Kaonavi's CPO on "Talent Intelligence": Turning Talent Data from "Records of the Past" into "Predictions of the Future"

Inoue's position is clear. Many companies have stopped at the stage of centrally managing their talent information. If data and AI can be combined to show "what to do next," HR can break free from relying on intuition and experience. The prerequisite, however, is that each company accumulates data. Throughout the interview, the conversation kept returning to how to bridge the gap between this "ideal" and the "reality that there still isn't enough data."

Asked what is happening to talent management as AI spreads, Inoue replied, "We're moving from an era of doing things by intuition and experience to one where we use data and AI to move people and make the most of them." The intuition of veterans and past successes are not bad in themselves. But they often lack reproducibility, and the big advantage of adding data and AI is that it makes decisions reproducible, Inoue said.

One example Inoue gave was signs of attrition. There are limits to detecting the warning signs that "this person might quit" through intuition and experience alone. If you know what kinds of things a person says often or how they are behaving, you can act ahead of time. Inoue sees uses like this as one of the positive sides of AI. When Iijima checked, "So it would cut down on situations where managers spend three hours talking among themselves about how 'that person seems a bit at risk,' without knowing whether the discussion means anything?" Inoue agreed.

Iijima asked whether the pace of these changes differs between Japan and other countries. Inoue said there are differences, but the biggest one is in how data is used. Overseas, digitization of data has advanced. Japanese companies do have data, but it is often managed in Excel or on paper. Inoue believes that if this information is properly digitized, it can be used in ways that also suit Japanese culture.

How data is collected varies from company to company, Inoue said. Some of the companies that come to Kaonavi for advice still manage things in Excel or on paper. Even among those that have introduced a system, some are simply filling in the required fields for the sake of it. Inoue said they feel these gaps in data utilization keenly.

Amid labor shortages, companies have to think not only about hiring but also about how to get the most from the people they already have and how to keep them from leaving. When Iijima raised this point, Inoue said the level of the inquiries they receive has changed considerably recently.

Previously, needs centered on things like "our internal DX has made no progress at all, so we want to digitize" and "we want to make operations more efficient." Recently, more sophisticated inquiries have been increasing, such as wanting to use data more for individual careers or for placement decisions.

Inoue framed this change as a series of stages. First, data was gathered through centralized management of talent information and greater operational efficiency. As the next step, inquiries about how to use that data increased. And recently, inquiries about the stage after that have been growing. Analytics features have made it possible to understand company trends, but deciding "so what should we do next?" has been slow to progress. Because expectations are high in this area, Kaonavi is designing its products in a way that leads to talent intelligence, Inoue said.

According to Inoue, the original concept of talent management is about centralizing talent information and deciding how to use the accumulated data. Use here means goal setting and evaluations based on those goals, skills management for each individual, placement, and so on. Recently, however, the meaning has been interpreted more narrowly, and there are still many cases where people think "data centralization = talent management."

The data collected centers on past evaluations, skills, personal interests and preferences, aptitude test information, and the like. Inoue explained that talent management is a system for creating a state in which all information related to a person is in one place, and for managing it.

On how it is actually used, Inoue said the following. A company that handled goal setting and evaluations internally in Excel digitizes them with Kaonavi and automates the workflow. Or résumés and work histories from before hiring were saved as PDFs somewhere, so that no one could see them even when someone wanted to; Kaonavi consolidates them so that anyone can view them. In reality, however, many companies stop there, Inoue said. They have not yet reached the stage of designing how to use the data and, based on that, deciding what data to store.

Inoue described talent intelligence as AI drawing intelligence out of talent-related data to change organizational decision-making. There is now a box for storing data, data has accumulated, and tools for visualization and analysis exist. Until now, however, it was only possible to analyze the current situation. The ideal is a state in which the system shows what to do going forward, Inoue said.

Yet on the company side, even data centralization is not yet sufficient. Asked why Kaonavi is nonetheless putting forward talent intelligence at this point, Inoue explained that they want to create a "cycle model." Suppose, for example, that a company's attrition rate is rising. Traditionally, things ended with visualization, and HR had to work hard to figure out why attrition was high and how to bring it down. With talent intelligence, Kaonavi aims to provide support all the way to offering suggestions such as "isn't this a possibility?"

