ACG Resources | Why AI Can’t Do What I Do: The Real Limitations of Artificial Intelligence in Recruiting

Let’s Be Honest About the Hype

I’ll be honest: I considered not writing this piece. Any recruiter who goes on record questioning AI right now risks sounding like someone who didn’t see the internet coming. That’s not where I’m coming from. I use technology every day. I’m always looking for ways to work smarter. But after thirty-plus years in recruiting — placing professionals in financial services, compliance, legal, and real estate — I’ve got some observations I think are worth putting on the table, because I don’t think they’re being said clearly enough.

The pitch for AI in recruiting usually goes something like this: it’s faster, it’s cheaper, it removes bias, it scales. And look, parts of that are true. AI can source candidates at a volume no human could match. It can handle scheduling and workflow logistics. Those are real process improvements. But somewhere along the way, the conversation jumped from “AI can help with recruiting” to “AI can do recruiting.” Those are not the same thing, and I think many companies are going to learn that the hard way.

Having had as many years in this industry as I have, I’ve experienced countless transitions.  I started when we used paper applications and file cabinets, index cards and notebooks for job orders, and typewriters for correspondence — rotary phones on the desk, no voicemail, no email, no internet. (I know. I sound ancient.) Over the years, we moved from electric typewriters to word processors, then to computers — my first one ran on a rudimentary DOS program that anyone under fifty won’t recognize from a description. We submitted resumes by fax before email existed. Each of those shifts felt significant at the time, and each one genuinely changed how we worked. But none of them replaced the part of recruiting that matters. Job boards were supposed to make us obsolete. Then LinkedIn. I remember when a well-maintained database felt like the future. The tools kept evolving; the fundamentals never did. That part is still human. Here’s why I believe that, and why I think it’s going to stay that way.

A Resume Is Not a Person

The most fundamental problem with AI-driven recruiting is this: it reads resumes, not people. And in this business, those are two very different things.

A resume tells you where someone has been. It doesn’t tell you how they handled a manager who undercut them at every turn, or how they responded the last time a deal fell apart at the finish line, or whether they’re the kind of person who takes ownership when something goes wrong or finds someone else to point at. Those things matter enormously. In many cases, they matter more than the credentials or the title.

I’ve placed candidates over the years who would never have cleared an AI screen. Their backgrounds were too unconventional, their career paths too nonlinear. But I’d talked to them. I knew what they were made of. And I knew what the client needed, which wasn’t always what the job description said. Some of those placements became the client’s best hires. That’s not a data point an algorithm has access to.

What makes a placement work — really work, the kind where the client calls you six months later to say thank you — almost never comes down to whether the candidate checked every box on the spec. It comes down to fit in the deepest sense of the word. Do they think the way this team thinks? Can they build credibility quickly in this culture? Are they excited about this, or are they just looking for a paycheck and a title upgrade? You find out the answer to those questions in a conversation. Not in a keyword match.

You Can’t Automate Trust

My clients don’t come back to me because I have better technology than the next firm. They come back because they trust my judgment. They know that when I put someone in front of them, I’ve already done the work — I’ve assessed the person, I’ve thought hard about the fit, and I’m standing behind that candidate. That trust took years to build. You can’t replicate it with a platform, no matter how sophisticated.

The same goes for candidates. The people I most want to reach — the ones who are employed, performing well, not actively looking, but open to the right conversation — those candidates do not respond to automated outreach. They get dozens of generic InMails a week. They ignore them. What they respond to is a call from someone they know, or someone who was referred to them by someone they know. They’ll have a real conversation because they’re talking to a person who they believe has something specific and relevant to say to them.

And those conversations go places that matter. A candidate will tell me — in confidence — that they’re not in a rush, but they’ve been feeling restless. Or that compensation isn’t the issue, but they haven’t felt challenged in two years. Or that they’re open to a move, but their spouse has concerns, and the timing needs to be right. None of that shows up anywhere a machine can find it. It comes out because the person on the other end of the call has built enough credibility to make the conversation feel safe.

