GS Paper III 2026 · Q16
15 marks · 250 wordsWhat is agentic Artificial Intelligence (AI) ? Explain its working. Describe its applications with suitable examples. Discuss the advantages, risks and challenges associated with agentic AI systems.
Approach · directive: “what / explain / describe / discuss”
What it asks · Define agentic AI, explain how it works, give applications, and weigh its benefits against its risks and governance challenges.
The question has 4 parts — answer each
- What is agentic AI: define it
- Explain its working
- Describe its applications with suitable examples
- Discuss the advantages, risks and challenges of agentic AI systems
Open with · Agentic AI means systems that pursue a goal on their own — planning steps, using tools and adapting — rather than only answering a prompt.
Cover
- Working: a large language model as the reasoning core; breaking goals into sub-tasks; memory; tool use (APIs, browsers, code); feedback loops; multiple agents coordinating.
- Applications — business: customer-service agents resolving cases end to end, coding agents, research assistants, supply-chain and IT operations.
- Applications — public: grievance handling, help with welfare applications, farm advisories in Indian languages, health triage, fraud detection.
- Advantages: productivity, round-the-clock service, personalised delivery at scale, handling complex multi-step work.
- Risks: errors compounding across steps, unintended actions, security threats such as prompt injection and misuse of credentials, privacy breaches.
- Challenges: accountability when an agent acts, bias, job displacement, energy and compute needs, concentration in a few firms, cross-border regulation.
- Governance: human oversight for high-stakes actions, audit trails, India AI Governance Guidelines (2025), DPDP Act, 2023, AI Safety Institute.
Close with · Agentic AI should be deployed with humans in the loop, so that autonomy adds capacity without diluting accountability.
Add value (verified)
- India's AI Governance Guidelines take a risk-based approach and do not permit unrestricted deployment of high-risk AI. PIB — MeitY on India AI Governance Guidelines (19 December 2025) ↗“The Guidelines do not allow unrestricted deployment of high-risk AI systems. It adopts a risk-based, evidence-led and proportional governance approach.”
Question: UPSC's CS (Main) 2026, GS Paper III — paper ↗. Approach: Minimalist IAS, checked 30 Sept 2026 (how we verify) — UPSC publishes no model answers. ·
Model answer · 305 words (UPSC limit 250) · Minimalist IAS
Agentic AI refers to systems that pursue a goal on their own, planning steps, using tools and adapting to results, rather than only answering a single prompt.
How it works
- A large language model serves as the reasoning core: it breaks the goal into sub-tasks, keeps a memory of context, calls tools such as APIs, browsers and code, checks results and iterates until the goal is met.
- Several agents may coordinate, each with a role, such as one that plans, one that executes and one that verifies.
Applications
- Business: customer-service agents that resolve a case end to end, coding agents, research assistants, supply-chain and IT operations.
- Public services: grievance handling that routes and follows up complaints, help with welfare applications, farm advisories in Indian languages, health triage and fraud detection.
- Illustration: an agent asked to arrange a journey checks availability, compares fares, fills the form and, with the user's approval, pays and files the ticket.
Advantages
- Productivity and round-the-clock service; personalised delivery at scale; capacity for complex, multi-step work that plain chatbots cannot complete.
- For India: scarce expertise is scaled, since one agent can serve lakhs of citizens in their own language, narrowing the gap between entitlement and delivery.
Risks
- Errors compound across steps; unintended actions with real-world effects; security threats such as prompt injection and misuse of credentials; privacy breaches through broad data access.
Challenges
- Accountability when an agent, not a person, acts; bias; job displacement; energy and compute needs; concentration in a few firms; regulation across borders.
- Governance response: human oversight for high-stakes actions and audit trails; India's AI Governance Guidelines (2025), which take a risk-based approach and bar unrestricted deployment of high-risk systems; the DPDP Act, 2023 for personal data; an AI Safety Institute.
Agentic AI should be deployed with humans in the loop, so that autonomy adds capacity without diluting accountability.
Written by Minimalist IAS from facts checked at source (how we verify) — a little fuller than exam length, so every part of the question is covered; in the hall, keep the structure and trim the detail. UPSC publishes no model answers: compare your structure and coverage with this, then write your own.