In 2019, Hyundai Motor Group abolished its regular recruitment process in favor of a more flexible hiring model. This shift marked a transition from a time when companies hired large numbers of general talent for later placement to a model focused on selecting candidates who can be immediately deployed in specific roles. The key criterion for corporate hiring has now shifted from 'how many will be hired' to 'what tasks will they be assigned.'
As Jensen Huang, CEO of NVIDIA, stated, 'In the future, IT departments will be like HR departments for AI agents.' AI is becoming a new 'digital member' of companies. However, the methods for implementing AI still often resemble outdated, vague recruitment practices. Many organizations introduce general-purpose AI without clearly defining the roles it will fulfill or the outcomes they expect.
This trend is driven by a fear of missing out (FOMO) in the competitive AI landscape. Companies feel anxious when they hear that competitors are adopting AI, leading them to implement systems without first determining the specific tasks for AI. This is akin to expecting a new employee to handle everything from sales to finance and customer service without any job description or onboarding manual.
Recently, there has been a shift in this atmosphere. As the initial excitement over maximizing AI usage—known as tokenmaxxing—subsides, companies that have experienced significant AI costs are finding relief in a joy of missing out (JOMO) mindset. They are realizing that simply using more tokens does not equate to better performance. Tokens represent both costs and AI working hours. Producing reports that no one reads with hundreds of thousands of tokens is far less valuable than using fewer tokens to accurately resolve a customer inquiry or automate repetitive tasks. The focus should be on how efficiently AI is utilized to generate actual business results, not just on usage volume.
The key to enabling flexible AI hiring lies in clear job descriptions and the tacit knowledge of experienced professionals. Creating a repository filled with internal documents does not automatically make AI a competent employee. Beyond written regulations, the nuanced understanding of quality assessment developed through years of experience, the sequence and criteria for customer interactions, and the reasoning behind responses in exceptional situations are the true competitive advantages of AI. Ultimately, the success of flexible AI hiring depends not on the sheer volume of data collected but on how precisely the context of work-related tacit knowledge is translated into data.
This principle applies equally to both screen-based AI agents and physical AI operating in the field. The mechanical 'operation' of a robotic arm following a predetermined path is fundamentally different from its ability to autonomously assess product conditions, classify defects, and decide on subsequent actions. Just as a person does not become a veteran by merely reading a manual, AI must learn from real-world experiences and judgment criteria to effectively perform tasks. Only AI that has absorbed the tacit knowledge of veterans can provide credible responses and genuinely take over human tasks.
It is also important to be cautious of the approach that insists on establishing perfect governance before implementation. AI governance should not be a set of regulations completed at a desk; rather, it should be an operational system that applies tacit knowledge through specific tasks, repeatedly validating and improving performance and risks. A realistic AI implementation strategy involves starting with small tasks, accumulating successful cases, and gradually expanding the scope of application.
Now, the question surrounding AI adoption should shift from 'how advanced an AI can we create' to 'which tasks will yield the highest return on investment (ROI) when AI is deployed.' If engineering performance does not translate into business results, it cannot be considered a good investment. The era of introducing general-purpose AI out of vague FOMO is coming to an end. Just as companies hire flexibly, AI also requires specific roles.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.