500 CVs. And Still: “We Can’t Find the Right Person.” What I’d Change About Hiring in the Age of AI

Something about today’s job market doesn’t add up.

When I post a job, I can receive hundreds of applications.

And among them are good people.

Experienced professionals. People with solid careers behind them. People who have built teams, solved difficult problems, managed change and delivered real results.

Yet I still hear companies say:

“We can’t find good people.”

Then I look at the market from the other side.

I talk to friends, former colleagues and experienced professionals who are looking for their next opportunity.

Sometimes I have been on that side myself.

And I hear:

“There are no good jobs out there.”

That is the paradox.

Companies are looking for people.

Good people are looking for companies.

We have LinkedIn, job platforms, applicant tracking systems, databases, algorithms and now AI.

We have more technology than ever to help the two sides find each other.

So why are they still missing each other?

After almost 20 years, I think part of the answer is this:

We may not have only a talent shortage. We may also have a matching problem.

And AI can help us solve it.

But it can also make a bad matching process much faster.

AI Doesn’t Know What “The Right Person” Means Until We Do

This is where I would start.

Imagine we create a job description with 20 requirements and treat almost all of them as essential.

The system searches for them.

Efficiently.

But what if we defined the profile incorrectly?

What if only six of those requirements are genuinely essential?

What if the best candidate comes from another industry?

What if they have a different job title?

What if their experience is highly relevant but their CV uses different language?

A better filter doesn’t automatically fix the wrong question.

Sometimes it simply applies it more efficiently.

So before ATS, algorithms or AI, I would start somewhere else.

With the actual problem the person needs to solve.

1. Don’t Start With the CV. Start With the Problem.

Before writing the job description, I would ask the hiring manager one question:

“Twelve months from now, what must this person have achieved for you to say this was a great hire?”

Not:

What experience should they have?

But:

What must they be able to do?

Build a team?

Grow revenue?

Transform a process?

Enter a new market?

Stabilize an operation?

Implement a new system?

Change a culture?

Once we understand the outcome, we can search for people capable of producing it.

Not simply people whose CVs look most like the job description.

2. Separate “Must Have” From “Can Learn”

For every role, I would put requirements into three categories:

MUST HAVE
Without this, the person genuinely cannot do the job.

CAN LEARN
Skills or knowledge a strong professional can acquire within a reasonable period.

NICE TO HAVE
Useful, certainly. But not a reason to eliminate an otherwise excellent candidate.

This sounds simple.

But I think many companies would be surprised by how many things currently sitting under must have actually belong somewhere else.

Every unnecessary mandatory requirement is another opportunity to eliminate someone who could have been excellent.

And then we wonder why the shortlist is so small.

3. Use AI to Discover More — Not Just Eliminate Faster

This is where I think AI can become genuinely valuable.

Instead of asking only:

“Which of these 500 candidates best matches this job description?”

I would also want to ask:

“Who has transferable skills?”

“Who has solved a similar problem in a different industry?”

“Who doesn’t have the job title we expected but has relevant experience?”

“Who is close to the profile and could learn what is missing?”

And perhaps most importantly:

“Which potentially strong candidates am I about to miss because of the criteria I chose?”

That changes the role of AI.

It stops being only a filter.

It becomes a discovery tool.

And that, to me, is a much more interesting use of the technology.

4. Look at Some of the People the System Didn’t Choose

If I were using automated screening extensively, I would create a simple control.

Periodically, I would take a sample of candidates ranked lower by the system and review them myself.

Maybe 20.

Maybe 30.

Not hundreds.

Just enough to test the filter.

If I repeatedly discover people there whom I would have interviewed, I have learned something important.

Maybe I don’t have a talent problem. Maybe I have a filter problem.

Technology should be tested against human judgment, not simply trusted because it is efficient.

5. Don’t Confuse a Good CV With a Good Candidate

AI has changed the candidate side too.

Today, a candidate can take a job description and use AI to adapt a CV.

They can identify keywords.

Improve the wording.

Restructure their experience.

Create a polished cover letter.

Prepare answers for likely interview questions.

There is nothing inherently wrong with this.

Candidates should use the tools available to them.

But it creates an important question for employers:

Are we selecting the strongest professional — or the professional who is best at creating an application that our system likes?

Those are not necessarily the same person.

The more easily CVs can be optimized, the more I would focus on evidence of performance.

What did you build?

What did you change?

What problem did you solve?

What was actually your contribution?

What result did you produce?

And perhaps:

Here is our problem. How would you approach it?

A CV should open the conversation.

It shouldn’t become the final verdict on someone’s potential.

6. Bring the Human Into the Process Before It’s Too Late

I am not suggesting that recruiters manually interview 500 applicants.

That isn’t realistic.

Technology is necessary when volumes are high.

But there will always be candidates in a grey zone.

They don't look perfect on paper.

Their career isn't linear.

Their industry is different.

Their title doesn't quite match.

Yet something in their experience suggests they may be worth understanding.

Sometimes I would rather have a 15-minute conversation than apply another three filters to their CV.

Or use a relevant work sample.

Or give them a real problem the company is facing and ask how they would think about it.

Suddenly, we are no longer evaluating who writes the best CV.

We are beginning to see how the person behind the CV thinks.

7. Measure Whether Your Matching System Actually Works

Recruitment shouldn't finish when the contract is signed.

If I wanted to know whether my hiring process was good, I wouldn't look only at:

Time-to-hire.

I would also ask:

Who is performing well after six months?

Who is performing well after a year?

Who stayed?

Who grew?

Who created more value than we expected?

And then I would ask something even more interesting:

How did our recruitment system originally rank those people?

Imagine discovering that some of your best employees were not your highest-ranked candidates.

That is valuable information.

Because the objective of recruitment isn't to produce a beautiful shortlist.

It is to make a good hire.

After Almost 20 Years, This Is the Matching Process I Would Use

If I had to reduce my approach to one simple framework, it would be:

PROBLEM → SKILLS → POTENTIAL → AI → HUMAN JUDGMENT → MATCH → RESULTS

PROBLEM
What does this person actually need to solve?

SKILLS
What must they genuinely know from day one?

POTENTIAL
What can they transfer, adapt or learn?

AI
Use technology to manage volume, identify patterns and discover candidates you might otherwise miss.

HUMAN JUDGMENT
Let people evaluate context, nuance, motivation and potential.

MATCH
Ask whether the fit works in both directions: candidate for company and company for candidate.

RESULTS
Look at what happens after the hire and use that evidence to improve the next one.

Then repeat.

That is where AI becomes especially useful: not as the final judge, but as part of a system that keeps learning.

Perfect Matching Is Not a Perfect CV

After almost 20 years, I don't believe there is a formula that guarantees the perfect hire.

People are too complex.

Organizations are too complex.

But I do believe we can significantly improve the probability of a good match.

And paradoxically, now that we have AI, I think we need to pay even more attention to the human part of recruitment.

Let AI help us see more, not make us see less.

Use it for volume.

Use it for analysis.

Use it for discovery.

Use it to challenge your assumptions.

But keep human judgment where it matters.

Because the objective of recruitment isn't to find the CV that looks most like the job description.

It is to find the person who can create the most value in that role — and the role in which that person has the best chance to succeed.

That, to me, is real matching.

So the next time 500 applications arrive and the conclusion is:

“There’s nobody suitable.”

I would ask one more question before reposting the job:

“Is the right person really not there — or did our process simply fail to see them?”