Let me give you a concrete example — one that has stayed with me for decades, and one that I think makes this point better than any argument I could construct.

Early in my career — long before computers, long before any of the sourcing tools that recruiters take for granted today — I was working on a search for a client who was opening an operation in Bermuda and needed a COO to run it. The assignment had two distinct challenges. First, finding someone with the right operational background and leadership credentials. Second, finding someone willing to pick up their life and move their family to Bermuda. That second part sounds easier than it is. Most people, when it comes down to it, don’t move. They think about it, they get excited, and then they don’t move.

I worked through my network and eventually surfaced a candidate with strong qualifications — someone I knew personally, not just professionally. His name was Dick. And what I knew about Dick, beyond his resume, was that he was an avid boater. It was a serious passion, not a casual hobby. When I approached him about the Bermuda opportunity, I didn’t lead with the title or the compensation package. I led with the fact that he’d be living on an island in the Atlantic. He was interested from the first conversation.

We went through several rounds of interviews. The client liked Dick. Things progressed. And then, as often happens, we got to the offer stage, and he started to get cold feet. It’s a common pattern — the closer a candidate gets to deciding, the more the reality of what they’re committing to sets in. Families get nervous. Comfort zones push back.

That year, during negotiations, we had a late spring snowstorm. I remember it clearly. I picked up the phone and called Dick. Not to talk through the offer. Not to go over the terms again. I just asked him: had he had enough of winter? Was he ready to go boating?

He laughed. And then he said yes. Dick accepted the offer, relocated his family to Bermuda, and built a life there. Years later, he and his family told me it was the best decision they ever made.

I’ve thought about that call a lot over the years. What closed that deal wasn’t a competitive compensation package or a well-crafted offer letter. It was knowing the person — really knowing him or her — well enough to understand what they wanted from their life and to say the right thing at exactly the right moment. No algorithm was finding that candidate. No automated outreach was having that conversation. And no AI-driven process was closing that deal.

AI has no memory of the favor you did someone three years ago. It doesn’t know that you went to bat for a candidate when the client was uncertain. It can’t account for the goodwill that accumulates over decades of doing this work honestly. That is the real competitive advantage in recruiting, and it has nothing to do with software.

The Nuance Problem

Every search is different. I know that sounds obvious, but I don’t think people outside the business fully appreciate what it means in practice. The same job title at two different organizations can be an entirely different role — different cultures, different expectations, different internal dynamics, different definitions of what “good” looks like in year one. Figuring that out requires a real conversation with the client. Multiple conversations, usually.

When a client tells me they need someone who is a “self-starter,” I don’t take that at face value. I ask what it specifically means to them. Sometimes it means they’re understaffed and need someone who doesn’t require handholding. Sometimes it meant the last person in the role waited to be told what to do, which drove everyone crazy. Sometimes it means something else entirely. The words on a job description are a starting point, not a spec sheet. Getting to what’s underneath them requires dialogue and experience, not pattern-matching.

Then there are searches that involve a level of sensitivity that requires real judgment — situations where the incumbent doesn’t know they’re being replaced, or where there’s organizational politics that could blow the whole thing up if handled wrong. Those searches require discretion, timing, and the ability to read people accurately throughout a process that can last months. That is not something you hand off to an algorithm and check on periodically.

Regarding Bias

One of the strongest arguments for AI in recruiting is that it removes human bias from the screening process. I understand the appeal of that idea. Bias in hiring is a real problem, and if technology could genuinely address it, that would be meaningful. But I think this argument is more complicated than it’s usually presented.

AI models learn from historical data. Historical data reflects historical decisions. And historical hiring decisions in many industries and organizations reflect patterns we would now recognize as biased. When you train a model on that data, you’re not eliminating bias — you’re encoding it. The difference is that it’s now harder to see and harder to challenge, because it presents itself as an objective output rather than a human judgment call. That’s not an improvement. In some ways, it’s worse.

