How freee Escaped the "Sea of Information": The askAny Model and a Knowledge Foundation for the AI Era

Shinmura began by explaining a challenge common to fast-growing startups. As headcount grows, information inevitably grows too. But because there is a limit to how much information one person can handle, even as the organization gets bigger, knowledge only gets managed within the range an individual can manage. Add remote work, and points of contact between people decrease, communication thins out, and the flow of information dwindles.

What Shinmura especially emphasized was the relationship with AI. Many startups are trying to use AI to gather information, but simply having information isn't enough; without context, AI can't handle it well. According to Shinmura, Atlassian's view is that companies aiming to use AI but failing to get results may be struggling with the knowledge management that has to come first. Iijima, the moderator, added that another factor may be that during growth phases, key people at the department-head level are hired one after another, each bringing their own way of doing things.

At freee, employees who joined in 2022 or later account for about 70% of the total. Inamura also joined in April 2022, and says the number of employees has multiplied several times since then.

It was an everyday occurrence for new hires to find themselves not knowing "what information is where" or "who to ask about what," and Inamura recalls that people were literally "drowning in a sea of information" all over the company. Because documents couldn't be found, people tended to create them from scratch, the so-called "reinventing the wheel."

What made Inamura feel a sense of urgency was that if growth continued like this, the cost of searching and the cost of resolving people's problems would rise in proportion to the organization's expansion. That would slow down the business and drive up costs. It was around this time, Inamura says, that momentum began building within the company to treat this not as something the IT department could fix on its own, but as a company-wide issue.

Around 2022, when Inamura joined, cross-team questions and work requests were mainly made through Google Forms or the internal social network. The social network had a group for each team, and people would post questions and requests there. But that didn't tell anyone who to ask about what. So "askAny" was launched as a service to consolidate inquiries to back-office functions such as accounting, HR and labor, and the IT department in one place, starting with the goal of eliminating the runaround.

The way it was launched reflected one of freee's core values, "Output → Thinking." As Inamura explains, rather than overthinking and hesitating to release something, putting out output first gets you feedback, both good and bad, and in the end lets you build something good in a short time. askAny, too, started out running on existing tools. The team received inquiries from each team on the internal social network, routed them to the appropriate team, and accumulated Q&As in a spreadsheet.

The launch went well, but the Q&As piling up in the spreadsheet remained knowledge for the askAny team's internal use only. It never got to the point of being published internally so employees could look things up and solve problems themselves. Since the operation itself was judged to be working, the team decided to bring in a tool that could scale further, and made the switch after about three months.

Responding to Iijima's question, Shinmura described the typical pattern of this kind of ad hoc tool adoption and its risks. Teams tend to start operating with tools they had used before, or tools brought in by someone on the team who knows them well. This is all the more likely when an organization is expanding rapidly and new teams keep appearing, because each team focuses on getting its own culture up and running as quickly as possible.

The result, however, is that the knowledge accumulated in those tools becomes fragmented and can't be used company-wide. Tools adopted at the team level sometimes can't be maintained as a platform, which can leave past data unusable, Shinmura says. The reasons for adoption include cost and familiarity with the tool, but the biggest cause Shinmura pointed to was that no one is designing how knowledge management should be carried out across the company. Without that design, tools become siloed, data gets scattered, and, as mentioned earlier, context is lost when you try to make use of it with AI. Shinmura's view is that for companies trying to grow by leveraging knowledge, a proliferation of separate tools is a major risk.

When it came to selecting a tool in earnest, Inamura was facing two main challenges at the time. One was handling questions, inquiries, and work requests smartly. The other was accumulating the knowledge generated in the course of handling them as a knowledge base so it could be used effectively across the company.

Inamura cited three deciding factors for choosing Atlassian products: "being able to start small," "having scalability," and "being able to start cheaply."

On the low-cost side, Jira Service Management's licensing model was a big factor. Only the members who actually handle tickets need a license, and general employees can use it essentially for free. The company already had a sizable number of employees at the time, but only a dozen or so people were handling tickets, so it could be launched with license fees for just that number of people. Furthermore, by combining it with Confluence and building up articles, employees could find knowledge and solve problems on their own before asking a question. Inamura says this was also appealing.

