansion. But we've been doing it. It's going to be better next year, and probably even today. Would ANZ be falling to next year plan as well? We're starting to test new countries. That's happening. I'm sorry, did you want to pick, choose someone else? Or may I just- Yeah, you can go ahead. Yeah. Let me ask you, going back, we were actually your investors at the IPO, Wade, many, many years ago, and it was very different Thank you. Yeah. It was great, right? It was great, and it was a very different company before Indeed. It was just as good, but very, very different. And what it was, it was the place where great Japanese entrepreneurs could express themselves and could build a great business. I wonder, as you think about Recruit x Indeed, whether with AI there is a new opportunity to be a leading technology company in Japan again, and whether you care to think about old Recruit, kind of Tokyo Recruit rather than just Indeed, global Indeed. Great question. Thank you. Is it okay to go over time a little bit? Yes, if you have time. Okay. Okay. Thank you. I think me personally, I really believe that all the internet technology change is just the bedrock for the AI change. AI innovation is such a great opportunity for all IT companies, and I just want to own. HR is relatively easy to monetize it because it is crystal clear there are so many manual tasks, and we have any other two-sided marketplace models, and we are trying to copy what Indeed is doing. If we can find a manual tasks, okay, can we have AI to automate it? That is what we are trying to do. We will see. I know you have a lot of questions, but it is about time. Thank you very much for your time today, Idekoba-san, and also thank you all for joining us today. Have a great day. Thank you. Thank you so much.