A recruiter at a staffing company in Noida posted an operations associate role on a Monday morning. By Wednesday, she had 340 applications. She needed to shortlist 10 by end of day because the hiring manager had blocked Thursday for interviews. Without any AI assistance, her process would look like this: open each application. Scan the resume. Check experience. Check skills. Check location. Check salary expectation. Decide. Next. At 2 minutes per application (and that’s fast), reviewing 340 profiles would take over 11 hours. She doesn’t have 11 hours. She has until 5 PM, and she’s also screening for 4 other open roles today.
So she doesn’t review 340 profiles. She reviews 30 to 40. The ones at the top of whatever sorting the system gave her. The rest sit in a queue she’ll probably never open. Application number 45 might be the best candidate in the batch. Doesn’t matter. The recruiter ran out of time before she got there.
That’s the problem AI candidate shortlisting software is solving. Not replacing the recruiter’s judgment. Replacing the 10 hours of manual scanning that happened before her judgment could even begin. The recruiter still decides who to interview. The AI decides which 30 to 40 profiles she should be looking at instead of making her scroll through 340 to find them herself.
This blog explains how that process works on Apna. Not the abstract “AI in recruitment” version. The mechanical version. What actually happens between a recruiter posting a job and the shortlist appearing on her screen.
Table of Content
- The Manual Shortlisting Problem (And Why It Breaks at Scale)
- What Apna’s AI Actually Does With 340 Applications
- The AI Calling Agent That Screens Before the Recruiter Wakes Up
- What This Means If You’re a Candidate
- FAQs
Let’s dive in!
The Manual Shortlisting Problem (And Why It Breaks at Scale)
The bottleneck in hiring isn’t finding candidates. It’s sorting them.
A company posts a customer support role in Bangalore. Within 48 hours, 280 people have applied. Maybe 60 of them are genuinely qualified. Maybe 25 are strong fits. But identifying those 25 out of 280 requires a human to open each profile, read it, assess it, and make a decision. At scale, that doesn’t happen. The human opens 30 to 50. The rest get whatever attention is left after the shortlist is built, which is usually none.
This creates a problem that affects both sides and nobody talks about it enough. The recruiter misses good candidates because she physically can’t review the full applicant pool. The candidate sends a qualified application and hears nothing because application number 73 landed after the recruiter already found her 10. Both sides lost. Not because of a bad match. Because of a capacity problem disguised as a hiring outcome.
At a 15-person startup where the founder reviews every application personally, manual screening works. At a company that receives 200+ applications per role across 8 to 10 open positions simultaneously, it doesn’t. The maths doesn’t work. The recruiter either sacrifices thoroughness (skips most profiles) or sacrifices speed (takes 2 weeks to shortlist). Either way, good candidates fall through.
AI candidate shortlisting software exists because this capacity gap became structural. Not a nice-to-have. A necessity. When a platform processes over 2.7 crore job applications in a single quarter (which Apna did in Q3 2025), human-only screening isn’t just slow. It’s physically impossible.
What Apna’s AI Actually Does With 340 Applications
The recruiter posts the role. Specifies the requirements: 1 to 3 years of experience, customer support background, English communication, CRM familiarity, Bangalore location. Applications start coming in. Here’s what happens next, behind the screen.
The AI reads every incoming application. Not the way a recruiter reads. Not subjectively. Structurally. It extracts data points from each profile: current role title, years of experience, listed skills, location, education, work history pattern. Then it compares those data points against the requirements the recruiter specified.
But this is where it gets more sophisticated than basic keyword matching. Older ATS systems matched text literally. “Customer support” in the listing. “Customer support” on the resume. Match. “Client servicing” on the resume. No match. That kind of rigid matching missed good candidates constantly because the same skill gets described 5 different ways across different industries and different resumes.
Apna’s matching works differently. The AI has learned from millions of successful hires on the platform. It knows that “client servicing” and “customer support” and “customer care executive” and “helpdesk associate” are functionally related. It knows that a candidate with “Freshdesk” and “Zendesk” in their skills probably has CRM experience even if they didn’t literally write “CRM familiarity.” It knows that someone who’s been in a customer-facing role for 2 years in Hyderabad and has indicated willingness to relocate is relevant to a Bangalore listing, even though the location doesn’t exactly match.
That intelligence, learned from the patterns of which candidates got hired for which roles across millions of interactions, is what separates AI-driven matching from the dumb keyword comparison that older systems relied on. The AI doesn’t just check whether the words match. It checks whether the person matches.