As for scalability, handling things on the internal social network was flow-based, with the problem that the history of responses just scrolled away. Neither knowledge nor response history accumulated well. By turning requests into tickets, Q&As and work history would be accumulated as a database, making it possible to aggregate, analyze, and improve while looking at the data. Inamura judged that this would lead to expanded use later on, in other words, to achieving scalability. Iijima nodded, saying that whether something can be started cheaply and kept in use for the long haul is a point that many executives, Iijima included, care about.

Shinmura then introduced the product lineup. The concept behind Atlassian's products is to unleash the full potential of teams. A team here means not only small units such as projects or departments, but the entire company viewed as one big team.

Among the main products, Jira handles task management and the implementation of business processes, while Confluence handles managing and sharing knowledge. There are also Jira Service Management, products that support management, and products that support software development. According to Shinmura, combining these makes it possible to link knowledge with "whose work it is" and "which organization it belongs to." This gives information context, and helps build the foundation for knowledge management that can be leveraged by AI.

On scalability, Shinmura explained that the products can cover everything from organizations of a few people to companies with tens of thousands. Beyond handling user counts, another feature Shinmura cited is the ability to add capabilities such as user management and security as the company grows.

Inamura explained that askAny evolved by combining Jira Service Management and Confluence around May 2023, and showed a graph on screen.

Before adoption, inquiries were handled on the internal social network. Because inquiries were consolidated, their number (the sum of orange and red on the graph) kept rising for a while. When Jira Service Management was introduced in May, the blue line began climbing steadily. Blue represents the number of views of knowledge accumulated in Confluence. According to Inamura, the data confirmed that as knowledge was viewed, the self-service resolution rate rose and the actual number of tickets fell.

freee named this mechanism, which accumulates knowledge to raise the self-service resolution rate and reduce the questions and inquiries flying around the company, the "askAny model." The effort originally started as a way to address "not knowing who to ask" and "the runaround," but once these results became known within the company, its use spread. Teams such as accounting and security, which had previously handled inquiries within their own departments, began using Jira Service Management themselves rather than leaving it to askAny, and started deliberately organizing knowledge in Confluence.

Iijima mentioned that when asking ChatGPT beforehand, the answer had suggested that Atlassian products are aimed at engineers, and asked what they were actually like to use.

Inamura said that the message that accumulating knowledge would make everyone's work easier had gotten through clearly within the company, and then offered a candid impression. Inamura had never used Atlassian products before joining freee and found them hard to get into at the adoption stage. Because they have so many features, there were also voices within the company saying they didn't know how far they needed to go in mastering them.

On the other hand, what Inamura came to appreciate after using them was how easy it is to integrate the products with each other. There are many tools for ticket management, project management, and knowledge management, but Inamura believes few tools do it all in one. The appeal of Atlassian products is that, when combined, they can be managed and used as if they were a single tool. freee's mission is "Make small businesses the stars of the world," and its vision is an "integrated management platform" that lets anyone run a business. Inamura said there is something in common with Atlassian in aiming for various tools to connect on their own so that the work of the people using them becomes easier.

On the keys to successful adoption, Shinmura said that adopting a tool is only the starting point. You need to embed the tool into business processes and get to the point where the organization uses it well. According to Shinmura, among customers who make good use of the tools, the team driving the effort walks alongside the frontline teams. Rather than just teaching how to use the tool, they think through together how to use it, what ways of working will make the tool effective, and how to feed the tool's results back into the work.

When asked what distinguishes organizations that master the tools from those that don't, Shinmura again emphasized this walking alongside. Shinmura added that the keys to reaching mastery are not having one-way communication from the driving team to the front lines, but instead gathering up the issues that arise as people realize new ways of working with the tool and working together all the way through to improvements, and being able to spread the hard-won know-how gained there across the company.

At freee, knowledge management was also incorporated into the company-wide OKRs. According to Inamura, Confluence was introduced in May 2023 amid growing momentum from 2022 to 2023. At that time, however, it was not yet a company-wide tool, and each organization was using its own separate knowledge management tool.

