The AI Chasm
Getting a design interview used to be about convincing another person that you were worth talking to. You sent in your resume, someone looked at your portfolio, and if something caught their attention, you got a conversation where judgment happened. Someone could ask questions, notice something unexpected, connect experiences that didn't appear related on paper, or simply think, "That's interesting. Tell me more."
Increasingly, the hiring process begin the same way anymore. A resume written by an AI is submitted, another AI evaluates it, screens, and only then a human might become involved. The candidate's first challenge is no longer, "How do I convince the hiring manager?" It is, "How do I get the AI to let me reach the hiring manager?"
That creates a paradox. The things that make it easy for an automated system to recognize you are not the things that make you interesting to a human. An AI benefits from familiar terminology, recognizable job titles, matching skills, conventional qualifications and experience that maps neatly to a job description. A human can value the opposite: unusual experiences, unconventional thinking, curiosity, potential, personality and a point of view that doesn't fit neatly into a predefined category. To get past the AI, you may have to make yourself more generic. To get noticed by the human, you need to show what makes you distinctive.
To get through the AI, you need to fit in. To get noticed by a human, you need to stand out.
The Proxy Becoming the Decision
I have experienced a version of this problem long before AI became part of hiring. I was shortlisted for a job with a large Tech company, but they decided not to proceed because my graduation was not full-time. I didn't reach the stage where I could have a functional conversation with someone and demonstrate what I could do. I was rejected before that conversation happened. I tried to reason, I pointed out that I had already worked for another company of the same group. I pointed out that my degree was completely unrelated to the design role anyway, so if an unrelated degree was acceptable, it seemed difficult to understand why the mode of study should determine whether I could be considered.
Sadly, that didn't change the decision. The qualification had become the decision. It didn't matter that the qualification was only a proxy for what the company actually wanted to know: whether I could do the job. The rule was easier to apply than the judgment required to decide whether I was an exception. That distinction matters enormously when we start talking about AI. A degree is a proxy. Years of experience are a proxy. Job titles are proxies. Keywords are proxies. We use them because the thing we really want to evaluate, someone's ability to perform, is harder and more expensive to evaluate directly.
The danger begins when satisfying the proxy becomes more important than demonstrating the capability the proxy was supposed to measure.
The Cold Efficiency of AI
The real problem of AI isn't that AI is incapable of evaluating people. It is that AI is extremely good at applying the rules we give it. A human can look at a rule and occasionally decide that the situation is unusual enough to warrant an exception. A machine doesn't have the same natural inclination to say, "I know the rule says this, but let's look at the evidence." If a design degree is mandated, a candidate without one can be rejected. If five years of experience are required, someone with four can be screened out. If the system is looking for particular titles or keywords, an unconventional career can become difficult to recognize. The AI isn't malicious, it is doing what it was asked to do. That is precisely what makes the cold efficiency of AI so powerful and, in certain hiring situations, so problematic.
Humans have always followed rules blindly. My own experience is evidence of that. At HCL, I saw a situation where the team and leadership wanted to hire an exceptional designer, but HR would not allow it because he did not have a graduation degree. The problem therefore did not begin with AI. Organizations have always created qualification criteria and used them as filters. AI changes the economics of applying those criteria. It makes proxy decisions faster, cheaper, more consistent and easier to scale. More importantly, it can make those decisions easier to defend. A recruiter can say, "The candidate did not meet the criteria," and the system provides an apparently objective explanation for why the person disappeared from the process.
The danger isn't that AI will always make the wrong decision. It is that it can make a proxy decision with cold efficiency.
Taking a Chance on The Misfit
The irony in my own career that makes this personal. Someone chose to ignore my degree issue. That was not the only time someone took a chance on me. My earlier jobs involved people giving me an opportunity when I didn't have the right qualifications, years of experience or polished portfolio. Someone looked beyond the obvious signals and decide that there was enough potential to take a chance. Those decisions were made by people. Those decisions that could not have been justified through a neat checklist, but they were decisions that helped shape my career.
I wonder what would have happened if I had started my job search as a fresher in today's environment. There was less formal education for the profession I entered, and much of what I learned came through experience and opportunity. If an AI had been instructed to reject candidates who didn't meet the qualification criteria, I may never have reached the person who was willing to take a chance on me. While every unconventional candidate should not be hired, this demonstrate something more important: potential can only be discovered if someone takes the effort to look for it and is willing to take a decision based on their gutfeel.
