Budget 2024 Expectations: Artificial Intelligence

Debashish Banerjee, Partner Shrenik Shah, Partner

Current environment

• Driven by the global buzz around AI and analytics, India’s tech and digital talent, and push from the government, India is emerging as a powerhouse in this area. Estimated to grow at a year-over-year CAGR of more than 30 percent, India Inc. has adopted AI in various segments. India might be behind a couple of global matured markets, but far ahead among developing economies. The application of AI brings in 3–10X improvement in business processes. Hence, if used effectively, the technology will significantly contribute to India’s ambitious goal of reaching a US$5 trillion economy.
• The central and multiple state governments have launched various schemes to push AI adoption in India. Some of these schemes/initiatives are “AI for all” by Niti Ayog and Chair at GPAI in the current G20 summit. MeitY also plans to launch, Bhasini Programme, and YuvAI for skilling Indian youth by the central government. Similarly, multiple state government schemes (notably Tamil Nadu, Telangana, and Karnataka) have been launched to grow and develop skills, propel India Inc. and investments in this area and regulate and govern the use of AI.
• While multiple initiatives have been introduced to make India a global AI hub, a few areas need to be considered. First, accelerate programme execution. AI research and such programmes are being launched by almost all mature countries. It is a topic of innovation where speed (and possibly that is the only thing) matters. Second, start reaping the benefits by launching specific programmes under Public-Private Partnership (PPP) with crowd-sourcing ideas, strict timelines and scaled-up solutions for the country. Third and possibly most critical is the focus on the two foundational elements of AI—data (collection, storage and analytics) and trustworthy use of AI.

Expectations

Top five asks

Ask #1: Accelerate research and development for the use of AI in various areas, including tax

• While research publications and patents in this area have improved from India, we are still way behind in terms of innovation, citation index and global exposure. The government should boost investments in AI or encourage private investments via tax cuts and identify new research areas, such as quantum computing in AI, digital nudge for social good using AI, Explainable AI (XAI), Gen Chem and Computational Biology and smart and connected cities. Governments’ role in promoting academia-industry collaboration will be key. Choosing one or two of the transformational areas mentioned above that still need solutions and require a faster path to completion.
• A dedicated pool of officers across direct and indirect taxes should have the requisite knowledge of AI and its use. This will enable them to use AI in tax assessment, tax administration and taxpayer facilitation. In the long run, the use of AI by tax officials will help focus on non-compliant taxpayers, including early detection of non-compliance trends and taking preventive measures.

Ask #2: Promote the PPP model with value-based implementation

India’s AI strategy focuses on social impact and has called out three sectors as a priority—agriculture, education and healthcare. The key is to identify specific use cases that the government can implement; these cases should have the maximum impact on the society at large. List the key challenges in these sectors, crowd-source ideas, implement those ideas (possibly under the PPP model with value-based impact) and scale the solution to be the service provider to the globe. Some ideas could include the following:

• Digitising lawsuits and judgements in the tax area and implementing a triage model using efficient reviews on tax, property, marital and other easy-to-review litigations.
  
• Using lifestyle-based data to predict future health scores (disease propensity models) for the population to reduce the burden on healthcare and proactive mitigation.
• Digital nudge using AI to prevent fraud on government schemes, financial and tax-related transactions and social issues, such as Swachch Bharat.
• Quality education in local languages for rural India using large language models (LLMs) specific to Indian languages possibly, under the national language transition mission.
• Use of satellite images from ISRO on the crop insurance claims process to make it easy for farmers.

Ask #3: Focus on centralised data repositories for the country

India has taken various significant steps in this area with Adhaar, CoWin, portability of insurance policies and multiple other initiatives. However, to boost AI and analytics further, we need to take the following steps:

• Improve the trust of customers and organisations (locally and globally) that this data is protected and secured. We must change the negative perceptions and narratives that may exist today.
• Completeness of data in several areas, such as having an Electronic Health Records (EHR) and centralised medical record database such as those in other mature countries for effective healthcare solutions.
• Limited but allowable access to the data for research, crowd-sourcing new ideas, and use by industry (using the latest tools, such as creating and opening model APIs and split learning).

Ask #4: Focus on trustworthy AI
Establishing a governance mechanism is crucial to ensure trustworthy AI, covering everything from data sourcing and storage to the application of AI methods. This is a challenging task, particularly in advancing research and creating Intellectual Properties (IP), especially considering India's diverse landscape.

Defining governance and creating frameworks in several areas, including the ones mentioned below, are crucial:

• Data privacy and accountability
• Transparency, attributability, reliability and robustness of models
• Fair and impartial use of tools
• Safety and security of technology stacks
• Ownership of results

Ask #5: Provide support to develop hardware and supercomputing resources

• Encouraging PPP collaboration in establishing and maintaining high-performance computing centres accessible to researchers and start-ups would significantly accelerate AI development.
• Currently, fine-tuning of GenAI models (mainly LLMs) is prohibitively expensive as most computing resources reside outside India, and the data used in areas such as tax administration are voluminous. Government’s decision to support 50 percent of the cost of up to 10,000 GPUs set up in India is a step in the right direction but needs efficient execution.
• Support for companies to set up and scale indigenous cloud services can reduce significant AI development costs. Again, the Government’s Meghraj policy is the right step, but the focus should be on execution and localisation.

Policy recommendations and expected impact/outcome

Recommendation #1: Although the Digital Personal Data Protection Act (DPDP) is a great beginning, the industry needs a deeper dive to ascertain how each industry is using personal data and keeping a balance between fostering innovation and regulating data use. AI is still in its nascent stages in many ways. We have seen many good and unethical/fake/wrong uses of the technology. India must introduce penalties for unethical use, helping it shape up its image as a global AI hub. India’s diversity, cultural identity and values should be considered while drafting policies on AI.

Recommendation #2: Craft a committee/programme that will task each government department to prioritise innovation using AI in their respective areas. Work with research, industry, academia, or other departments and craft a value-driven PPP model to initiate time-bound solutions. This will boost the economy and foster innovation using ethical principles. In addition, the Government should consider a PPP model to spread AI education as a higher education curriculum.

Recommendation #3: Due to complexities, tax litigation takes several years to reach finality. Moreover, a significant amount of money is locked up in litigation. The use of AI technology to sharpen the tax administration’s decision-making process around filing or not filing an appeal is based on key parameters. With the use of Gen AI, tax administrators can summarise tax positions based on facts of the case, legal interpretations, and views of the courts. The use of AI can also help assess winning/losing probabilities for a case, for the tax administration. This can help authorities decide whether to pursue a particular case in litigation or not. AI can help decide whether to litigate a case and optimise the number of cases, saving time and energy of the tax administration. The number of lawsuits can also decrease significantly, reducing the pressure on the legal system.

Recommendation #4: Reduce friction in integrating AI technologies with existing tax systems. This will make fine-tuning large language models easier and develop customised tax language models that can enhance the performance of AI in tax compliance, relaying the correct information to taxpayers, etc.

Recommendation #5: Design a central committee of AI/ML experts in India whose role will be standardising the curriculum for the masses and suggesting/investing on new unexplored technologies. Collaborate with leading academic and industry players to develop and globalise solutions, starting locally in India and using Indian IP.