PaperlayerView exact revision on arXiv

Source-linked AI summary

From AI-Free Gates to AI-First Work: How Engineering Students in India Navigate University Education and Campus Placements Without AI Guidelines

Samhitha Goutham, Ziwei Gu

arXiv:2610.29140v1cs.HCcs.AI

TL;DR

Research has offered limited evidence about how Indian engineering students navigate campus placements that largely exclude AI on the way to AI-first work. The authors interviewed 23 students and recent graduates, finding that they used AI to clear gates and prepare for work while worrying about readiness and what grades and screen scores certify.

  • Problem

    Research on AI-mediated transitions into technical work largely draws on Western and East Asian settings, and existing Indian engineering-student studies did not examine placements.

    Source: §2.4 · p. 5

  • Method

    The authors interviewed 23 current students and recent graduates of Indian engineering and computing programmes about AI across coursework, examinations and placements.

    Source: §1 · p. 2, §3.1 · p. 5

  • Results

    Participants described AI-excluding education and placement gates before AI-first work, using AI to clear assessments and prepare while worrying that dependence could leave them unready.

    Source: §1 · p. 2, §5 · p. 15

  • Takeaways & Limitations

    The authors argue that AI-assisted clearing weakens what grades and screen scores signal, while leaving students to carry readiness concerns without a shared account of readiness.

    Source: §5 · p. 15

    The sample excludes less-resourced, rural and regional-language-medium institutions, includes only two women, and relies on self-reports and students’ perceptions of other stakeholders.

    Source: §6 · p. 19

Abstract

from arXiv · show

Engineering graduates in India enter a technology sector that is being reshaped by generative AI, yet the route into that sector runs through examinations and campus placement screens that largely exclude AI. Institutions in India have an approved intake of nearly 2.9 million seats in engineering and technology for 2026-2027, but HCI research on how students prepare for AI-mediated work has drawn mostly on Western and East Asian settings. We interviewed 23 students and recent graduates of Indian engineering and computing programmes about how they used generative AI across coursework, examinations and placements, and how they judged their own readiness for work. Using reflexive thematic analysis, we report five themes, covering AI-first workplaces and AI-free placement screens, how participants prepared for placement, how they used AI in coursework, the role of faculty authority and enforcement, and participants' concerns about their readiness for work. Drawing on signaling theory, we discuss what grades and screen scores certify once AI can help produce them, and we offer recommendations for employers, institutions and AI tool designers.

1 Introduction

Indian engineering students pass through examinations and placement screens that largely exclude AI on the way to a technology sector being reshaped by generative AI. This study examines how students navigate that transition and what their experiences imply for the signals employers use.

1 Introduction

  • Students were left to decide how to use AI for coursework, placement preparation and workplace readiness without institutional frameworks for its use.Participants distinguished clearing assessments from learning what those assessments were intended to teach.Source: §1 · p. 1
  • Research on AI-mediated transitions into technical work has mostly drawn on Western and East Asian settings, while prior Indian studies did not examine campus placements.Campus placements organize much of the later years of Indian engineering degrees and shape where many graduates begin working.Source: §1 · p. 2
  • The study uses interviews with 23 current students and recent graduates of Indian engineering and computing programmes to examine AI use across their education-to-work transition.The authors use signaling theory to ask what grades and screen scores certify once AI can help produce them.Source: §1 · p. 2
  • Participants described AI-excluding placement screens alongside workplaces that expected AI, and worried that dependence on AI-assisted work could leave them unready.They also described preparing for placement mainly through self-directed study with AI as their coach.Source: §1 · p. 2
  • The paper contributes an interview-based account of this transition, a signaling-theory interpretation of AI-assisted assessment, and implications for employers, institutions and tool designers.It also identifies the responsibilities students take on when institutions and employers do not define readiness for AI-first work.Source: §1 · p. 2

2 Related Work

Prior research examines generative AI’s effects on technical work, student use and assessment, but leaves limited evidence about Indian students’ passage through campus placements into AI-first work. The paper situates that gap in research on education, hiring, institutional practices and student reliance.

