When Classrooms Fit the System, Not the Student

In many Indian classrooms, the gap is not always visible.

A student who cannot focus, another who struggles to read at a pace, or someone who processes language differently—these are often interpreted as a lack of ability, effort, or discipline. Increasingly, however, they are being understood differently: as variations in how the brain works.

India is home to millions of neurodivergent individuals, yet the scale remains under-acknowledged. ​India’s neurodivergent population is far larger than official data suggests. According to studies, one in eight children may have a neurodevelopmental condition, with ADHD affecting up to 5-8% of school-age children and autism estimated at roughly 1 in 100—figures widely believed to be underreported due to gaps in diagnosis and awareness. 

Despite this, access to diagnosis, support, and inclusive education remains uneven, particularly beyond urban centres.

For as far as we know, neurodivergent individuals, those with ADHD, autism, dyslexia and other neurodevelopmental conditions, have navigated systems that expect uniformity. Diagnosis often comes late, support is inconsistent, and access remains deeply unequal, particularly outside urban India.

Now, a new layer is beginning to reshape that reality. AI is not just entering classrooms and clinics; it is beginning to change how learning, behaviour, and cognition themselves are understood.

From Assistive Tools to Adaptive Systems

India’s AI accessibility push reflects a broader shift—from tools that compensate for limitations to systems that adapt to individuals.

The government’s India AI Impact Summit 2026 Compendium frames AI as a mechanism for inclusion, emphasising its role in addressing learning disabilities and enabling early intervention.

One of the telling examples is AlphaKhoj, a NeuroAI-powered multilingual literacy platform designed for neurodiverse learners. It combines AI with cognitive insights to personalise how users engage with language, comprehension, and reading.

Designed for neurodiverse users, including those with dyslexia and other learning disabilities, it adapts literacy by adjusting reading formats, pace, and comprehension pathways.

In the Indian context, this becomes even more critical, where multilingual learning environments can significantly increase cognitive load. By combining multilingual support with adaptive learning, Alphakhoj reduces this burden, allowing users to engage with content in ways that align with their cognitive processing.

A critical layer in this shift is not just what AI does, but how people interact with it. Increasingly, systems are moving towards voice-first, conversational, and multimodal interfaces that combine text, audio, and visual inputs.

For neurodivergent users, this is significant. Individuals with ADHD may benefit from audio-led engagement that reduces attention strain, while those with dyslexia may find spoken content more accessible than text-heavy formats. Structured, predictable interactions can also support users on the autism spectrum.

This shift is also becoming visible in how people consume knowledge online. Armenian AI-based audio processing software company Krisp recently launched Accent Converter for YouTube, a free Chrome extension that improves the clarity of accented speech in real time, aiming to reduce friction in understanding global content. 

Tools like this are attempting to reduce the cognitive effort required to process spoken content in real time. Vimal Nair, Chief Growth Officer at Krisp, explains, “When we look at how users engage… they’re not optimising for speed, they’re optimising for understanding.”

For users with ADHD or auditory processing challenges, this distinction is critical. “If too much effort goes into decoding speech, it becomes fatiguing and distracting. Reducing that load helps users stay focused on the content, not the delivery,” Nair notes, framing voice AI also as “an accessibility layer.”

The shift here could be fundamental. Instead of asking individuals to adapt to rigid systems, these platforms could potentially begin to reshape themselves around individual cognitive patterns, whether that means adjusting reading formats, pacing, or language complexity.

However, Alphakhoj’s promise risks outpacing evidence. Adaptive learning claims lack long-term validation, especially across India’s diverse contexts. Multilingual AI may also misinterpret cognitive differences as deficits, and over-reliance on automation could sideline educators, reducing nuanced, human-led intervention in complex neurodevelopmental learning journeys.

The Problem of Time

For clinicians, the most pressing challenge in neurodevelopmental care is often not treatment, but delay.

“AI has the potential to significantly reduce the delay between symptom onset and identification,” Sandesh Cadabam, Executive Director at Cadabams Group, tells AIM.

In India, early signs of conditions like ADHD or autism are frequently missed or normalised, particularly in non-urban settings where specialist access is limited. As Cadabam notes, “in non-urban settings, early signs are routinely missed or normalised.”

By the time families reach clinical care, critical developmental windows may already be narrowing.

AI, Cadabam explains, introduces a different model of observation—one that is continuous rather than episodic.

“What is particularly promising is the shift from one-time clinical evaluation to continuous, observation-based screening,” he says.

This shift allows early indicators—attention variability, social responsiveness, communication patterns—to surface well before a family has navigated the path to a specialist.

Through mobile-based interfaces, school-integrated tools, and even messaging platforms, AI systems can track behavioural patterns, communication signals, and developmental markers over time. Attention variability, social responsiveness, and language patterns can surface earlier, often before a formal clinical pathway is even initiated.