Inamura believes that "fragmented tools lead to fragmented communication," and a policy was set to consolidate knowledge management tools into one. The decision to roll out Confluence company-wide was made at the start of 2024. As its effects became visible, the keyword "knowledge management" was included in the company-wide annual OKRs starting "July of last year." As specific goals, "halving the inquiries flying around the company that require a human to respond" and "eliminating reinventing the wheel" were set for the whole company, becoming a message from management. Iijima took this to mean that the commitment to go as far as building it into OKRs is likely important for maximizing the effectiveness of the tools.

Still, Inamura laughed that just because something is in the company-wide OKRs doesn't mean everyone will naturally use it, and recalled that steady, persistent efforts to promote usage were needed.

As basic measures, the team held company-wide briefings on basic operations. Since new hires vary in whether they have prior experience, training for new joiners covered how to use the tools as well as internal rules and guidelines. The team also held events to get everyone trying it out together.

But that alone only gets people to the state of "being able to use it." What Inamura prioritized was letting people experience firsthand how using it makes things more convenient. In parallel with building out Confluence, the company was also developing an AI chatbot. The form that is running today is a system for asking questions of Confluence. The more Confluence is used, the more the chatbot grows, which made it easy to show the benefits of using it.

Another challenge was migrating existing documents. Gathering manuals and knowledge that had been scattered across Google Drive and elsewhere into Confluence takes a great deal of effort. While each team was asked to proceed on its own initiative, some teams couldn't finish the job. So Inamura's team created a system and structure to handle the migration on their behalf, taking on the work on a request basis. Inamura cited this as another form of "walking alongside."

Looking ahead, Inamura said there is a sense that considerable progress has been made in building the foundation and putting it to use. A sense has spread among employees that the internal AI chatbot grows and becomes more useful the more it is used, and it now handles about 10,000 questions a month. The halving of inquiries set in the company-wide OKRs has also been achieved.

According to Inamura, however, the current knowledge foundation is merely at the stage of having centralized manuals and rules that already existed. The next move is to turn tacit knowledge that remains with individuals, such as the skills and best practices of high performers, into explicit knowledge and make it usable. Inamura believes this will lead to employee growth and help customers. Inamura wants to use AI for this. Until now, knowledge has been written, accumulated, and used by hand, but knowledge accumulating naturally in the course of work, with AI providing practices suited to the problem at hand: Inamura sees this becoming feasible and said the company will take on that challenge.

For startups that recognize the importance of an information foundation but haven't gotten started, Inamura, prefacing the remarks by saying it would be presumptuous to give advice, raised two points. One is to start small first, and then quickly think about the next move based on what you learn and act on it. The other concerns AI. Even though AI has become this widespread and will grow further, AI won't solve everything. Inamura believes that a company's own unique knowledge and practices are precisely what will become a source of competitive advantage in the AI era. That is why the sooner you build a knowledge foundation, the better, and Inamura said they would recommend starting immediately if things are chaotic right now.

Shinmura said that what Atlassian is aiming for going forward is to grow not just as a tool provider, but as a platform that helps improve customers' ways of working as a whole. Behind this is the concept of the "System of Work." As Shinmura explained, it is Atlassian's philosophy on how technology-driven companies should work, and it supports how well technology teams and business teams can collaborate and how much impact the company as a whole can make on the market. Shinmura said Atlassian wants to evolve into an integrated platform that supports the performance of the entire organization, rather than performance at the project or department level.

In closing, Inamura pointed out that building a foundation tends to be postponed as an investment. In business, investment in building and selling comes first, and as a result the foundation is often neglected and the business slows down. Because AI has greatly increased the return on investing in the foundation right now, Inamura urged companies to put just as much effort into building it.

Shinmura said that a knowledge foundation may at first feel like a major cost for a startup, but in a society where AI is the norm, it becomes not merely an investment for building strength but an indispensable platform. Noting that Atlassian also offers free licenses for small teams of around 10 people, Shinmura encouraged companies to try things on a small scale without making a big investment right away, build up successes, and expand gradually. Doing so, Shinmura concluded, may eliminate the need to spend heavily on reorganizing and reintegrating information once the organization has grown large.