From BYJU’S to upGrad, Indian edtechs are leveraging AI to enhance learning

Cracking a job interview can be challenging, with numerous steps, including technical questioning rounds that could leave candidates feeling overwhelmed. But what if there was a way artificial intelligence (AI) could help? 

For instance, Ronnie Screwvala-led upGrad has designed a GPT-powered chatbot, which aids its learners in cracking job interviews. Dubbed ChatGPT interview assistant, this tool harnesses the power of GPT technology to simulate mock interview scenarios for job seekers by tapping into a pool of past interview data accumulated by upGrad’s placement experts.

This is one of the several ways edtech firms are leveraging AI to enhance learners’ experiences.

Emerging technologies like AI offer personalised learning experiences and boost student engagement, thus helping edtech firms in retaining students on their platforms. 

According to founders of edtech firms and industry experts, AI can also play a crucial role delivering targeted feedback and support, streamlining administrative tasks, and empowering innovative teaching practices.

“There are numerous use cases for AI, with some already evident and others yet to be discovered. As we delve deeper into this field, we will realise that generative AI has the potential to revolutionise the entire edtech sector,” says Utpal Chakraborty, an AI & Quantum Scientist, who is a Gartner Ambassador for Data and Analytics.

Aiding learning journey

Recently, BYJU'S introduced a suite of AI transformer models—BADRI, Math GPT, and TeacherGPT. BADRI, which stands for BYJU'S Attentive DateVec Rasch Implementation, identifies students’ strengths and weaknesses, and delivers customised questions and learning videos to address the weaknesses.

Its other AI model, TeacherGPT, is an AI-powered assistant that provides personalised guidance to students and evaluates their responses. BYJU’S says the AI model gently steers students towards the correct answers, fostering independent problem-solving.

Dev Roy, Chief Innovation and Learning Officer at BYJU’S, says, these models represent a significant milestone in the company’s pursuit to educate students not only on ‘what to learn’, but also on ‘how to learn’.

Bengaluru-based edtech firm CueMath considers AI assistants to have the potential to generate personalised hints for teachers, aiding children in achieving their learning objectives.

“At Cuemath, our aim is to teach maths intuitively and in fun ways through our unique pedagogy ‘Cue don’t tell’, which means that teachers should cue or hint students to help them navigate through the solution on their own instead of dictating it to them,” says Manan Khurma, CEO and Founder of CueMath.

The Peak XV-backed maths learning startup is currently working on multiple use cases and experimenting with generative AI, with co-pilots for its teachers to create prompts that they can refer to, cue students, and help them solve the problem. 

Edtech unicorn Vedantu is also developing a teaching assistant powered by generative Large Language Models (LLMs), trained on Vedantu's extensive content repository, according to CEO and Co-founder, Vamsi Krishna.

Some of the focus areas—content, personalisation, doubt-solving, and counselling—are currently handled by teachers. Vedantu is developing an AI-powered application which can serve as an assistant teacher. It is accessible to tutors in the background and is expected to be available to students in about six months, Krishna says.

The AI assistant will be capable of curating relevant content tailored to a student’s needs instead of following a rigid rule-based approach. “It will greatly enhance the effectiveness and efficiency of teachers,” Krishna adds.

While several companies are prioritising various forms of teaching assistants, edtech firm Simplilearnis actively working on an advanced ChatGPT-powered chatbot that taps into a vast database of live interactions between learners, trainers, and teaching assistants, according to Co-founder and Chief Operating Officer, Kashyap Dalal.

Still in its early stages, Dalal says the 24/7 human-like support system has huge potential.

Simplilearn is also focusing on learners who have English as a secondary language. For instance, it can help individuals from the Middle East or Europe, whose primary language is French, German, or Arabic. Its ChatGPT tool offers real-time translation by providing transcripts and addressing learners' doubts in the local language.

Google-backed Adda247, which offers government test prep, is going a step further by utilising AI for content creation (including regional languages) and streamlining doubt-clearing processes, says Founder and CEO Anil Nagar. Similarly, upGrad is developing assignment bots to improve the learning experience, and is using ChatGPT to enhance coding modules, including grading answers and generating entire code bases. 

