Envisioning an AI-first public sector in India
Thought leadershipExplore how an AI-first approach can transform India’s public sector, enabling data-driven governance, improved service delivery and scalable digital infrastructure.
Introduction
AI has become a transformative force for businesses across industries. Since the release of ChatGPT in December 2022, we’ve seen the rapid rise of generative AI, marking a key moment in the evolution of artificial intelligence. Many organisations view AI as a way to gain a competitive edge, with growing interest across areas such as Machine Learning (ML), Natural Language Processing (NLP), Computer Vision (CV), Robotics, and now gen AI.
Business leaders increasingly recognise AI’s potential to reshape their industries. As confidence in its benefits grows, generative AI and AI/ML services are becoming central to technology strategies.
Artificial Intelligence (AI): Traditional AI vs. Generative AI
Artificial Intelligence services includes traditional techniques like Predictive AI and Causal AI. Predictive AI (or Analytical AI) uses algorithms to forecast future events based on historical data. Causal AI determines the causes and effects of events and is used in IT operations and AIOps.
Generative AI (Gen AI) services, leveraging foundational models such as large language models (LLMs), creates new content—text, code, images, audio, and more—by learning patterns from large datasets. This technology builds on existing data to generate novel outputs.
Gen AI use cases – 6C Framework
Gen AI might not be suitable for all use cases. It must be used complementary to Traditional or Discriminative AI. Our Gen AI use cases – 6C Framework – gives a broader direction for identifying use cases for Gen AI. According to our 6C Framework, its use cases fall under six broader categories.
Code
Conversational AIAgent (Customer Interactions)
Content
Concision (Extraction +Summarization)
Creativity
Classification
| Predictive AI | Causal AI |
|---|---|
|
Classification |
Root cause analysis |
|
Regression |
Anomaly detection |
|
Clustering |
Event corelation
|
|
Forecasting, Prediction |
Emergence of Hyper-modal AI (Gen AI + Predictive AI + Causal AI)
Hyper-modal AI combines Generative AI, Predictive AI, and Causal AI to address complex use cases. For example, in drug discovery, Predictive AI identifies potential drugs, while Causal AI analyses gene interactions to find the most effective treatment.
IT observability and operations tool maker Dynatrace recently announced their Hyper-model AI tool Davis. Historically, Davis has utilised Causal AI and Predictive AI for many AIOps use cases. Now they have augmented Davis with Generative AI capabilities, including the capability of interacting with the tool with natural language, Copilot AI Agent, and generating code for automation workflows.
Save
Gen AI Architecture Stack
- Specialised infrastructure: Building and training Artificial Intelligence and Gen AI Models require specialised hardware and infrastructure, including GPUs, FPGA, fast and efficient networking (InfiniBand, Hollow core fibre/ HCF, RoCE / RDMA over Converged Ethernet, RDMA/ Remote Direct Memory Access), modern storage solutions, including cloud-based storage, and overall next-generation data centre technologies.
- Data and cloud: Success of your AI programmes would depend on your data strategy. Build a strong data foundation. Consider data as a product. Utilise modern data architecture patterns. Think about data lineage, data quality and data governance. The specialised infrastructure for AI can be easily accessed through a cloud service provider. Apart from infrastructure cloud service, providers also provide other services, including machine learning as a service, foundational model/ LLM as a service, data services, AI security, governance, RAI, etc.
- Foundational models: Foundational models power Gen AI. Models could be categorised by modality. Single modality models can generate only a single output type, e.g. either text or image. Multi-modality models can generate content multiple output type such as code, text, image, and video. Models also could be categorised by the number of parameters – Large language models (LLM) and small language models (SLM). Models also could be open source or closed source.
- Model customisation: Models need to be customised so that the model can understand your organisation, your client and your sector. This could be done in a number of ways, including prompt engineering, RAG (retrieval augmented generation), fine tuning and pre-training.
- Orchestrator: Our orchestration solutions integrate various AI capabilities to streamline operations and enhance efficiency.
- Gen AI apps: Gen AI powered apps include existing apps where new Gen AI powered features are added, such as newly built intelligent apps, AI agents and chatbots, digital assistants, etc.
