How to Screen CVs Faster: 500 Applications, One Founder

The fastest way to screen CVs is to stop reading them in order. Decide what good looks like before the job goes live, score every CV against the same four or five skills, and spend real time only on the top tenth. That is how to screen CVs faster without being unfair to anyone, even when 500 applications land in a week.

Who this is for: founders and first managers who are hiring without a recruiter, and drowning in applications for one role.

Why reading CVs in order fails

When you open the first CV, you are fresh. By the eightieth, you are tired, and you are comparing each person to whoever you read last. The candidate at number 412 gets ten seconds and a worse mood. None of that has anything to do with how good they are.

Reading in order also hides the real problem: most of us start reading before we have decided what we are looking for. So the bar moves. A CV with a famous company name feels strong. A CV with a typo feels weak. Neither tells you much about whether the person can do the job.

The fix is not to read faster. It is to read less, and to read every CV the same way.

Step 1: write the scorecard before the job post

Before you post the role, write down what the person will do in their first 90 days. Not their title. The actual work. "Close the books every month by the 5th." "Run 30 customer calls a week." "Ship the Android app's payments flow."

From that, pull out the skills that matter. Then split them into two groups:

• Must-haves: without these, the person cannot do the job. Keep this list short, two or three at most.
• Strong pluses: these make someone better, but you could teach them.

Everything else, such as the college, the brand names and the number of years, stays off the scorecard unless it truly predicts the work. Years of experience is a weak signal on its own. Two people with five years can be miles apart.

Here is what a scorecard looks like for a first finance hire:

The 1 to 4 scale is on purpose. There is no middle score to hide in. You have to decide whether someone leans good or leans weak.

Write the job post from the scorecard

Once the scorecard exists, the job post almost writes itself, and it brings in fewer weak matches. Three things help:

• Lead with the work. Open with the three or four things the person will do in their first 90 days, in the same words as the scorecard.
• State the must-haves plainly. "You have filed GST returns for a company" filters better than "strong compliance knowledge". People who don't have it will mostly not apply, which saves both of you time.
• Ask one short question. For example: "In two lines, tell us about a month-end you ran." The answers are quick to read and tell you more than a cover letter.

Add the salary range if you can. It saves calls with people who were never going to accept, and it brings in more applications from people who fit. Say where the role is based and how many office days you expect. Say how long the process takes and how you'll get back to people. Then hold yourself to it.

Step 2: score five skills, not twenty

The temptation is to add every nice-to-have. Resist it. A scorecard with 20 lines takes as long as reading the CV properly, and the important skills get drowned by the trivial ones.

Five lines is plenty. Two or three must-haves, two or three pluses.

Now screen in two passes:

1. The must-have pass. For each CV, look only for evidence of the must-haves. If a must-have is clearly missing, close it. This pass is fast, often under a minute, and it removes most of the pile.
2. The scoring pass. For the CVs left, score each skill from 1 to 4 and add a one-line note. Sort by total score.

Your top 10 to 15 people go to a first call. Read their CVs properly now, because this is where the time is worth spending.

Keep the scores, even for people you reject. If someone asks why they were not shortlisted, you have an honest, job-related answer. And if your top picks all drop out, the next ten are already sorted.

Step 3: the 15-minute first call

The first call is not the interview. It is a quick check that the person is real, interested and in range. Fifteen minutes is enough, and it should be the same for everyone.

Use the same questions every time:

1. What made you apply for this role?
2. Tell me about the work in your last job that is closest to what this role needs. (Then one follow-up on a must-have.)
3. What is your notice period?
4. What salary are you expecting? (In India, ask for expected CTC and current CTC if they are happy to share. Explain the offer later in in-hand terms too. Our post on CTC vs in-hand salary shows how.)
5. What questions do you have for us?

Write the answers into the scorecard straight after the call, before the next one. Memory blurs quickly when you do six calls in a row.

A good first call moves the person to a proper interview within a week. A slow process loses good candidates, because good candidates have other offers.

Using AI to screen CVs fairly, with a person deciding

AI can read 500 CVs in minutes and score them against your scorecard. Used well, that saves days and can even make screening more consistent, because the software reads number 412 with the same attention as number one.

Used badly, it does real harm. So set some rules before you switch anything on.

A person makes every decision. The AI suggests a score and shows its reasons. A person reads the reasons, can disagree, and decides who moves forward and who does not. Never let software reject people on its own.

Score against the written scorecard, nothing else. The criteria should be the same job-related skills you would use yourself. Name, photo, age, gender, religion, caste, marital status, address and the college's prestige have no place in the scoring.

Check for patterns. Every few weeks, look at who is being scored low. If a group of people is consistently scored lower for reasons unrelated to the job, stop and fix the criteria.

Tell candidates. Say on the careers page that software helps read applications and that a person reviews every decision. People deserve to know, and in many places the law is moving that way. The EU's AI Act treats AI used for recruitment as high-risk, with strict duties for employers who use it there. In India, the Digital Personal Data Protection Act asks you to tell people how their personal data is used, with most duties phasing in through 2027.

Keep data only as long as you need it. Delete CVs of people you will not consider again after a sensible period, and say how long you keep them.

The point of AI here is to give you back the hours you spent reading weak matches, so you can spend them talking to strong ones. Not to take people out of the decision.

Step 4: getting back to everyone

Most companies never reply to rejected applicants. It is one of the easiest ways to stand out as an employer.

You do not need a personal note for 490 people. You need three short templates:

People who reach the interview stage should never hear nothing. A quick call to say no, with a kind reason, is remembered for years.

How the numbers work out

Here is what one founder's week can look like with this approach, for a role that drew 500 applications:

These are planning figures, not measured averages. Your numbers will differ by role. The point is the shape: most of the time goes to the people most likely to be hired. With AI doing the first read against your scorecard, the first two rows shrink from hours to a review of its suggestions.

Common mistakes

• Writing the job post before the scorecard. You end up hiring for the post, not the work.
• Too many must-haves. If everything is a must-have, nobody qualifies, and you start making exceptions at random.
• Letting brand names do the work. A famous company on a CV says where someone sat, not what they did there.
• Changing the questions between calls. You can no longer compare people fairly.
• Letting a tool decide. AI should suggest and explain. A person should decide.
• Leaving people hanging. Silence after an interview damages your name more than any rejection.

Questions people ask

How long should it take to screen one CV?

With a scorecard written before you start, most CVs take under two minutes. You are only checking them against four or five skills, not reading every line. The ones that score well get a closer look. The ones that clearly miss a must-have can be closed quickly and politely.

Is it fair to use AI to screen CVs?

It can be, if a person makes every decision, the AI scores against the same written criteria for everyone, and you check its results for patterns. Tell candidates that software helps read applications. Never let a tool reject people on its own, and never let it judge things that have nothing to do with the job.

What should I ask in a first phone screen?

Keep it to 15 minutes and ask everyone the same four or five questions: why this role, what they did in their last job that matches the work, their notice period, their salary expectations and one question on your biggest must-have skill. Write the answers in the scorecard straight after the call.

Should I reply to every applicant?

Yes. A short, kind rejection costs you seconds if you send it in batches, and it protects your name. Candidates talk, and many of today's rejections are next year's customers, referrals or hires. Send it within a week of deciding, and never leave someone waiting after an interview.

Sources

• European Commission: AI Act, regulatory framework (opens in a new tab)
• Ministry of Electronics and IT: Digital Personal Data Protection Rules, 2025 (opens in a new tab)