In the face of AI, there are two strategies: embrace the algorithm, or embrace being human.
As a job board operator, this has been weighing on me. I see hints that AI will disrupt the industry and transform the hiring process, perhaps not for the better.
Ultimately, I’ve decided that TokyoDev will not use AI to influence the hiring process. We will instead focus on doing what AI cannot: building a supportive community, crafting high-quality resources to help individuals become stronger candidates, and leveraging our deep expertise to help companies showcase what makes them unique, attracting the talent they seek.
Challenges posed by AI
Since AI has entered common usage, it has created a number of issues in the hiring industry, both for candidates and for companies.
AI makes it incredibly easy to send customized job applications
Since online job applications became common, candidates have been able to apply for countless positions in a short period of time. Anecodetally, I’ve heard of people submitting 1,000+ job applications. The initial downside to this approach was that candidates couldn’t spend a lot of time customizing those applications, so they tended to be low quality. Now AI tools have emerged that not only allow candidates to automate their job applications, but to submit a fully-customized application based on the job description and their resume.
While I believe these tools won’t yield the best results, for many, the temptation to use them is just too high, particularly for candidates who aren’t getting job offers. This means that positions are receiving an increasing number of applications that seem to be tailored to them and are superficially passable.
The more people use such tools, the worse the problem becomes. This is particularly true for positions that support remote work globally. For example, by some estimates, there are 28 million software developers worldwide. If even only 0.01% of all developers were to use AI tooling, a position could potentially receive 2,800 such unthoughtful, AI-generated applications.
Only after human reviewers have invested some time in reading the AI-generated application will they see the cracks in the facade. Countless human hours will therefore be spent filtering out AI-generated applications, unless recruiters take the next logical step.
Large application volume makes AI screening attractive
This can lead to a technological arms race: as candidates seek to lower the effort spent on job applications, those screening the applications will seek to do the same. Recruiters have always spent only a short amount of time reviewing a given application: some estimates put it at seven seconds. In our previous hypothetical of 2,800 applications, that adds up to almost 5 ½ hours of review. This is a time sink that recruiters will jump at the chance to cut down.
Now, AI-based screening tools are starting to be built into Applicant Tracking Systems (ATS). These tools don’t automatically accept or reject candidates. Rather, they opt to assist in evaluation, usually by assigning scores based on how well candidates match the job description. You can see an example of how this works with Workable’s overview of their AI screening assistant.
Even though these tools do not explicitly reject candidates, they implicitly do so. If a recruiter sees a candidate has a low match score, they’re likely to just click the button to reject the candidate themselves. If they weren’t going to trust the evaluation and were going to read the application themselves, why would they bother to use the AI evaluation to begin with?
AI is an existential threat to job boards
Before TokyoDev was a job board, it was my personal blog with a mailing list. Despite the mailing list only having a couple hundred subscribers on it, and me only posting opportunities every once in a blue moon, several people got jobs through it. This meant we had an incredibly high signal to noise ratio.
Since turning the mailing list into Japan’s first job board for international software developers, the number of applications we process has skyrocketed. On our job board, positions that accept overseas candidates will often receive hundreds of applicants from us alone. While our clients have compared the quality of our candidates favourably to competitors, the fact remains that our signal to noise ratio has worsened.
I can see that, as AI-created applications become more common, they could grow to be an existential threat to our business. Our value lies in the superior candidates that we provide to our clients; if all of the applications through us are the same as other job boards’, why would clients choose us? Furthermore, these automated tools make it likely that a company’s own ATS will be getting ever-increasing volumes of applications. If their ATS does a good enough job of highlighting qualified candidates, there will be no need to bother with third party sites such as ours.
Why TokyoDev does not use AI screening
In the face of these challenges, I’ve seen other job boards begin to implement their own AI-based screening, scoring candidates in a similar fashion to ATS. After all, it’s incredibly easy to use services like ChatGPT to do this: any AI will happily give you back such a score, whether or not it’s meaningful or would correlate with what a skilled human evaluator would give.
Screening candidates would be illegal
In Japan, recruiting is a regulated industry. We’re regulated as a job board, not a recruitment agency. That means we’re prohibited from screening candidates or modifying the applications they make.
Some job boards in Japan have registered themselves as recruitment agencies, allowing them to do such screening. While TokyoDev could do this too, I’ve intentionally not pursued this option, as getting involved with screening would take away from our primary focus: sharing opportunities in Japan with international developers.
Transparency is a guiding principle
My belief in the importance of transparency is why we’ve been explicit about how we make money, and how that influences our operations.
An AI-based screening system is fundamentally not transparent. Such a system would mean putting candidates at the whim of an AI, the processes of which cannot be deeply or thoroughly inspected.