The output of all this processing is a ranked list. Not a pass/fail gate. A ranking. The candidate who matches the requirements most closely sits at position 1. The one who matches least sits at position 340. The recruiter opens her dashboard and sees the top matches first. She’s not scrolling through 340 profiles hoping to stumble on the good ones. The system surfaced the good ones and put them at the top.
She still makes the final decision. She still reads the profiles. She still uses her judgment about who seems like a fit beyond what the data can capture. But instead of spending 11 hours scanning 340 applications to find 10 good ones, she spends 45 minutes reviewing 30 pre-ranked profiles and shortlists her 10 from there. Same quality of shortlist. A fraction of the time.
And the system also categorises candidates into “matched” and “non-matched” buckets automatically. The recruiter can focus entirely on the matched pool. The non-matched pool still exists in the system if she wants to expand the search later, but the initial shortlisting happens from a curated set rather than the full firehose.
One thing worth being specific about because it matters for how candidates think about this: the AI doesn’t just rank based on the job description the recruiter wrote. It also factors in what has historically worked. If 70% of successful hires for similar customer support roles on Apna had 1 to 2 years of experience (not 3+), the system weights that pattern. If candidates who listed specific tools (Freshdesk, Intercom, Zendesk) got shortlisted at higher rates than candidates who just wrote “customer support skills,” the system learns that too. Every hire on the platform teaches the algorithm something about what a good match actually looks like for that type of role. The matching gets better over time because the data gets deeper.
The AI Calling Agent That Screens Before the Recruiter Wakes Up
This is the part that’s genuinely new. Not AI matching, which has existed in various forms for years. AI that actually talks to candidates.
Apna launched an AI Calling Agent built on its proprietary agentic AI stack. Here’s what it does in practice. The moment a recruiter posts a job on Apna, the AI Calling Agent can begin contacting matched candidates by voice. Automatically. In English, Hindi, or regional languages depending on the candidate’s profile and location.
The AI calls the candidate. Introduces the role. Asks screening questions that the recruiter configured (or that the system generated based on the job type). “Do you have experience in customer support?” “Are you comfortable with rotational shifts?” “What is your current notice period?” “Are you open to working from our Bangalore office?” The candidate answers by voice. The AI processes the responses using voice analytics, evaluates them against the screening criteria, and auto-shortlists candidates who meet the threshold.
The recruiter wakes up Tuesday morning, opens her dashboard, and sees a list of candidates who’ve already been contacted, screened, and sorted by the AI overnight. Not just matched on paper. Spoken to. With their availability confirmed, their screening answers logged, and their interest level assessed.
The numbers on this are striking. The AI Calling Agent can run over 10,000 simultaneous voice interviews with an 80%+ candidate connection rate, compared to the industry average of about 30% for manual recruiter calls. A recruiter trying to screen 100 candidates by phone would need 3 to 4 full days of calling (factoring in no-answers, callbacks, rescheduling). The AI does it overnight. For all 100. Simultaneously.
This matters especially for the hiring that happens at volume. A logistics company in Gurgaon hiring 50 delivery partners. A BPO in Hyderabad filling 30 customer support seats. A retail chain onboarding 100 store associates across 8 cities before Diwali. These are roles where the recruiter doesn’t need to personally evaluate each candidate’s career narrative. She needs to know: can they communicate? Are they available? Will they show up? The AI screens for exactly those questions at a speed and consistency that manual calling can’t match.
For the candidate, this changes the experience in a way that doesn’t get discussed enough. Traditional hiring: you apply, wait 3 to 5 days, maybe get a call, maybe don’t. With the AI Calling Agent: you apply at 10 PM, get a call at 10:30 PM (the AI doesn’t sleep, doesn’t take lunch, doesn’t go home at 6), complete a 3-minute screening conversation, and by the next morning you’re either on the shortlist or you know the role isn’t a fit. That speed of feedback, even when the answer is no, is better than 5 days of silence.
What This Means If You’re a Candidate
Everything above was explained from the recruiter’s side. But candidates interact with this system too. And understanding how AI candidate shortlisting works on Apna changes how you should set up your profile, which is the part that directly affects whether the algorithm surfaces you or skips you.
The AI reads your profile the way it reads every profile: structurally. It extracts your role title, your skills, your experience duration, your location, your work preferences. It compares those against the requirements of every open listing on the platform. If the match is strong, your profile gets surfaced to the recruiter. If the match is weak, it doesn’t.
Which means the single highest-impact thing you can do as a candidate on Apna isn’t applying to more jobs. It’s making your profile readable to the system that decides which recruiters see you.
Your headline should name the role you’re targeting. Not “Seeking opportunities” or “Passionate professional.” Those match no recruiter’s search. “Customer Support Executive | 2 Years | Freshdesk, English, Hindi” matches every recruiter searching for customer support candidates with CRM experience.