“None of These... Let Me Explain.”
AI-driven hiring can create an experience similar to calling an IVR with a unique problem. It operates in a narrow spectrum and expects you to have a problem from a set of predefined list. If it is there, it is efficient, if not, then good luck to you getting it resolved, or even to get to talk to someone who understands that. The system may not really be asking, "What's your story?" It is explicitly asking, "Which of my predefined categories do you belong to?"
That matters most to freshers, career changers, unconventional designers and people with non-linear careers - exactly the kind of people who need a human interaction instead of an AI because there can be gem hidden there the AI will just weed out. A senior designer often has years of experience, recognizable titles, established companies, projects and outcomes. Those are strong signals that a machine can understand. A fresher's strongest qualities may be curiosity, self-directed learning, unusual projects, unconventional thinking and potential. Those qualities are much harder to reduce to a set of predefined criteria. The people who most need a human to say, "I'm not sure this candidate fits the profile, but there is something interesting here," may be the people least likely to survive the first automated filter.
The Human Override That Nobody Wants to Use
So, are you thinking to keep a human in the loop? It can help, but there is no guarantee. With the efficiency of AI comes the expectation of a higher output. How often will recruiters genuinely revisit data that has already been validated by AI when the expectation is that AI should make them more productive? If the machine has already screened hundreds of candidates, the organizational expectation may naturally become that the recruiter should trust those results.
The other challenge is the penalty for making the wrong decision. Imagine a recruiter overriding the AI, hiring a candidate the system rejected, and that candidate subsequently proving to be a poor fit. How kind would you think the management will be to that recruiter when they come to know a candidate rejected by AI was hired?
This creates an asymmetry, the penalty for hiring the wrong candidate is visible, attributable and immediate. The cost of not hiring the right candidate is usually invisible. If an exceptional designer is rejected because they don't have a degree, nobody may ever know what was lost. The candidate disappears and the organization moves on. But if a recruiter takes a chance on a candidate who turns out to be wrong for the role, there is a person who made that decision and a decision that can be questioned. Following the AI therefore becomes a safer decision, even when everyone involved says that humans have the final say.
This is why AI makes proxy decisions easier. The issue isn't simply that a recruiter believes the AI is infallible. The AI gives them a defensible reason for not taking a risk. "The AI rejected him" is a much easier explanation than "I disagreed with the system, took a chance on someone who didn't meet the criteria, and I was wrong." When the incentives are structured this way, the human override can exist in the process without actually preserving much human judgment. The human technically has the authority to disagree with the machine, but the organization may give that person very little reason to exercise it.
When Efficiency Replaces Serendipity
There is nothing inherently wrong with making hiring more efficient. Companies cannot interview every person who applies, and recruiters have always needed ways to manage large volumes of candidates. AI can help with that. It can reduce repetitive work, conduct structured first-level interviews, organize information and help recruiters focus their time. I would not want to remove those capabilities. The question is what happens when efficiency becomes the dominant objective and the system becomes optimized for finding candidates who resemble the profile we already know how to recognize.
Human judgment has a frustrating characteristic that makes it difficult to measure: sometimes it creates serendipity. Someone sees a candidate who doesn't quite fit, asks another question, hears an unexpected answer and changes their mind. Someone notices that a person's experience in one field might be valuable in another. Someone takes a chance. Sometimes that decision is wrong. But sometimes it is precisely the decision that creates an exceptional hire. My own career contains several such moments. I benefited from people who were willing to look beyond the obvious criteria and make a judgment that wasn't guaranteed to work.
What happens when the people who most need human judgment are the people least likely to survive the automated filter?
That is the question I think we need to ask as AI becomes a bigger part of hiring. The answer isn't to reject AI or pretend that human hiring has always been fair. It hasn't. My own experiences with TCS and HCL demonstrate that humans can follow bad proxies just as rigidly as machines can. The difference is that humans also have the capacity to recognize when the rule shouldn't apply. AI gives us extraordinary efficiency, but efficiency doesn't tell us when an exception is worth making.
If we are not careful, we may build hiring systems that are extremely good at identifying people who fit the rules while becoming progressively worse at discovering people who could have exceeded them.