Those job boards utilizing AI screening could attempt to frame this as a good thing; they could say that “AI highlights your strengths,” or that it allows exceptional candidates to shine. But even so, it wouldn’t make sense to ever show how well a candidate was scored by an AI, as revealing that information to applicants would encourage and allow them to game the system, reducing the quality of applications that the client eventually receives. The process would need to be kept mysterious, which is antithetical to the way I’ve always run TokyoDev.
AI tools have been shown to be biased
Amazon used AI to screen candidates, but scrapped it after they uncovered that it preferred male candidates, and was doing things like penalizing resumes that included the word “women’s” in it. While AI has progressed a lot in the last 10 years, and it may theoretically be possible to create an unbiased tool, I think it is quite challenging to actually do so, let alone definitively prove that it doesn’t have unacceptable bias.
Tackling the gender divide in our industry is an important issue to me. TokyoDev supports communities that empower women in technology, and provides companies with advice on how to improve gender diversity. It would be counterproductive for us to inject additional bias into the hiring process.
Companies developing ATS have more resources to invest in AI
TokyoDev is a niche job board with finite resources. We’re not going to be able to build a better AI-powered screening system than companies with development teams producing ATS products. I wouldn’t want to come up with some haphazard solution myself, either.
If I were to implement some AI screening system, and it failed to identify even a single qualified candidate who might have gotten the job otherwise, that would weigh too heavily on my shoulders.
How TokyoDev works to improve the process
This is not to say that we’re doing nothing in the face of AI. Rather than leveraging AI to reject poor candidates, TokyoDev is working to increase the number of high-quality candidates.
For example, when candidates are troubled by an unclear job description or application process, they can reach out to me directly, and I am usually able to cut through the bureaucracy and find a truthful answer. This is something AI cannot do.
On a broader scale, we invest in the hiring ecosystem by building our community, educating applicants, and helping companies highlight their best qualities to attract top talent. We also work to better the software industry as a whole in Japan; these improved work conditions will then attract more qualified candidates.
Community building
I’ve been volunteering at and organizing tech events since 2010. As we’ve seen success with TokyoDev, I’ve worked to reinvest back into the local developer community, through sponsorship and more. We’ve also started our own community, both online through a Discord server, and through hosting in-person events. Our primary motivation is the value these create for our members, but they also have a positive impact on our business.
Because a community promotes peer-to-peer knowledge sharing, it’s helped our members become better candidates. A prime example is the “resume review” channel in our Discord, where members can post their resume and get feedback from others. I’ve seen firsthand how members have used the channel to develop an insta-reject resume into one that successfully passes screenings.
We also educate our members on immigration-related processes, settling into life in Japan, and successfully dealing with diverse personal and professional stresses.
Besides that, many talented developers seek out our events, to hone their skills and make connections. By having these people in our audience, it’s more likely that they’ll choose to use our job board over others, which means we have higher-caliber applicants.
Informing candidates
Alongside the peer-to-peer education members of our community get, we also invest in creating high-quality resources that can help people become better candidates. Some examples include our articles on writing resumes for jobs in Japan, using recruitment agencies in Japan, and a guide to software engineer visas in Japan.
In addition, we write articles about Japanese work culture, the process of moving to Japan, and how to live here comfortably with pets, children, and more. This is because Japan is in need of talented and senior developers, and we want to do everything we can to attract those kinds of people to the Japanese companies we work with. Good candidates are hungry for information so that they can make informed decisions; the smoother and more appealing we can make the process of immigration, the more top-notch international candidates will be interested in applying.
Besides our articles, we also conduct an annual survey of international developers living in Japan, which enables people to better understand the market. These resources help all potential candidates, regardless of where they ultimately apply.
Helping companies highlight themselves
In-demand candidates have the choice of where they want to work, so they won’t bother applying to a company that doesn’t look attractive to them.
This is especially important in Japan, where English-speaking postings are often created by using machine translation, and can be missing crucial information for international engineers or contain confusing, ambiguous English.
We also write developer stories to help companies demonstrate that their own employees have a positive view of their work environment. These stories highlight the culture of the company in a genuine way, and provide an “insider” perspective for would-be candidates.
Conclusion
TokyoDev has grown to be a much greater endeavour than my personal blog ever set out to be. Through it, I have been able to help hundreds of developers find their paths to Japan. We have made measurable change in our industry by educating people about their rights, supporting diversity, and assisting companies in understanding how to attract and retain international engineers.
AI screening contains inscrutable processes that are biased against certain classes of applicants and prevent communication on a meaningful, human level. Building it into TokyoDev would turn it into a cold, sterile service that runs counter to the warm and welcoming place we currently are.
Instead of turning to a machine to overcome modern hiring challenges, we will continue to embrace our humanity, by connecting with people on a personal level to grow and promote high-quality candidates. We will do this because we believe it’s not only the right way to hire, it’s genuinely the most effective, transparent, and helpful method.
.png)