AI ethics, security and governance – Responsible AI (RAI) and Explainable AI (XAI)
As AI systems and agents are being used in critical and impactful use cases from identifying the right candidate for a job to underwriting insurance to different healthcare and life sciences, use cases ensuring that these systems are unbiased, fair, safe, secure and reliable should be the topmost priority. These systems should not violate intellectual property (IP) and should abide by AI and data-related laws, regulations and compliances (e.g. the newly formed EU AI Act, Artificial Intelligence and Data Act (AIDA by Govt of Canada), GDPR, etc). Overall, AI systems should protect and reinforce positive human values.
Our responsible Artificial Intelligence services framework covers all aspects of AI ethics and governance. We utilise this framework in all our Gen AI and AI/ML projects.
1.
Fairness and equality
2.
Inclusivity & non-discrimination
3.
Reliability and safety
4.
Security and privacy
5.
Intellectual property
6.
Regulations & compliance
7.
Transparency, explainability and interpretability
8.
Protection and reinforcement of positive human values
9.
Accountability
10.
Governance
Scale your AI Initiatives – Our AI adoption framework
We can help you end-to-end, starting from AI readiness assessment to identifying and prioritising the optimum use cases for you to Gen AI POC to design and build. Our AI ethics and governance services include all aspects of Responsible AI, FinOps for AI workloads and Sustainable AI.
|
Drive your AI initiatives through Businessand Customer focus:
|
Take a business centered approach. Identity AI usecases and carefully judge all the aspects includingvalue delivered, cost benefit analysis, ROI, TCO etc.
|
|
Consider between Generative AI vsTraditional AI vs both:
|
Do not try to paint all your use cases with Gen AI. GenAI might not be suitable for all use case. Take a hybridapproach of Gen AI + Traditional Discriminative AI
|
|
Be curious; Explore the potentials ofnewer areas of AI and Gen AI:
|
AI is still an emerging field. New models, approaches,tools and services are getting introduced every week.Explore the possibilities of new emerging areas of AI
|
|
Bake in AI Ethics, Governance, Security,Privacy, Trust, and Risk:
|
Bake in Responsible AI, Explainable AI and AIInterpretability from inception phase. Utilize best of theclass AI Ethics and Governance tools.
|
|
Take a People Centered andMultidisciplinary approach for AI:
|
Successful AI adoption require a multidisciplinary team including team members with psychology, philosophy,and social sciences background; and not just datascientists, AI specialists, and prompt engineers.
|
|
Build foundational capability in Cloud,Data and Automation:
|
Scaling your AI initiative would require you to build astring foundation in Cloud, Data and Automation. YourAI models would perform as best as your data & cloudfoundations are.
|
|
Focus on scaling your AI initiatives fromPOC to projects across your org:
|
While many enterprise are doing POCs on AI/ Gen AI lessof it is getting productionized. Build a strong strategyand execution approach for scaling your AI initiatives
|
|
Think about Operational aspects fromstrategy and inception phase:
|
AI operations needs a different mindset. Build an AIoperating model and consider all operational aspectsincluding monitoring, scaling, cost, security, process etc.
|
Our Gen AI and AI/ML services
We can help you end-to-end, starting from AI readiness assessment to identifying and prioritising the optimum use cases for you to Gen AI POC to design and build. Our AI ethics and governance services include all aspects of Responsible AI, FinOps for AI workloads and Sustainable AI.
Our insights
Quality in the age of AI
ArticleFor decades, consulting has been defined by structured thinking, rigorous analysis, and carefully developed recommendations.
Pragmatic ethics: India’s calibrated approach to AI deployment
ArticleIndia, home to 1.4 billion people, stands at a pivotal moment in harnessing artificial intelligence (AI) not as a Western luxury, but as a necessity for survival and scale. The recently concluded India AI Impact Summit 2026 in New Delhi underscored this ambition, shifting global discourse from risk-obsessed regulation to developmental.
Our resources
Flyer
Generative AI and AI/ML services
Grant Thornton Bharat provides a comprehensive suite of GenAI and machine learning services designed to support organisations at every phase of their AI adoption journey. Whether you are in the early stages of exploring these technologies or working towards advanced integration, our experts deliver strategic insights and solutions to help you leverage AI for enhanced business outcomes.
Whitepaper
Rise of AI infrastructure
Having a well thought through architecture stack is important for adopting and scaling AI and Gen AI programs across your organization. The AI architecture stack comprises of specialised infrastructure for Gen AI, cloud and data services, model as a service layer (provides a catalogue of foundational models from different providers, and model customisation), middleware (model orchestration tools and APIs), and AI powered Intelligent App. Read our whitepaper to know more about the AI architecture stack.