Other applications

While numerous edtech companies have started harnessing AI for learning purposes, many of them initially utilised this emerging technology for other areas of their business, like sales and marketing.

Mayank Kumar, Co-founder and MD of upGrad, states that the firm has used AI in sales and marketing to evaluate user engagement and improve outcomes based on user interactions.

Similarly, Simplilearn is leveraging AI for its sales and marketing strategies by using ChatGPT to create brochures, email templates, and targeted outbound communications for learners, Dalal says.

Simplilearn also aims to automate project evaluation to streamline workflows, enabling teaching assistants to provide value-added services and feedback.

Meanwhile, Cuemath is exploring generative AI to aid its curriculum team by generating content for maths learning, practice, and exam preparation.

According to industry experts, AI can enhance the business of edtech companies through product differentiation, scalability, cost efficiency, targeted marketing, customer engagement, and predictive analytics.

However, drawbacks do exist. Limited human interaction in AI-powered solutions can reduce personal and social aspects of learning while student data collection raises privacy concerns.

Utility and investment

As companies continue to explore the potential of AI in edtech, many have opted for the convenience of utilising large language models (LLMs) like ChatGPT or other AI tools powered by LLMs to enhance different aspects of their business.

“Open-source solutions are gaining significant traction, and within the next four to five months, they are expected to become more cost-effective. This is why our tech teams, particularly for large-scale deployments and applications used by a significant user base, are leaning towards open-source options. However, for smaller applications, the tech team continues to utilise GPT,” Vedantu’s Krishna notes.

Vedantu invests close to $2 million per year on AI across multiple projects, Krishna estimates. The investment will grow based on the value creation of each project.

According to Krishna, Vedantu’s data science and machine learning team consists of around five members. The final application involves other teams as well, so the overall team size is larger. 

upGrad is currently using existing models and aims to build its own in due course. He explains that previously, building models required substantial investments of around $5-10 billion, and at least a billion dollars for a “solid model”. However, these days, the development of models happens at a significantly lower cost, ranging from $1-2 million.

Like Vedantu and upGrad, Simplilearn is also using tools that leverage ChatGPT API. “We have a team of around 15 individuals who were specifically hired for their AI and machine learning skills. Many of our employees have been upskilled to use ChatGPT API and other tools,” Simplilearn’s Dalal remarks.

Approximately 60 individuals, accounting for around 30-40% of Simplilearn’s product and technology team, are involved in AI-related areas, he adds.

In comparison, BYJU’S has chosen a more advanced approach by developing its BADRI model internally. It is currently subjecting its models to rigorous testing at BYJU’S Labs, a facility with advanced technology and expert staff.

A future with AI in education

According to industry experts, human guidance is also essential. Teachers play a crucial role in creating meaningful learning experiences and fostering student development.

“Generative AI is a real shift,” upGrad’s Kumar says. He believes that it will have a much more far-reaching impact on the overarching knowledge-based economy. 

“We will see many jobs getting impacted, a lot of roles getting redundant, and a fair amount of increase in productivity. But in the process, there's going to be massive displacement and re-allocation of jobs,” he explains.

Dalal says, while there may be some shift in roles, “our expectation is not to reduce the workforce or halt hiring in the coming year”. 

“Instead, evolution will likely lead to changes in the types of roles we hire for. The skill sets sought after today may differ from those of two years ago, indicating a natural progression in our hiring practices,” he adds.

What implications does this have for teachers? Is there a possibility of them being replaced by AI tutors?

Chakraborty, who is closely tracking the changes in AI, says, while it is unlikely that human teachers will be completely replaced, there may come a time when their roles evolve to focus on other aspects. Behind the scenes, they will continue to play a crucial role in tasks such as verifying and reviewing content generated by AI to ensure sensitive information is appropriately handled before reaching students, he elaborates.

Cuemath’s Khurma believes no technology, no matter how great or flawless, can ever replace the impact of a teacher. 

“Teachers have a mastery over the art of teaching and know how to customise the instructions to match the different styles and pace of learning of their students. Furthermore, the empathy that a teacher provides to a student can never be replicated by machines,” he says.

(Cover image by Nihar Apte; infographic by Winona Laisram)