A group of researchers  have come up with the  ‘Next-Billion AI Index: The compass for AI utility and adoption in the global majority’ .

This new study challenges the global AI industry’s obsession with benchmark rankings with its new framwork Next Billion AI Index Framework (Nexbax AI Index). It argues that the metrics dominating AI development in the world are essentially useless for the billion-plus users across India and the Global South.

The Nexbax AI index is developed by Ambrish Rawat, Kush R. Varshney and Jessica He from IBM research, Subhabrata Majumdar from Indian Institute of Management Bangalore (IIMB), Claudio Pinhanez from University of Sao Paulo, Yann Le Beux from YUX Design, Satyapriya Krishna from Harvard University, Rahul Gupta from Amazon and Rumman Chowdhury from Humane Intelligence.

Even in India’s AI policy discourse, NITI Aayog increasingly recognises that readiness depends on affordability, multilingual access, and compliance-by-design. According to the researchers, the Nexbax AI Index gives that intuition a rigorous and a measurable form.

Subhabrata Majumdar, assistant professor from IIMB, said, “This framework is developed with coauthors not only from India but people spanning from across Africa, Europe, and North America. We are trying to evaluate if the existing AI systems can be deployed, trusted, and sustained under the constraints that define adoption in markets like India, because Indian market has diverse users using devices with intermittent connectivity, low-cost devices, multilingual and low-literacy users, and informal small-business workflows.”

He said, “The project was initiated around six months ago during the India AI Summit in February 2026. The idea behind developing the framework was to guide the development of better AI systems so that these systems can solve the problems of people working at the grassroot level. For example, giving the most important health updates to ASHA workers who work in the remote villages of India where internet connectivity hasn’t seeped in yet. Such AI systems or developments must be brought in India to help deal with the real world challenges.”

The Nexbax AI Index is developed on the basis of 10 measurable dimensions organised under three themes: Effective Efficiency, Operational Practicality, and Societal Integrity. The 10 dimensions included cost effectiveness and price performance, resource efficiency, adaptability and customisability, interoperability, resilience, reliability and robustness, usability and automation, education and empowerment, trustworthiness and ethics, multiculturalism, inclusivity and pluralism, openness and collaboration.

The team conducted a formative expert evaluation of the first version of the index through one-hour user interviews with eleven experts (participants) working on AI technology and applications for customer usage. These 11 participants included founders of various companies’ developers, and product leaders.

 The goal of this evaluation was to assess the clarity, actionability, and perceived accuracy of the dimensions for technical and product stakeholders, and to iterate on them based on expert feedback. Participants tested the index by rating three illustrative technology configurations across the ten dimensions. They shared feedback on their experience using the dimensions, including where they saw value, points of confusion, and suggestions for improvement.

 Prof. Majumdar explained, “Overall, the experts we interviewed felt that the index was useful for reasoning about AI utility and adoption in next-billion settings from a developer and product perspective. One participant explained that the tools used to evaluate customers are beyond localisation, ease of use, reliability, the cost, the set up time — all those things are hit. Those are the same factors that the founders consider to pick their vendors. When asked to identify the most important decision-making considerations in their work, ten participants chose cost effectiveness, five chose usability and automation, four chose trustworthiness and ethics. It is worth noting that these dimensions span all three themes, providing early signal that each theme captures distinct and meaningful aspects of real-world decision making.”

The team found that from a usability standpoint, the experts generally found the dimensions and rubrics easy to apply, although some felt that it would have been difficult without the guidance and clarifications they received from the researchers during the study session.

In that sense, the value of nexbax is not only that it produces an index, but that it shifts the object of evaluation itself: from asking which AI systems are most capable in idealised settings to asking which systems create the conditions for useful, inclusive, and sustainable adoption where the next billion users live and work.

 The researchers have clarified that a future iteration of this work will include aggregated, system-specific ratings from a larger and more diverse stakeholder sample, including public-sector implementers, civil society organizations, local domain experts, and affected users.