AI-First Screening: What Your Students Now Have to Show

By the time a recruiter at one of India’s large IT firms opens a resume in a campus drive, the automated system has already scored it and ranked the candidate against everyone else in the pool. Reporting by the Economic Times in June 2025 documented how this plays out at HCLTech, Wipro, and Infosys: resumes are processed and ranked before any human reads them, and the recruiter’s shortlist is what that first pass produces. A student who does not clear the algorithm may not reach a recruiter’s view at all.

I have watched this shift build across our partner network. Across more than 2,000 colleges and universities FACE Prep has worked with in 18 years, the clearest pattern in shortlisting results is this: the students who pass the algorithm are the ones who have something to show, not just something to claim. NASSCOM and Indeed, in their joint report released in May 2026, found that 86 percent of Indian employers have seen some impact of AI on job roles and responsibilities, with 35 percent reporting significant redefinition. The same report records that the share of Indian job postings mentioning AI has risen consistently since 2023. The screening layer before a recruiter’s desk has been built to match.

The question for a college placement cell is not whether AI screening is happening. It is whether your students are visible to the algorithm in the way that produces a shortlisting, and whether the students who do clear it are prepared for the interview that follows.

What runs before the shortlisting call

The Economic Times report named three people who made this concrete. Ramachandran Sundararajan, the Chief People Officer at HCLTech, described the company’s proprietary AI-backed platform: it enhances job descriptions, screens CVs, assists interview panels, and provides a second opinion on candidates. The platform had processed more than 45,000 open positions. Sanjeev Jain, COO at Wipro, said the company began AI pilots in April 2025 covering resume screening, first-level interviews, and communication assessments. Shaji Mathew, the Chief Human Resources Officer at Infosys, described AI tools handling shortlisting and chatbot-based interviews.

Across these companies and the sector more broadly, AI is now involved in somewhere between 70 and 80 percent of initial screenings at large IT firms, according to the same reporting. The scale matters because it means the automated stage is not a supplementary check. It is the primary sorting mechanism. A recruiter at a large firm does not scroll through two thousand resumes from a campus drive. They review the candidates the algorithm cleared and ranked above a threshold.

Most placement cells have limited visibility into this because the shift happened inside recruiter systems, not at the campus level. From a student’s perspective, the process looks the same: submit a resume, wait for a call. The invisible stage between those two steps is what has changed, and a college that has not mapped it is preparing students for a screen that no longer exists while the one that does runs without them.

What the algorithm scores, and what it passes over

The scoring logic of a modern AI resume tool is not a mystery, even if exact weights vary by system and company. The algorithm reads the resume against the job description and scores on several dimensions: how well the candidate’s stated skills match what the job requires, whether there is evidence that those skills have been applied in real work, whether the experience and project sections describe outcomes rather than just activities, and whether the profile is consistent and complete.

A student who has spent three years studying a subject and lists it as a skill has described knowledge, not evidence. The algorithm reads both, but in ranking a large pool it weights evidence more. A project entry that reads “built a recommendation system that improved search response time on a 50,000-row dataset” is evidence. “Proficient in machine learning” is an assertion. In a pool of five hundred resumes, the student with the project entry ranks above the one with only the assertion, and that gap is what shows up in shortlisting rates before any recruiter looks at a single name.

GitHub activity is a visible evidence signal for coding roles. A consistent commit history, or a repository that shows incremental work on a real problem, tells the algorithm something about a candidate’s coding habits that a resume entry cannot replicate. For a student aiming at a software or data role, the absence of any public coding activity is a gap that a strong CGPA does not fill.

The preparation implication is specific: train students to build and document real work in Year 2 and Year 3. A student who arrives at a drive with a project they can walk through is visible to the algorithm in a way that a student with a stronger GPA but no project evidence is not.

The interview that follows an AI shortlist

The interview a student reaches after clearing the algorithm has also changed, in a way that matters for students who optimised their resume without matching it to real experience.

A recruiter who relies on AI screening knows the filter can be worked. A resume loaded with keywords from the job description scores well on a text-matching algorithm whether or not the skills are real. That is a known limitation of the tool, and it is why the human interview after the screen has been redesigned in most companies to verify the claims. When a recruiter asks a student to walk through a project they listed, they are not making conversation. They are testing whether the resume is accurate.

Earlier in 2026, Infosys deferred hiring tests for more than 20,000 candidates after detecting impersonation and malpractice during online assessments, according to the Times of India. I mention this not to suggest impersonation is the dominant risk on campus, but because it reflects a wider pattern: when automated processes handle volume at scale, recruiters tighten every verification stage that follows. The human interview after an AI shortlist is one of those stages.

A student who passes the algorithm because their resume accurately represents real project work is in a straightforward position. The project they listed is the project they did. The recruiter’s verification questions confirm rather than expose. For a student whose resume claims outrun their experience, the outcome runs the other way, and the exposure is greater now than it was when the interview was designed around a different profile.

