Education has already seen technology transform content delivery, learning management, administration and, more recently, personalised learning and AI tutoring. Yet one of the most important parts of the learning cycle has received comparatively less attention: Assessment—and the teacher at the centre of it.
Before a school can personalise learning, it needs reliable evidence of what students have actually understood, where misconceptions exist and where intervention is required. Much of that evidence still comes from tests, assignments and answer sheets.
Competency-based assessment is not new in India. The National Education Policy (NEP) 2020 has stressed that assessment should focus on testing conceptual understanding, higher-order thinking skills and learning outcomes, not rote memorisation.
The next question is whether the technology used to evaluate those assessments is ready for the same shift.
Much of the conversation around AI in education focuses on what students are taught. But changing curriculum will only be effective with AI assessment infrastructure.
A curriculum might promote critical thought, problem-solving, and advanced application. But if the test can only reward the context of textbook solutions, the education system is still communicating to students that they should have an eye on the exam pattern and answer key.
An obvious gap thus exists between what education is attempting to teach and what assessment is able to reveal.
{{/usCountry}}An obvious gap thus exists between what education is attempting to teach and what assessment is able to reveal.
{{/usCountry}}A modern assessment system needs to look beyond whether an answer is right or wrong. It should be able to identify:
- Whether a student has understood the underlying concept
- Where a particular misconception appears
- Whether an answer demonstrates reasoning or simply memorisation
- Which topics are creating difficulty across a class or cohort
- Where a teacher may need to intervene
This is where AI-assisted assessment can move beyond automating marks towards assessment intelligence turning every answer into evidence of conceptual understanding, misconceptions, reasoning patterns and learning gaps.
Assessment is also a workload problem. According to OECD's TALIS 2024 data, teachers spend around 9% of their working time, on average, marking and correcting student work. Forty per cent of teachers across OECD education systems report excessive marking as a source of stress.
This becomes particularly significant when assessments involve descriptive or handwritten answers. Unlike multiple-choice tests, these responses require interpretation. They also generate large amounts of unstructured information that is difficult to analyse manually at scale.
AI can take on some of the repetitive work and help with response analysis, automated test correction, identifying learning gaps and reducing teacher workload.
The key is that the teacher does not go away from that process. AI can deliver a first assessment and identify the pattern, but the teacher remains in place for judgement, context and final decisions.
This is perhaps the bigger change. A conventional examination produces a score. An intelligent assessment system can potentially produce a learning map.
For example, two students may both score 60%. On paper, they look similar. But their answer patterns may tell very different stories. One student may understand most concepts but make calculation errors. Another may remember definitions but struggle to apply them to unfamiliar problems.
The mark tells us how the student performed. Assessment intelligence begins to tell us why.
A useful AI evaluation system should be able to distinguish between these patterns rather than treating both students as simply “60% performers”. This is where assessment starts becoming an input for personalised learning insights rather than the final step after teaching.
AI tutors, personalised learning platforms and generative content tools will continue to evolve rapidly. But all of them eventually confront the same fundamental question: How do we know whether a student has actually learned?
Assessment is, therefore, more than an administrative function. It is the feedback loop connecting teaching, learning and intervention.
India's assessment reforms already point towards competency-based examination and more frequent measurement of learning. The next step could be building the technology and infrastructure capable of interpreting that evidence at scale.
The future of AI in education may not begin with an AI teacher or another layer of generated content. It may begin somewhere less visible but far more fundamental—with an assessment system that understands learning better.
The opportunity is not to take the teacher out of assessment, but to put the teacher back at its centre—supported by better evidence, greater consistency and timely learning intelligence.
(The views expressed are personal)
This article is authored by Swaminathan Ganesan, co-founder & CEO, Smartail.