In the world I work in — compliance, legal, financial services, among others— a non-traditional background can be exactly what a client needs and doesn’t know how to ask for. Someone who spent ten years on the operations side before moving into compliance often brings a practical perspective that a purely credentialed candidate never developed. An algorithm running a keyword screen would filter that person out. A recruiter who’s had a real conversation with the client and a real conversation with the candidate would know to make the connection.

When the Market Gets Messy

I’ll say something that probably resonates with anyone running a recruiting desk right now: the market is unpredictable in ways that don’t fit neatly into a model. Clients who were moving quickly have slowed down. Candidates who seemed ready to make a move are getting cold feet. Searches that appeared to be closing are stuck in limbo. Some of that is macro, some of it is company-specific, and a lot of it is just the general anxiety that comes with uncertainty. Whatever the cause, it creates a situation where you need to stay very close to what’s happening — not what the data says should be happening.

AI is a pattern-recognition engine. It works well when behavior is predictable and conditions are stable. When things get volatile — when candidates are hesitating for reasons that aren’t entirely rational, when clients are stalling because of internal conversations they haven’t told you about, when the rules of the game are shifting week to week — an algorithm doesn’t know what to do with that. It keeps optimizing for conditions that no longer exist.

What you do in that environment is call people. You find out what’s going on. You figure out whether a candidate going quiet means they’re losing interest or just had a brutal week. You find out whether a client’s hesitation is about budget or about something else entirely. And then you respond to what’s real, not what the pipeline report suggests. That’s relationship work. It’s not glamorous. But it’s what keeps searches moving when everything else would stall them out.

Where AI Actually Belongs in This Process

I want to be straightforward here, because I don’t want this to read as a blanket argument against technology. There is a real and useful role for AI in recruiting. Sourcing at scale, scheduling coordination, workflow tracking, pulling together market data — these are legitimate contributions, and they free up time that can be redirected toward the work that requires judgment. A recruiter who uses AI well can cover more ground and stay more organized. That’s a real advantage.

The problem isn’t the technology. The problem is how it’s being positioned. When AI is marketed as a replacement for recruiters rather than a tool for recruiters, something important gets lost. And I think some companies are finding that out after the fact — after they’ve built a process that’s efficient on paper but produces placements that don’t stick or lose good candidates because nobody took the time to talk to them.

The firms that will do well over the next decade will be the ones that strike the right balance. Use technology to handle volume. Keep humans at the center of anything that requires judgment, relationships, or nuance. That’s not a complicated formula, but it requires being honest about what each side of that equation contributes.

Where I Come Out on This

People hire people they trust. That’s been true for as long as I’ve been in this business, and I don’t expect it to change. The tools we use will continue to evolve. The underlying dynamic won’t.

AI can process a thousand resumes in a fraction of the time it takes me to read one. What it cannot do is earn someone’s trust, read a room, push back to a client who’s about to make a mistake, or tell a candidate something they need to hear but don’t necessarily want to hear. It can’t pick up on the thing a candidate didn’t say. It can’t notice that the hiring manager described the role three different ways in thirty minutes and figure out what that means. These are things that come from experience and attention, and no version of machine learning will replicate them.

At ACG Resources, we’ve built our practice on being personal and specialized. We know our markets. We have real relationships with the people in them. When I work a search, I’m not just matching keywords to a job description — I’m thinking about what I know about this client, this candidate, this moment in this market. That context is everything. And it’s not something you can download.

So no, I’m not worried about AI replacing good recruiters. I’m concerned about companies convincing themselves it can and finding out too late what they gave up in the process.

By Len Adams CPC, CTS |Founder/CEO ACG Resources

 

 

Compliments of ACG Resources – a member of the EACCNY