However, without enough data, it can't offer sufficient suggestions either. That is precisely why they want to release the features first and clearly show that "unless more data is collected here, we can't tell whether this possibility is correct," Inoue said. When Iijima summarized, "So it gives companies a reason they have to enter data, giving them a push?" Inoue responded, "Right now, the reason to enter data is precisely what's weak." Creating that reason also leads to a sense that the data can actually be put to use, Inoue said.

Inoue calls this the "Talent Intelligence Infinite Model." Intelligence is combined with data centralization to produce suggestions. Then more data is registered, and better suggestions are produced. This is the cycle they are aiming for, Inoue said. Inoue stated that this is quite a new idea, even compared with other companies. In terms of the product, it may look like a leap, but underlying it is the desire to increase the value of registering data, Inoue said.

Asked again what talent intelligence will make possible, Inoue answered, "Talent management and the management of talent information up to now have inevitably been about the past. What we want to do now is predict the future." Inoue said this is the easiest way to put it. Whereas things used to stop at a record of someone's past achievements and experience, the big difference is the ability to predict that "they could achieve better results in this position next" or "they might be able to aim for this kind of career," Inoue explained.

On how the predictions work, Inoue prefaced the answer by saying they have only just started a PoC and that this is still ahead of them. When you consult a general-purpose AI or use it as a sounding board, you often get generic answers. Kaonavi has a full set of talent information: not only a person's interests, preferences, and personality-related information, but also past evaluations, goal settings, and their results. Using all of that, it can offer suggestions about the best career for that person and the position in which they would shine.

Inoue also noted that Kaonavi has been adopted by about 4,500 companies. By packing in the track record of use at those companies and the know-how of the internal team members who support customers, they believe they may be able to achieve things a general-purpose AI cannot, and they are moving forward with the PoC on that basis, Inoue said.

As a challenge in advancing talent intelligence, Inoue first cited the technical side. Many companies are working on products that use generative AI, but generative AI itself is still evolving rapidly. There is a lot of uncertainty, and dealing with it is difficult, Inoue said.

The other difficulty is having to assume Japanese-style employment practices, Inoue said. Overseas, a wide range of things can be done around an individual's skills, experience, and career. Japan, by contrast, does not have as much labor mobility as other countries. Rather than what individuals will do, including changing jobs, the perspective tends to focus on how to maximize the potential of the people currently in the company. That is why Inoue feels it is difficult to offer suggestions in that context.

At this point, Iijima asked how the system can give advice when, in addition to the variable of people, variables such as company-specific rules, open positions, and who sits above someone also differ from company to company. Inoue's answer was that it is first important to properly register the company's vision and MVV in Kaonavi. Unless what each department is aiming for and which KGIs and KPIs it should pursue are registered around that axis, they won't connect to individual careers and goals. So Kaonavi is making it possible to register these properly. That is its current approach, Inoue said.

When Iijima said, "So the first step is realizing you have to do this, and those who enter the data come out ahead?" Inoue agreed. Many companies still do not fully understand the value of their data. They may be entering it as part of their daily routine, but Inoue said they want companies to enter data with an eye to how it will be used.

Adding to the explanation, Inoue said that while they want to provide value to companies through data and AI, they don't think that will make it possible to do all the work. Inoue believes that 20 to 30% of the work will always remain in human hands. Final decisions and the creation of internal HR rules are areas where people must continue to provide support. For that reason, the policy of moving forward on two fronts—the product and human support—remains unchanged, Inoue said.

Asked how this differs from the talent intelligence promoted by other companies, Inoue said that the part about using accumulated data in features is something they have long provided as talent management. Going further, to the point of offering suggestions on how a company should act to grow its business, is what they call talent intelligence.

Inoue gave attrition rates as an example. Even if a company's attrition rate is 10%, you can't tell whether that's high or low. Kaonavi has more than ten years of support experience as well as data specific to that company. Based on these, it first determines whether 10% is high or low. If it's low, it shows what the contributing factors are; if it's high, it shows where the problems lie and what actions would bring it down. Inoue is trying to provide this as the concept of talent intelligence.

As a product, it will mainly take the form of features that use AI agents. If it simply answered how to lower attrition, it wouldn't be much different from a general-purpose AI. So the aim is to first show where the company currently stands—whether its numbers are good or bad—and then, if the figure is higher than at other companies, suggest causes based on the data accumulated in Kaonavi. When Iijima confirmed, "So if you ask about something people who've been inside the company all along can't judge, it will advise you that 'this isn't bad,' 'you should improve this,' and 'you should also do this'?" Inoue agreed. The ultimate goal is to become infrastructure that supports decision-making, Inoue said.