Your skills section should list tool names and specific competencies. “Freshdesk, Zendesk, Excel, English Communication, Hindi, Chat Support” gives the AI 6 matchable data points. “Communication skills, team player, hard worker” gives it zero.
Your work preferences (location, role type, salary expectation, work mode) should be filled in and current. These are the filters the AI applies before it even begins matching your skills. If your location says Jaipur and the recruiter is searching for Bangalore candidates, you don’t appear, even if you’re willing to relocate but forgot to update that setting.
And when the AI Calling Agent calls you, pick up. Answer the screening questions clearly. The call takes 2 to 3 minutes. Your responses get logged. Your candidacy moves forward. Candidates who miss the AI call or don’t call back within the follow-up window lose the timing advantage to someone who picked up on the first ring. The AI reached out because you matched the role. Not answering is declining the match.
Fun Fact: During the Apna Safety pilot in August and September 2025, over 1,46,000 recruiters were verified on the platform, and the AI Calling Agent’s 80%+ candidate connection rate was roughly 2.5 times the industry average for recruiter-initiated phone screening.
The uncomfortable truth about AI in hiring is that it doesn’t make the process fair. It makes it faster. A candidate with a badly formatted resume and vague skills descriptions still gets ranked low. A candidate with a great profile who forgot to update their location preference still gets filtered out. The AI accelerates the process. It doesn’t fix the inputs.
Which is why the advice for candidates hasn’t changed even though the technology behind hiring has transformed. Write specific resume lines with numbers. List tool names instead of adjectives. Keep your profile updated. Respond quickly. These habits mattered when a human was manually reviewing your application. They matter more now that an algorithm is doing the first pass, because an algorithm is even less forgiving than a tired recruiter at 4 PM. A recruiter might read between the lines. An algorithm won’t.
Apna’s AI shortlisting system processed 2.7 crore applications in a single quarter. That scale means your profile is being evaluated, ranked, and either surfaced or buried every time a recruiter posts a matching role. The system runs whether you’re actively applying or not. Whether you check the app today or not. Your profile is either ready for that moment or it isn’t.
The recruiters who use this system aren’t outsourcing their judgment to a machine. They’re outsourcing the 10 hours of manual scanning that used to prevent them from using their judgment on anything more than the first 30 applications. The AI gives them back the time to actually read profiles. To notice things a keyword filter would miss. To make the kind of nuanced decision that only a human can make. But it can only surface the right candidates if those candidates gave it the right signals.
The AI is looking. Whether you’re ready to be found is up to you.
FAQs
How does AI candidate shortlisting software work on Apna? When a recruiter posts a role, Apna’s AI reads every incoming application, extracts structured data (skills, experience, location, role history), and ranks candidates by how closely they match the job requirements. The matching goes beyond keyword comparison. It uses patterns learned from millions of successful hires on the platform to understand that “client servicing” and “customer support” are functionally equivalent, for example. The recruiter sees a pre-ranked shortlist instead of scrolling through hundreds of unsorted applications.
What is the AI Calling Agent? An automated voice screening tool that calls matched candidates the moment a job is posted. It asks screening questions in English, Hindi, or regional languages, evaluates responses through voice analytics, and auto-shortlists candidates who meet the criteria. It can conduct over 10,000 simultaneous interviews with an 80%+ connection rate. Recruiters get a screened shortlist by the next morning without making a single phone call themselves.
Does AI replace the recruiter’s decision? No. The AI handles the sorting and initial screening. The recruiter makes the final shortlisting and hiring decisions. The system surfaces the most relevant candidates faster. The recruiter still reads profiles, conducts interviews, and uses judgment that no algorithm can replicate. AI removes the bottleneck. It doesn’t remove the human.
How can candidates improve their chances with AI-driven shortlisting? Make your profile structurally readable. Specific headline naming the target role. Tool-based skills instead of adjectives. Updated location and work preferences. Quantified experience bullets. And when the AI Calling Agent calls, answer it. The call takes 2 to 3 minutes and moves your candidacy forward immediately. Candidates who don’t answer lose the timing advantage to someone who did.
Is this only for entry-level roles? The AI matching works across role levels. But the AI Calling Agent is most impactful for high-volume hiring: customer support, logistics, retail, BPO, warehouse operations. These are roles where the candidate pool is large, the screening criteria are definable, and speed of hiring matters most. For senior or specialised roles with smaller applicant pools, the AI matching still ranks candidates but the AI Calling Agent is less commonly used since those roles require more nuanced human evaluation.
All the Best!