2 Related Work

  • The related-work review combines research on technical-work transitions, Indian and Global South student AI use, and assessment, enforcement and reliance.It closes by identifying the gap the study addresses.Source: §2 · p. 2, §2.4 · p. 5

2.1 From Education to AI-First Work

  • Research on education-to-work transitions describes AI redirecting junior engineering work into senior-AI workflows [25], while hiring research examines automated screens and campus placements.Signaling theory frames CGPA cut-offs and standardized placement tests as employer-facing signals under uncertainty.Source: §2.1 · pp. 3–4

2.2 Students and GenAI in India and the Global South

  • Indian student studies document uses and concerns about GenAI, while Global South education research highlights access, language and pedagogical context [14].The paper follows work emphasizing institutional context and learners’ existing strategies rather than treating Indian culture as essentialized.Source: §2.2 · p. 4

2.3 Assessment, Enforcement and Reliance

  • Assessment research distinguishes rule-based changes from task redesign, and studies examine concealment, selective AI use and AI-mediated help-seeking.Related work also raises concerns about unreliable detection, student reliance and novices’ ability to notice AI errors.Source: §2.3 · pp. 4–5

2.4 Research Gap

  • Prior work mostly studies Western or Korean settings, permitted or taught AI use, or Indian student use soon after ChatGPT’s release, without examining placements [14].This study addresses the gap with interviews about an exam-oriented system, placement screens that largely exclude AI and workplaces that expect it.Source: §2.4 · p. 5

3 Methods

The study uses semi-structured interviews with 23 current students and recent graduates, recruited purposively and through referrals, and analyzes their accounts through reflexive thematic analysis. The methods also describe translation and positionality practices, alongside participant protections.

3.1 Participants and Recruitment

  • The sample comprised 23 students and recent graduates of Indian engineering and computing programmes, recruited purposively and through referrals.Most participants studied in Chennai and all came from established institutions; the authors avoid institution-level claims resting on one or two accounts.Source: §3.1 · p. 5

3.2 Data Collection

  • The first author interviewed participants in English or Tamil between September and October 2026 using a seven-section guide covering coursework, placements, learning, institutional norms and career concerns.Most interviews lasted 30–45 minutes; Tamil–English interviews were about 20 minutes.Source: §3.2 · pp. 5–6

3.3 Transcription, Translation and Analysis

  • The first author corrected clear transcription errors, translated Tamil–English segments, and used inductive two-level coding within six phases of reflexive thematic analysis.Themes were treated as interpretive constructs rather than frequency counts, and participant IDs indicate who described each pattern.Source: §3.3 · p. 6

3.4 Positionality

  • The interviewer’s insider experience informed the study but carried a risk of assumptions shaped by their own trajectory.Reflexive memos, consultation with a second author uninvolved in data collection, and grounding claims in participants’ words were used to limit that risk.Source: §3.4 · p. 6

3.5 Ethics

  • The study received institutional review board approval, obtained informed consent from adults, and restricted recording and transcript access to the two authors.Institutions were described by type and professors’ names were removed from transcripts.Source: §3.5 · p. 6

4 Findings

The findings trace students’ experiences of AI across coursework, placements, and interactions with faculty, including how they prepared for work. They describe learning through peers, trial and error, and internships amid limited institutional support, alongside reported dependence, diminished confidence, unevenly distributed skills, and perceived benefits.

4.1 AI-Free Gates, AI-First Work

  • Participants contrasted AI-first workplaces with placement screens that excluded AI and often tested skills they saw as unlike the job.Workplace norms shifted toward AI during many participants’ degrees, while screens were widely seen as filters for general problem-solving; some doubted online tests were meaningful when candidates could use AI to cheat. Participants welcomed assessments that allowed AI or involved building applications.Source: §4.1 · pp. 7–8

4.2 Preparing for the Gate with AI

  • Almost every participant used AI to prepare for placements, often as a tailored coach because coursework and faculty offered little preparation for the screens.They used it to plan study, rehearse interviews, target companies and interviewers, and build practice tools; short notice made its help especially valuable. Access to stronger paid models was limited by cost, and some participants described added difficulty using AI in English.Source: §4.2 · pp. 8–9

