Nasscom Community

From Chatbots to Counsellors: Evaluating AI-Driven Mental Health Tools for India’s Tech Workforce

8 Mins read
254
0
Share

With increasing prevalence of mental illness and a growing understanding of the direct impact of poor mental health on productivity and retention, corporates are struggling to respond to its speed and scale. Every second person at the office reports depression, anxiety, or burnout due to work-related situations, and behind each of those individuals is an organisation that largely has no formal system to respond.
India’s tech workforce alone stands at over 5.4 million professionals, yet only 0.09% of India’s 1.1 million registered companies have implemented any formal Employee Assistance Programme (Rajasulochana et al., 2025). Traditional mental health models, while effective at an individual level, were never designed to scale across an organisation of 10,000 people, let alone an industry of 5.4 million.
The appeal of AI-driven mental health tools lies precisely in this gap. They promise scale, accessibility, and round-the-clock availability at a fraction of the cost of traditional interventions.
It is against this backdrop that several Indian tech organisations have turned to them as one potential response. What follows is an objective look at what research shows, what these tools can and cannot do, and what responsible implementation actually requires.

What “AI in Mental Health” Actually Means

“AI mental health tools” is not one thing. It covers a range of applications; each doing something different, backed by different evidence, and carrying different levels of risk. Before buying anything, corporate leaders need to understand what they are actually looking at.

  • Screening and Early Detection These tools use machine learning to spot employees who may be struggling. By analyzing questionnaire responses, mood patterns, or anonymized communication data, they help to identify early signs of distress. For HR teams, this means early warning signals by team or department, so you can address structural problems rather than wait for a crisis.
  • Conversational Agents The most visible category. These are apps that guide employees through CBT exercises, breathing techniques, mood journalling, and mental health education — all through a chat interface. The more credible platforms available in the Indian market operate within clinically defined boundaries. These are not open-ended chatbots. They are structured tools with a defined scope — and their value is entirely contingent on clinical design rigour. Poorly scoped tools in this category carry meaningful risk of harm.
  • Intelligent Triage and Routing The practical value here lies in reducing administrative burden on human professionals, so that trained counsellors spend more time delivering care and less time on logistics.
  • Progress Monitoring AI that tracks mood week on week, flags when someone has stopped engaging with their exercises, and prompts follow-ups. This has some utility as a post-care support layer, though its effectiveness depends heavily on whether employees continue to engage with the tool at all.
  • Psychoeducation at Scale The simplest layer, and the most underused. AI-delivered content that teaches employees how to recognise burnout, manage difficult conversations, or protect their sleep under deadline pressure delivered to everyone, regardless of role or location.

These tools are low-cost and easy to access, which means more employees are likely to try them. But its adoption and sustained engagement in workplace settings still require active effort.  The real organisational value, however, lies upstream. Whether delivered through digital tools, structured workplace training, or peer education, an awareness-first approach is how organisations build population-level mental health literacy before a crisis emerges.
Understanding these distinctions matters because the risk profile is completely different across each layer. Psychoeducation carries near zero risk. Autonomous therapeutic conversation carries real clinical responsibility. Procurement decisions that treat these as equivalent are how organisations end up with tools that are very unsafe.
The distinction matters: awareness and first aid are not therapy, and neither is a psychoeducation chatbot. All three are, at best, early points of contact in a longer period of care.

What Indian Research Actually Shows

The evidence base for AI mental health tools in India is growing. Three findings are directly relevant to anyone making deployment decisions.

  • People are more open to it than you’d expect It is important to read this finding in context first: the following numbers reflect, in large part, the absence of accessible alternatives — not necessarily a preference for AI over human care. A peer-reviewed survey conducted under formal ethics clearance, found that 61% of Indian users with anxiety were open to AI-assisted support, and 49% felt AI actively reduces the stigma of seeking help (Varghese et al., 2024).  That is a finding about the access gap, not a validation of the technology itself.
    Employees who would never raise a concern with a manager will often engage readily with a lower-stakes first touchpoint — digital or otherwise — if confidentiality is credibly guaranteed.
  • Users want AI and human care together, not instead of each other The same study found that the hybrid model — AI combined with access to a human professional was seen as the most compelling option by Indian respondents (Varghese et al., 2024).
    The same study also found that participants expressed greater comfort with a human counsellor than with an AI platform, and that privacy concerns were meaningfully higher for AI tools — particularly among women (Varghese et al., 2024). People turn to AI because human care is hard to access, not because they prefer a chatbot. For corporate deployment, this means one thing: AI should be the entry point into care, not the endpoint.
  • Engagement numbers from real deployments are mixed Published service evaluations of AI mental health tools deployed in healthcare settings show encouraging engagement rates — with 80.1% of onboarded users completing a minimum of two sessions, averaging 10.9 sessions over approximately four weeks ((Chang et al., 2024). It is worth noting, however, that session completion is a measure of engagement, not clinical outcome. Whether sustained engagement translates into measurable improvements in employee wellbeing requires more robust longitudinal evidence than currently exists in the Indian context.
  • Where Indian deployments have struggled.  The optimism above must be read alongside evidence from the same context. A qualitative study involving urban Indian psychiatrists, patients, technology experts, and health-tech CEOs, found that almost all respondents cited ethical, legal, and accountability gaps as barriers — with Indian psychiatrists specifically flagging the absence of clinical validation and the risk of liatrogenic harm. The study’s conclusion was direct: deployment of AI-enabled mental health platforms in India is not feasible without first addressing these structural gaps (Thenral & Annamalai, 2021).

