Why 40% of New Hires Leave Within 90 Days, and How AI Fixes Onboarding

Introduction

Hiring is expensive. Losing a new hire within three months is devastating.

Yet, across Indian enterprises, this is exactly what keeps happening. Studies by SHRM India and LinkedIn's 2025 Workforce Report place early attrition in Indian companies between 30% and 40%. In IT services, BPOs, and fast-scaling startups, the number often climbs higher. The cost of each failed hire, factoring in recruitment, training, lost productivity, and team disruption, runs between 50% and 200% of the annual salary for that role.

The problem is rarely the talent. It is almost always the onboarding.

Most Indian organisations still treat onboarding as a two-day orientation programme followed by a PDF handbook and a wish for the best. In an era where employees expect personalised, continuous learning from day one, this approach is a retention risk hiding in plain sight.

AI is changing this. Not with flashy gimmicks, but with structured, scalable, private infrastructure that transforms how new hires learn, adapt, and grow.

---

The Problem

The first 90 days of employment are a make-or-break window. Research from Gallup shows that only 12% of employees globally believe their organisation does onboarding well. In India, the picture is bleaker. A 2024 survey by PeopleStrong found that 62% of Indian employees felt "abandoned" after their first week at a new job.

Here is what broken onboarding typically looks like in Indian enterprises:

• Generic content. Every new hire, regardless of role, department, or experience level, receives the same induction material.
• No feedback loops. L&D teams have no visibility into whether new hires actually understood the training.
• Language and communication gaps. In a multilingual workforce, English-only training materials leave large sections of the team behind.
• Overloaded managers. Direct supervisors are expected to fill the gaps, but rarely have the bandwidth.
• Zero personalisation. There is no mechanism to adapt the learning path based on an individual's pace, strengths, or skill gaps.

The result is predictable. New hires feel disconnected, underprepared, and unsupported. They leave. And the cycle repeats.

---

What AI-Powered Onboarding Actually Looks Like

This is not about replacing HR teams with chatbots. It is about giving L&D professionals the infrastructure to deliver onboarding that is personalised, measurable, and consistent, without multiplying headcount or budget.

AI-powered onboarding systems built on private LLM infrastructure can do what no static training programme can: adapt in real time to the learner.

A properly designed AI onboarding platform can assess a new hire's baseline skills on day one, generate a customised learning path, deliver training in the employee's preferred language, evaluate comprehension through interactive assessments, and flag at-risk hires to HR before they disengage.

All of this runs on-premises, with no data leaving the organisation's servers. For sectors like banking, healthcare, and government, where data sovereignty is non-negotiable, this is the only viable approach to AI in education and training.

---

How It Works

Step 1: Skill Assessment at Entry.

The system evaluates each new hire's current capabilities through adaptive assessments. This is not a static quiz. The AI adjusts difficulty and focus areas based on responses, building a detailed skill profile within minutes.

Step 2: Personalised Learning Paths.

Based on the assessment, the LLM generates a tailored onboarding journey. A software engineer joining a fintech firm receives different modules than a relationship manager at the same company. The content adapts to role, department, seniority, and even regional context.

Step 3: Communication and Soft Skills Training.

This is where tools like VaakShakti come in. One of the most underestimated reasons for early attrition is communication anxiety, particularly in client-facing and cross-functional roles. VaakShakti provides AI-powered voice tutoring that helps employees build confidence in professional communication, practise real scenarios, and receive instant, private feedback. No judgement, no audience, just structured improvement.

Step 4: Continuous Assessment and Manager Alerts.

The system tracks engagement, completion rates, and comprehension scores in real time. If a new hire is falling behind or showing signs of disengagement, the platform alerts the assigned manager or HR partner with specific, actionable insights.

Step 5: Workflow Automation for L&D Teams.

LLM workflow automation handles the repetitive operational load: scheduling check-ins, generating progress reports, sending reminders, and compiling feedback. L&D teams spend less time on administration and more time on strategy.

---

Real-World Example

Consider a mid-size Indian IT services company onboarding 200 freshers every quarter. Under the traditional model, L&D runs a week-long classroom programme, distributes manuals, and assigns mentors. By day 60, roughly 35% of the batch has either resigned or been flagged as underperforming.

With an AI-powered onboarding system running on local LLM deployment, the same company can deliver individualised training to all 200 hires simultaneously. Each fresher gets a learning path calibrated to their aptitude. Communication coaching through VaakShakti helps those who struggle with client interactions. Managers receive weekly dashboards instead of anecdotal updates. The L&D team, instead of running the same slides for the fourth quarter in a row, focuses on improving content and outcomes.

The shift is not theoretical. It is operational, measurable, and sustainable.

---

Why This Matters in India

India's workforce is young, growing, and increasingly mobile. The National Skill Development Corporation estimates that over 10 million people enter the workforce annually. Companies are competing fiercely for talent, and yet the systems meant to retain that talent remain largely manual and outdated.

There is also a uniquely Indian dimension to this challenge. India's linguistic diversity means that a training programme designed in English may not land effectively in Lucknow, Coimbatore, or Guwahati. AI infrastructure for education and corporate training that supports multilingual delivery is not a luxury. It is a necessity.

Private LLM infrastructure makes this possible without depending on third-party cloud APIs that introduce latency, cost unpredictability, and data exposure. For Indian enterprises navigating the Digital Personal Data Protection Act, keeping training data on-premises is both a compliance advantage and a trust signal.

---

Common Myths

Myth: AI onboarding is only for large enterprises with big budgets.

Reality: Modular, on-premises AI systems can be deployed incrementally. A company can start with communication training alone and expand from there.

Myth: AI will replace L&D professionals.

Reality: AI handles the repetitive and data-heavy work. L&D professionals focus on what they do best: designing learning experiences and mentoring.

Myth: Employees will resist AI-driven training.

Reality: When training is personalised, available on demand, and free from the anxiety of being judged in a group setting, adoption rates consistently exceed expectations.

Myth: Cloud-based AI tools are good enough.

Reality: For any organisation handling sensitive employee data, local LLM deployment provides the control, speed, and compliance that cloud solutions cannot guarantee.

---

Conclusion

The 90-day attrition problem is not a hiring problem. It is an onboarding problem. And onboarding, at its core, is a learning design challenge, one that scales poorly with manual methods and thrives with intelligent automation.

Setidure Technologies builds the AI infrastructure that makes this transformation possible. From private LLM platforms that personalise learning at scale to VaakShakti, our AI-powered communication training tool, we help Indian enterprises keep the talent they worked hard to hire.

If your organisation is losing new hires faster than it can train them, it is time for a different approach.

Write to us at admin@setidure.com to explore how Setidure's AI tutoring and onboarding infrastructure works.