4.3 Clearing, Not Learning

  • Participants often used AI to clear coursework assessments, treating marks as valuable and understanding as task-dependent in a system they saw as emphasizing recall.AI helped generate exam-ready answers and target likely questions, while assignments were triaged by perceived value, deadlines, and whether understanding would be checked. Participants described scoring and learning as separate tracks, with grades tied to opportunity and AI helping produce some of the grades placement screens filtered on.Source: §4.2 · p. 9, §4.3 · pp. 10–11
  • AI also served as a tutor when teaching assumed prior knowledge, supporting learning beyond the assessed curriculum as well as preparation for placements and work.Participants used it for basic explanations, repeated questions, practice, and curiosity-driven learning, though some admitted copying solutions without thinking them through. They were more likely to do work themselves when AI was unreliable or when they expected to explain their work in person.Source: §4.3 · pp. 10–12

4.4 Answering Upward, Sharing Sideways

  • Faculty authority shaped AI use through grading and uneven, often unstated rules, while institutional responses relied on detection and limits rather than guidance.Many participants deferred to professors over AI, sometimes adapting answers to the expected format; acceptance of AI rarely changed teaching or assessment. Students described bans, silence, or permission alongside detection tools, online tests open to AI, and efforts to evade checkers.Source: §4.4 · pp. 12–13
  • Participants proposed changing assessment rather than relying on stricter detection, including application-oriented questions and separate tests of unaided knowledge and AI use.P9 suggested pen-and-paper exams for checking what students know and online assessments for testing how well they apply AI.Source: §4.4 · p. 13
  • Peers openly shared prompts, outputs, and tool knowledge, but this informal learning spread unevenly; eight participants wanted faculty guidance and five saw little value in it.Professors often determined whether AI was permitted without saying so, while participants said no one taught them what to learn for work.Source: §4.4 · pp. 13–14

4.5 Ready for What?

  • Many participants said AI had become their first response to tasks and felt it was weakening their ability to think, learn, or perform basic skills.Others disagreed: P10 saw delegation as raising the level of work, while P9 said AI-assisted debugging made them more knowledgeable.Source: §4.5 · p. 14
  • Participants linked AI-assisted work to doubts about ownership and job readiness, including concern that they might not answer questions about it or meet company requirements.Some described underconfidence presenting AI-assisted projects in placement interviews; P14 feared being unable to work without AI.Source: §4.5 · pp. 14–15
  • Participants learned AI literacy through trial and error, friends, and internships rather than institutions, and their ability to use AI well was uneven.They developed rules of thumb for trust, but participants with greater experience or coding skills described using AI more effectively.Source: §4.5 · p. 15
  • Participants assembled readiness for AI-first work themselves, often with misgivings, and said their institutions left gaps in managing reliance on AI.They learned to prompt and catch some errors, but this literacy did not extend to managing their own reliance.Source: §4.5 · pp. 14–15

5 Discussion

The discussion argues that AI-free assessments and placement screens sit between education and workplaces that expect AI, weakening what scores certify and leaving readiness largely to students to assemble. It connects this mismatch to uneven preparation and proposes changes for employers, educators, and tool designers.

5.1 When Clearing Produces the Signal

  • AI-assisted clearing weakens the grades and screen scores employers use as signals, while participants also described losing confidence in whether they understood or owned their work.The discussion argues that rehearsable answers can be forecast and compressed with AI, and that signals protected only at the moment of measurement can still be produced elsewhere.Source: §5.1 · pp. 15–16
  • Harder interviews may raise preparation costs without restoring a link to the job; affordability-related effects on selection in India remain a hypothesis.Participants reported free tiers falling short and few students being able to pay for better models, but the authors state their data cannot test this in India.Source: §5.1 · p. 16