As for what's coming soon, Inoue said they have publicly announced that they will release AI agents. Among the themes they have named are the aforementioned attrition and placement decisions.

Inoue considers placement decisions extremely difficult. You can't know whether moving a certain person to a certain department will improve the company's performance. Until now, such decisions have been made by intuition and experience, or by hearing in a 1-on-1 what the person wants to do and wanting to transfer them to a department where their motivation will rise. But you can't know whether, as a result of the transfer, the person will deliver results, or whether they will struggle to adapt to the new environment and fail to do so. Using data to raise that probability has become a major theme, Inoue said. Inoue added with a laugh that they have about ten other themes in mind but can't reveal them yet.

Iijima, who also runs a company, suggested that executives will likely use it themselves more often. Inoue explained that Kaonavi's main targets are frontline managers such as department heads and section chiefs, as well as each individual employee. Inoue then noted that HR involves several layers—executives, HR directors, the HR department, frontline managers, and employees—and that executives surely want to see the state of the data.

One particularly big concern Inoue cited was succession planning. Who will be entrusted with heading this department next, or, from a management perspective, who will be the next executive officer? If the system could present candidates and show, should a given person be promoted, what perspectives they currently lack and how they should be developed, that would be very valuable to executives, Inoue said. Inoue said this is an area they also want to tackle from the perspective of senior management.

Asked whether the role of people will also change as AI spreads, Inoue answered that it "will change quite a bit." Until now, a lot of time was spent gathering data, leaving little time for the final decision. From now on, data will be gathered and possibilities and suggestions will be generated automatically. Since final decisions will be made by management and HR team members, they will be able to concentrate their time there, Inoue said.

As for HR departments, Inoue pointed out that although they should be spending time on dialogue with each individual, they are burdened with a lot of administrative work. In March, for example, they are swamped with handling April transfers, year-end evaluations, and deciding next year's salaries, leaving no room to think about the next initiatives. Inoue expects that such administrative work will decrease considerably through automation, increasing the time available for talking with people and for thinking about which HR initiatives can solve problems.

Iijima noted a difference: overseas, with job-based employment, there is a strong tendency to think in terms of combining individual abilities, whereas in Japan, achieving results as a team is more common. Iijima asked how this affects the use of AI. Inoue also acknowledged that in Japan, people often work in teams. Inoue then said that since the COVID-19 pandemic, more companies have adopted remote work (though recently there has also been a move back to the office), making it very important to visualize what team members are thinking and what they are good at. If that remains invisible, people get assigned work they don't want to do or aren't good at, and individual productivity doesn't rise, Inoue said.

Inoue also mentioned that the working population is said to decline by 40% over the next 40 years. Unless the way Japan works changes, it will no longer be possible to maintain current productivity. Raising each individual's productivity is important for the way Japan works, and Inoue said they feel strongly about wanting to support that.

Iijima remarked that "the gap between companies that have adopted this and those that haven't seems likely to widen," and asked what HR and executives should prepare now. Inoue answered, "Honestly, it's the data itself." Whether or not a company has introduced a system doesn't matter much; Inoue wants companies to start collecting information about their people first, whether in Excel or on paper. That is the real first step, Inoue said. When Iijima added that it's probably also important to register the data in a form that's easy to use later, Inoue agreed and said Kaonavi can offer advice even to companies that are still a long way from introducing a system.

On whether they actually see effects on attrition and sales, Inoue replied, "Of course we do." Attrition rates are visible, easy-to-understand numbers, but the immediate effects of transfers and placements are harder to see. That's because you can't run an A/B test comparing the same person placed in one department versus another. Even so, whether a placement was good or bad ultimately shows up in the department's results and, in turn, becomes visible as a contribution to the company's overall sales. Inoue therefore thinks it should be possible to create a situation where the results are easy to understand.

Finally, asked about the medium-term vision, Inoue introduced Kaonavi's stated vision of being a solution that aims to "maximize the power of people in Japanese organizations." Concretely, the plan is to build a structure in which intelligence sits on top of the talent platform in the form of various features.

As a result, Inoue said they want to create a society in which each working person can discover their own potential. Inoue said they would be glad if Kaonavi could become one of the keywords for that. They want to create a state in which individuals realize, "I had this kind of potential," or "This career path was open to me too," and can feel excited about their own growth. That is Inoue's outlook for the next 3 to 5 years.