On the Use of AI at HELLO WORK as Looking Toward the Future

 

On April 22, the Ministry of Health, Labour and Welfare issued a press release titled “On the Use of AI at HELLO WORK as Looking Toward the Future.”

Initially, starting in this fiscal year (FY 2025), AI will be trialed at 10 HELLO WORKs across Japan. The aim is to assess the potential of AI in performing job placement services, which have traditionally been carried out by humans. When thinking how deeply AI has become embedded in society, it seems to be natural that HELLO WORK would consider making use of it.

 Limiting the trial to just 10 locations feels narrow—I can’t help but think it would be better to conduct it nationwide, across all 544 offices. Even if it’s just a pilot program, large-scale data inputs would provide clearer insight into AI’s effectiveness. But, anyway, it must be a positive step forward.

 While AI can be applied to various tasks, its most meaningful use at HELLO WORK will likely be in matching job seekers with employers. That is apparent in the explanatory diagram included in the press release about this year’s pilot program. Although human staff will still be involved, the goal appears to be enhancing the accuracy of job matching by using AI.

 By the way, when we talk about matching job seekers with employers, what exactly means “being matched”? In other words, what information about job seekers and job providers is being input into the AI? The press release briefly touches on this, noting that one of the technical risks in introducing AI is the importance of securing, selecting, and evaluating the dataset. It states that “unless appropriate input data is selected, evaluated, and secured, the intended model capabilities cannot be achieved.”

 From a career consultant's perspective, basic resumes and work histories are insufficient. The key information that should truly be leveraged here is found in the Job Card. Although not yet widely adopted, the Job Card contains not only objective data like educational background and work history but also subjective narratives from the job seekers themselves. If AI can understand and utilize these narratives to provide career information, the job matching process would become much more effective. In that context, the people who support the creation of Job Cards would play a crucial role. I wonder whether this aspect has been considered in the design of the trial program—this is something I’ll be paying close attention to.

 As for employers, there are many pieces of information to consider—salary, benefits, industry, etc.—but perhaps this initiative could also encourage more companies to shift toward job-based employment. That is, to clarify their job descriptions (JD). Aligning this AI-based matching with the Ministry’s broader employment policy makes sense. If AI can also understand the contents of job descriptions, matching would become even more effective.

 As I write this, a thought crosses my mind: what if AI-based matching becomes so advanced that humans can no longer understand why a particular job is being recommended to a particular person? Even if it’s just for reference, would people be able to accept it? Take, for example, assessment tools used to measure job aptitude—they usually provide understandable categories, like “you are suited for research” or “you are suited for creative work.” Humans can comprehend these because they’re based on a limited number of recognizable types. But AI can classify data in much more granular ways, far beyond our cognitive capabilities, which could make it unclear whether a result is right or wrong.

Then again, maybe it’s pointless to worry about such things now. After all, stopping mid-thought is a uniquely human privilege.