5.2 AI-Free Gates for AI-First Work

  • AI-free gates assess unaided skills even as workplaces expect AI use, leaving students without a shared definition of readiness or guidance on what to learn instead.Participants also identified accounting for, explaining, and standing behind AI-assisted work as a skill neither gate assessed.Source: §5 · p. 15, §5.2 · pp. 16–17
  • The mismatch has particular weight in India, where routine work passed to Indian IT companies is the kind AI absorbs first, and participants’ concerns persisted among students who entered college with AI already present.The discussion argues that changing assessments without incorporating AI, or incorporating AI without reforming assessments, would each leave a mismatch.Source: §5.2 · p. 17

5.3 Readiness as an Individual Burden

  • Students assembled readiness through sophisticated but uneven self-directed preparation, while unstated faculty rules encouraged concealment and peers shared AI knowledge openly.Participants who navigated most effectively tended to be motivated, confident with AI, and sometimes able to pay for better tools; students who wanted guidance more often described dependence or access barriers.Source: §5.3 · p. 17
  • Because institutions and employers leave readiness undefined, the discussion argues that readiness should not be left to students alone.The authors note that their data cannot separate factors associated with who arrives best equipped and offer them as hypotheses.Source: §5.3 · p. 18

5.4 Implications

  • The authors recommend AI-aware placement screens and clearer skill expectations, explicit course-level AI rules and reasoning-focused assessments, and tools that support answer ownership.Suggested tool designs include asking students to commit to an answer before seeing the model’s and making AI-work provenance visible to students.Source: §5.4 · p. 18

6 Limitations and Future Work

The authors delimit the findings by the sample, retrospective and self-reported accounts, translation, and lack of interviews with institutional or employer stakeholders. They propose recruiter interviews, longitudinal research, broader surveys, and intervention studies to examine the patterns and proposed designs.

6 Limitations and Future Work

  • The sample excludes less-resourced, rural, and regional-language-medium institutions, includes only two women, and relies on retrospective, self-reported accounts and translated interviews.Faculty, administrators, recruiters, and managers were not interviewed, so their positions appear only through students’ perceptions; tools, offers, and policies were changing rapidly.Source: §6 · p. 19
  • Future work could interview recruiters, follow students who entered college with AI present, survey broader institution and regional contexts, and test proposed assessment and placement designs.The proposed intervention examples include AI-permitted build tasks paired with oral explanation.Source: §6 · p. 19

7 Conclusion

The conclusion describes AI-free gates leading to AI-first jobs, with students using AI across assessment and preparation largely on their own. It argues that employers and institutions share the task of defining and directly assessing readiness for AI-first work.

7 Conclusion

  • Interviews with 23 Indian engineering and computing students and recent graduates traced AI use from coursework through placement screens into work.Participants described gates built for a world without AI and jobs that expected AI from the first day.Source: §7 · p. 19
  • Participants used AI to clear gates, prepare for them, and learn where teaching fell short, often on their own initiative and with access shaped by affordability.Professors’ stances on AI were seldom explicit, and tool quality depended on what students could afford.Source: §7 · p. 19
  • The conclusion argues that AI-assisted clearing weakens what grades and screen scores certify, making readiness definition and direct assessment a shared task for employers and institutions.Participants were already attempting to assemble readiness themselves.Source: §7 · p. 19

Disclosure of AI Use

The authors report using Claude as a writing and organisation aid while retaining analytic decisions and responsibility for the paper. They reviewed, revised, and verified the AI-assisted material.

Disclosure of AI Use

  • The authors made the analytic decisions, including coding, themes, and interpretations.They also conducted the interviews and literature review.Source: Disclosure of AI Use · p. 19
  • Claude helped collate excerpts, draft prose from author notes and outlines, edit text, check quotations against transcripts, and format the paper.Source: Disclosure of AI Use · p. 19
  • The authors reviewed, revised, and verified AI-assisted material and take responsibility for the paper’s content.Source: Disclosure of AI Use · p. 19

A Participants

Participant institutions are described by type, while gender is omitted to protect anonymity; programmes and years follow participants’ descriptions.

A Participants

Loading 2610.29140v1…