Originally published in IMC Journal, July-August 2025 issue (cover story).
The dawn of artificial intelligence (AI) era has been compared to various game-changing innovations in human history, like fire, the wheel, electricity, and the internet. Imagine the excitement that human beings had when they realized the potential and power of their discoveries and inventions. Look around now, and you find that these innovations have woven themselves into the fabric of our lives; they do not stand out independently but power a lot of the work and leisure that human beings accomplish.
Such will be the case with AI too. The current excitement about AI seems to suggest that it will alter dramatically, and rather quickly, the entire edifice of our living. The advent of this new technology is expected to change everything: how we study, how we work, how we think, how our relationships shape up, and indeed how we go about our daily lives.
What we need to consider, hence, are two aspects: (a) how significant can the change be? and (b) how quickly can it come?
How deep can the changes be
A flurry of news items around us coming with deep intensity and repeated frequencies suggest that quite a few things are changing. With global AI companies reaching trillions of dollars of market capitalization, smart money is betting that the gains from this technology will be exceptionally large. Some companies have announced job cuts or paused hiring on the back of AI; others are highlighting expected enhancements in productivity. Consumers are using the technology to get information and companionship. Across the board, there is a perceptible fear of missing out.
Dig a level deeper, and a report from MIT notes that a large proportion (95%) of AI pilots in companies are failing. There are concerns that the amount of capital expenditure and energy required to power artificial intelligence can be stupendously large. Activists worry about the ethical use of AI, especially as it relates to deepfakes and faux therapies, indicating a wider social concern.
The tussle between the possibilities and challenges means that the equilibrium of what the changes will be and how deep they will they permeate is yet to be determined.
How quickly can it come
Given the fast-changing technology, the possibility set of changes is increasing. While there is no Moore’s Law equivalent in AI yet, the rapid change in technology means that various stakeholders talk of an artificial general intelligence (AGI) being only a few years away. Be that as it may, technological innovation and its seepage into the general public for everyday use can see a significant timing difference.
The first mobile phone call in India between Jyoti Basu, then Chief Minister of West Bengal, and Sukh Ram, then telecom minister of the Government of India, was made in 1995; India reached a billion telephone subscribers in 2015 – a good two decades later. This required massive investments and innovations – in both technology and policy.
It is prudent to assume that AI diffusion will also take time. It may not take as long as two decades as a lot of digital infrastructure is already in place. However, as the recent Airtel-Perplexity partnership shows, the uptake for top-end AI services, even when available for free to the final consumer, leads to slow and limited adoption. While the formal numbers are not yet out, app- tracking data suggests a 1-2% of the subscriber base of Airtel (upwards of 350 million) have taken up the offer to download and start working out of Perplexity.
Eventually as the use cases become clearer and the general understanding of how to effectively use AI becomes deeper, expect that a larger number of customers will start to incorporate this service in their daily lives, work, school, etc.
What should be done till then? Get the basics right
The core building block of AI is high-quality data. The deeper and cleaner the database on which the AI runs, the better the output will be.
High quality data allows for building models that can be both explanatory and predictive. Such models can help identify the interventions that need to be done. What AI can help do is build deeper models, identify patterns which may not have been visible to the human mind, and generally be one step ahead on the possibilities.
For example, say we want to understand how every district or location in India is performing economically. Answering this question means that the AI should have access to reliable and updated datasets on which models and inferences can be built. These datasets will come from a wide variety of primary and secondary sources. These could include, among many other possibilities, understanding the prevailing business scenario, employment across age levels, registration data for property and vehicles, cropping patterns, freight movement, electricity consumption, weather, etc. Building the right data pipeline, which remains refreshed to feed the AI is critical – the output will be as good as the inputs that go in.
Alternatively, say the citizens want to understand which government scheme can they benefit from. A cursory usage of the free large language models may give him/her a list of such schemes and possibly some understanding of the details. However, robust response will require knowing all the applicable schemes, keeping it updated, identifying the actual form, and building the process of getting the right information into the form, and submitting it to the appropriate authorities. Again, while AI can play an important role, we note here the importance of building the right coordination mechanisms across a wide range of data points and keeping them updated. Making AI accessible to people in their local languages may require some more effort.
To manifest a better Bharat with AI, hence, we need to start with the basics of data governance – and get them right. We should expect massive gains in productivity over a period of time when AI starts to become second nature for our daily use – till then, we should be clear we are not blinded by FOMO so much that we miss out on the actual, hard-working opportunities.