But this is not a replacement for care. Cadabam is explicit about the limits. “These tools must be positioned as screening aids, not diagnostic systems,” he stresses.

The Risk of Simplifying Complexity

AI systems are capable of adapting content and interaction styles to suit individual needs, particularly important for neurodivergent individuals whose cognitive processing may differ from standard models.

But this capability comes with its own risk.

“Neurodevelopmental conditions are highly individual in their presentation,” Cadabam says. “When AI reduces behaviour to simplified patterns without clinical context, it misses the nuances that define effective intervention.”

At Cadabams, this tension is addressed through a clear division of roles. AI handles consistency—tracking patterns, enabling engagement, and maintaining continuity between sessions. Clinicians remain central to interpretation and decision-making. As Cadabam emphasises, “The clinician remains central to interpretation, formulation and decisions that require judgement.”

“The most effective use of AI in this space is as an augmentation layer,” he says, “one that extends the reach and consistency of skilled clinicians, not one that tries to replicate them.”

The distinction is subtle but critical. AI can observe and assist, but it cannot yet understand in the way care requires. “Short-term regulation support in a controlled environment is not therapeutic equivalence,” Cadabam cautions.

Rewriting Access to Learning

“If people have access to content, their life just improves,” Akshita Sachdeva, Co-Founder of Trestle Labs, said in an earlier conversation with AIM.

That insight sits at the core of Trestle Labs’ approach. Their platform enables users to interact with content in ways that align with how they process information, listening instead of reading, translating across languages, or navigating material at their own pace, for the visually impaired.

“People who may not come from an English-speaking background or who may face literacy barriers… it helps everybody,” Sachdeva points out. During this conversation, she discusses Kibo, an assistive technology for the visually impaired that also addresses literacy barriers faced by neurodivergent individuals.

For neurodivergent learners, this flexibility is not a feature—it is a necessity. The ability to shift formats, reduce cognitive load, and engage with content differently can determine whether learning is accessible at all.

Scaling Inclusion

For Prateek Madhav, CEO of AssisTech Foundation, the challenge is not just building solutions—but ensuring they reach those who need them.

“We have created a whole system… which is very divided. It is us, and it is them, and that’s not fair,” he said in a previous conversation with AIM.

Many individuals remain outside formal systems of support—not because they do not exist, but because those systems are inaccessible, fragmented, or poorly understood.

Technology, in this context, offers a way to bridge that divide.

Meanwhile, Nair sees this playing out at scale in markets like India. “There’s a large English-speaking population consuming global content, while also producing content at a massive scale… that creates a two-sided comprehension gap.”

“Technology is such a force multiplier… the same reuse of technology should be used to help people with disabilities across the world,” Madhav says.

What the Evidence Shows, and What it Doesn’t

Despite the momentum, AI in neurodivergence remains an evolving field.

Current evidence suggests improvements in immediate outcomes—better engagement, reduced behavioural escalation, improved responsiveness in structured settings.

But long-term impact remains less clear.

“What the field currently lacks is longitudinal outcome data,” says Cadabam, pointing to the absence of evidence around sustained behavioural change without parallel human support.

This distinction matters. A system that helps a learner engage in the moment is valuable. But whether that engagement translates into sustained learning or behavioural outcomes is a different question.

The Challenge of Context

Perhaps the most complex challenge lies in something less visible: bias.

In India, neurodevelopmental assessment is deeply contextual. Language, culture, classroom norms, and socioeconomic conditions all shape how behaviour is expressed and interpreted.

“AI systems are pattern-driven,” Cadabam says. “When those patterns are trained on narrow datasets… the risk is not just inaccuracy but systematic misinterpretation.”

Multilingual exposure, for instance, may be misread as a language delay. Cultural communication styles may be flagged as deficits. Variations in attention may be over-pathologised in one context and overlooked in another.

For Nair, the goal is clarity. “This isn’t about replacing how people sound… it’s about making them easier to understand without losing what makes their voice their own,” he says, describing it as “comprehension without conformity.”

A Shift in Control

What AI ultimately changes is not just systems, but relationships.

Historically, the burden of adaptation has fallen on individuals to learn, behave, and fit into predefined structures. AI can redistribute that burden.

It allows learners to engage with content in formats that work for them. It enables early signals to surface without requiring formal clinical pathways. It provides interaction modes, voice, text, and multimodal interfaces that reduce cognitive load.

In doing so, it shifts control.

This shift is subtle, but its implications are profound. When systems begin to adapt to individuals, rather than the other way around, inclusion moves from accommodation to design.

India’s AI ecosystem is moving quickly, but its long-term impact will depend on how carefully it balances ambition with responsibility.