Separately, a 2025 mixed-methods study of telepsychiatry users at a tertiary care teaching hospital in north India found a 56.4% dropout rate once in-person care became available again — with patients citing concerns about therapeutic continuity, technological difficulty, and lack of privacy (Shakya et al., 2025). High sign-up rates and session completion figures, then, do not guarantee that digital tools hold up against the alternative when one exists.

A Practical Deployment Framework

The most evidence-supported architecture for workplace mental health is stepped care matching support intensity to need level, rather than routing everyone through the same door. It is how India’s own National Mental Health Programme is designed, and it is directly applicable to enterprise contexts.

Level 

Who It Serves 

AI’s Role 

Human Involvement 

Universal 

All employees 

Psychoeducation, mood tools, digital wellness 

Not required 

Selective 

Early stress signals 

Guided CBT, conversational support, 24/7 access 

Trained peer supporter or MHFA trained manager  

Indicated 

Mild-to-moderate distress 

Smart intake, counsellor matching, session prep 

Access to care counsellor delivers care 

Clinical 

Moderate-to-severe 

Risk detection, emergency escalation 

Psychiatrist / crisis services 

This stepped architecture only functions if each level has a human capable of responding when the system flags a concern. When an employee disengages from a digital tool, stops responding to check-ins, or begins showing signs of distress that an algorithm can flag but not interpret — that is precisely where trained peer supporters have a functional role. They are not therapists. They are the human bridge between a system-generated alert and an appropriate, compassionate response.
The majority of your workforce stressed but not clinically unwell receives meaningful, evidence-informed support at minimal cost. Your limited human counsellors focus on employees who genuinely need them. And the AI layer generates anonymised, population-level data that tells your HR team which projects, teams, or periods are producing the highest distress enabling systemic fixes, not just individual interventions.

What to Look for When Evaluating a Platform

  1. Demand peer-reviewed clinical evidence. Ask for published efficacy studies, not case studies or testimonials. A 2025 meta-analysis of 31 randomised controlled trials covering nearly 30,000 participants found that AI chatbots demonstrated statistically significant, if modest, reductions in depression, anxiety, and stress symptoms (Feng et al., 2025) . That kind of evidence base — peer-reviewed, large-sample, RCT-grounded — should be a baseline expectation for any platform you consider.
  2. Verify DPDP compliance and privacy architecture. Under India’s Digital Personal Data Protection Act 2023 and DPDP Rules 2025 (Ministry of Electronics and Information Technology, 2023), mental health data is classified as sensitive personal data requiring explicit, layered consent. More practically: your employees will not use a tool they believe their manager can access. Individual conversation data must be completely inaccessible to employers to make this a contractual obligation, not a vendor assurance.
  3. Confirm crisis protocols. Every AI mental-health platform must have a defined, human-led protocol for high-risk disclosures. What happens when a user expresses suicidal ideation? The answer must be: immediate routing to a trained human, with local emergency resources provided. If a vendor cannot clearly articulate this process, do not deploy their tool.
  4. Insist on linguistic coverage. A platform available only in English is inaccessible to a meaningful portion of India’s tech workforce, particularly across development centres in smaller cities. Hindi, Tamil, Telugu, Kannada, and Marathi coverage should be expected for any serious pan-India rollout.
  5. Measure outcomes, not activity. Session count is not a health outcome. Ask vendors for data on validated clinical measures changes in PHQ-9 or GAD-7 scores across user cohorts over 90 days. If they cannot provide this, their wellbeing claims are unverifiable.

The Leadership Variable

No platform reaches the employees who need it most without visible leadership commitment. India’s Corporate Health Study 2026 — published by Truworth Wellness in November 2025 and drawing on insights from over 300 organisations — found that while the majority of organisations have moved towards structured, policy-backed wellbeing frameworks, execution continues to lag significantly behind governance intent (Truworth Wellness, 2025). The gap between policy and reality is almost always a culture problem, not a technology problem.
Utilisation of wellbeing services, digital or otherwise, consistently rises in organisations where this norm is set from the top. Psychological safety is created by senior behaviour, not app subscriptions.
For organisations still evaluating their approach, the leadership question is not which platform to choose — it is whether the culture exists to make any tool usable. Reference wellbeing initiatives in all-hands communications. Share anonymised aggregate insights. Let mental health support be visible, not a shameful last resort.

The Bottom Line

AI-driven tools can play a legitimate supporting role in an organisation’s mental health infrastructure. The more useful question for any CHRO or L&D leader is not whether to adopt a digital tool, but what gap it is genuinely addressing, what evidence supports that claim, and what human infrastructure must already be in place for that tool to function responsibly. Technology can reduce the distance between an employee and a resource.
But it is trained, present, accountable human beings who determine whether that resource actually reaches the people who need it most.

Sources:

This article is the intellectual property of MHFA India Private Limited. Any reproduction, distribution, or use without prior written permission is strictly prohibited.

254
0
Share