The practical implication for a placement cell is that the goal of preparation is to help students build the experience that makes the resume accurate, so the verification interview is a demonstration of real work rather than a stress test of unsupported claims.

Building the project evidence trail

The preparation shift is structural. It starts in Year 2, not in the final semester.

A student who begins building a project portfolio in the pre-final year arrives at a campus drive with twelve months of documented, defensible work. A student who is told to add a project to their resume in September of the final year is rushing to produce something in three or four weeks that an algorithm built to distinguish genuine applied experience from a last-minute addition will not rank highly.

Three things create an evidence trail that the algorithm reads. The first is a project that describes a real problem, a real approach, and a real result. It does not need to be original research. A student who implemented a text classifier on a public dataset and documented what they built, what the result was, and what they would do differently has produced more of the right signal than most of their cohort. The second is a coding activity record. Many college students only write code when an assignment requires it. A daily practice habit builds a visible history of sustained effort. DOJO, which we use with partner colleges for daily coding practice, produces this kind of consistent activity record alongside the skill, which is the combination that reads well in the screening stage. The third is outcome-focused language on the resume itself: not “worked on a machine learning project” but “built a sentiment classifier in Python, tested on 10,000 reviews, achieving 82 percent accuracy on a held-out set.”

The mapping work matters alongside the portfolio work. Not every company that visits campus uses AI screening, and the companies that do run different tools weighted differently. A placement cell that has spent fifteen minutes with each of its regular recruiters asking what the initial screening process looks like has a map. A college without that map is preparing students for a generalised version of the problem rather than the specific version each recruiter runs.

A college in Karnataka that changed what it measured in Year 2

A college we work with in Karnataka, around 1,600 students across a typical branch mix, ran a placement season two years ago that looked acceptable by its main metric. The overall placement count was reasonable, the services drives had delivered as expected, and the leadership was broadly satisfied.

What the placement head noticed, and could not immediately explain, was that a set of GCC and product-company drives had produced almost no shortlistings. Students who cleared the aptitude test were not progressing through the resume screening stage. The shortlisting rate for those drives was low enough to be visible, and no one had an account of why.

The diagnosis took a few weeks. The students were qualified. CGPA was at the bar. Aptitude scores held up. What was missing was project evidence. The resumes listed the technologies the students had studied and the courses they had completed. None of them carried a project that described what was built, what it produced, or what the result had been. For the companies running structured AI screening, those resumes ranked low in the pool, and the recruiter never reached them.

The fix was not a crash course before the next season. It was a change to what the college measured in the pre-final year. The placement cell introduced a portfolio checkpoint for all students registering for placement support: before the final year began, each student was expected to complete one documented project, on any topic they chose, with a short write-up covering the problem they worked on, the approach they took, and the result they got. Faculty advisors reviewed the write-ups not for engineering correctness but for specificity: could the student defend this in a conversation?

By the following season, most students heading into drives carried that one project, described in enough detail that the resume could carry it accurately. The shortlisting rate for GCC and product-company drives improved. The students who reached the interview stage could walk through the project because it was a project they had done. The conversion from shortlist to offer moved in the same direction.

What a placement cell can act on before the season opens

Three moves cover most of the practical work.

The first is to map the screening process for each recruiter on your list. Which companies route initial applications through AI scoring? What does the scoring weight: skills match, project evidence, coding activity, certifications? The best source for this is the recruiter themselves, asked directly before the season. Most will tell you, because they want better shortlists, not worse ones, and a college that understands the process is a college worth a longer conversation.

The second is to audit your students’ resumes before they go into any drive. Not for grammar or format, but for evidence. For each skill listed, is there a project, an outcome, a specific result behind it? If the answer is no, the student has an assertion without evidence, and the algorithm will treat it accordingly. This audit is more useful than any last-minute training, because it shows each student exactly which gap they need to close.

The third is to build the portfolio expectation into the pre-final year for next season’s batch. A placement cell that waits until the final year to introduce this requirement is always working against the clock. The students who arrive at a drive with a year of documented project work are in a structurally different position, and making that shift is a college-level decision, not a student-level one. A companion piece on building AI readiness into the preparation cycle covers how to sequence this alongside existing training.

---

The screening has automated, and that has not made the eventual hiring decision less human. An algorithm sorts the pool. A recruiter reviews the shortlist. An engineer on the panel asks a student to walk through what they built. That last stage has become more searching in 2026, not less, because companies know the earlier stages can be worked by a well-formatted resume. A college that trains students to build real work and to describe it clearly is preparing them for all three stages, not just the first one. The institutions that make this shift in the second and third year teach a habit that serves students in the drive, and then again in the role, which is the longer measure of what a placement programme is actually building.