AI-driven air quality management systems can enable faster, better decision-making, provided they are secure, safe.

Published on: Aug 25, 2026, 07:17:21 IST

Prefer HTon Google

Share via
Copy link

In a little over two months, winter will descend on the Indo-Gangetic Plain, and with it, the familiar smog. For over three crore people in Delhi-NCR, this means tracking the Graded Response Action Plan (GRAP) directives of the Commission for Air Quality Management in National Capital Region and Adjoining Areas (CAQM), and planning their activities. The CAQM will rely on the Air Quality Early Warning System’s (AQEWS) forecasts to decide which stage of GRAP to impose.

The AQEWS is remarkably accurate. According to a study by the Council on Energy, Environment and Water (CEEW), it can predict days with air quality index (AQI) of 300 and above with over 80% accuracy. A series of crises — the Great Smog of 2016, the pollution episodes of 2017, the deadly dust storms of 2018 — gave birth to this system. Today, seven other cities beyond Delhi have their own AQEWS: Pune, Bengaluru, Hyderabad, Ahmedabad, Kolkata, Mumbai, and Jaipur.

Crises gave us a generation of digital air quality management tools. The current spring of artificial intelligence (AI) offers newer, better ones. But we must tread with caution. How these AI-enabled systems are designed, used, and funded will determine whether they become a long-term paradigm shift or end up in the graveyard of obsolete technology.

First, adopt AI-driven air quality decision support systems (AQDSS) and incorporate end-user feedback from day one. Air quality data comes from monitoring stations, models and forecasts, low-cost sensors, emission inventories and custom surveys. Gathering it on a single platform, analysing it, and drawing actionable insights is challenging. Traditional decision support systems such as the AQEWS simplify these tasks, but still require skilled personnel to navigate them and convert data into action.

Such AI-enabled systems bring air quality managers closer to action by eliminating complex intermediate steps. Through simple natural language queries — like one would type on a mobile phone — managers can ask for insights while the AI runs complicated analytical tasks in the background, presenting only what matters. But involving the end user in system design must be non-negotiable. A CEEW study found that even well-designed decision support systems could fail if they do not account for how end-users engage with them. Regular feedback from decision-makers ensures systems are designed around their actual requirements.

IIT Gandhinagar and CEEW recently developed a forthcoming AI application called VayuChat that integrates data from over 300 low-cost air quality sensors installed at construction sites in Thane, Maharashtra, and helps the municipal corporation identify sites that exceed pollution thresholds or fail to relay data. The system incorporated feedback from officials in the mockup stage itself. The Brihanmumbai Municipal Corporation’s collaboration with IIT Kanpur on an AI-enabled decision-support system that draws real-time insights from 75 sensors across Mumbai is another good model of end-user collaboration.

Second, build capacity to use AI-driven systems well and to intervene when they go wrong. Replacing traditional DSS with AI does not eliminate the need for capacity-building. The quality of output depends on the clarity and depth of the user’s questions. Personnel still need a deep understanding of air quality and training on feeding effective prompts. These systems are not foolproof either. If a system meant to monitor 5,000 construction sites daily has an error rate of even 1%, it would flag 50 sites as problematic when they are not. Having humans in the loop who can judge AI output and make final decisions is essential. Agencies adopting AI-powered systems must ensure adequately trained personnel are in place before widespread deployment.

Third, ensure these systems are secure and invest in local AI models and applications. Air quality management involves sensitive data that could relate to government finances, enforcement patterns, and compliance histories. Such data should be analysed locally using offline AI models and not transmitted to external data centres. Security aside, bespoke offline models will also liberate government organisations from the usage constraints that come with online models. The best AI models available today have been developed abroad and carry risks of restrictions, such as the recent US export ban on Anthropic’s advanced models. This calls for developing cutting-edge indigenous models. That also carries a risk of failure, and government bodies should have an appetite for it by supporting research institutes and startups developing better models and applications.

The Centre’s IndiaAI Mission is an encouraging step in the right direction. With an outlay of over $1 billion and 38,000 GPUs, it aims to de-risk the development of innovative applications across agriculture, health care, weather forecasting and governance — all domains closely linked to air quality. With a comprehensive plan that includes indigenous AI models, hosting innovative datasets, upskilling programmes, and access to computational resources, the mission aims to position India as an AI leader by 2035, the year AI is expected to add $1.7 trillion to India’s economy. The third phase of the mission, between 2027 and 2029, aims to roll out pilot projects in high-readiness sectors to test AI solutions in real-world settings. Air quality is a national problem. AI models and applications to better manage it should be considered a priority in this phase.

AI-driven AQ management systems can enable faster, better decision-making, provided they are secure, safe, and genuinely tailored to end-users’ requirements. Without attention to these factors, even the best systems could end up unused. The air we breathe deserves better.

Arunabha Ghosh is CEO, the Council on Energy, Environment and Water (CEEW), member, Commission for Air Quality Management (CAQM), and chair, AI and Climate Expert Group for the India AI Impact Summit. Mohammad Rafiuddin is programme lead at CEEW. The views expressed are personal

Get Current Updates on India News, Elections 2024, Lok sabha election 2024 voting live , Karnataka election 2024 live in Bengaluru , Election 2024 Date along with Latest News and Top Headlines from India and around the world.

See Less