एजेंटिक कृत्रिम बुद्धिमत्ता (ए आई) क्या है ? इसकी कार्य-प्रणाली को समझाइए। इसके अनुप्रयोगों का वर्णन उचित उदाहरणों के साथ कीजिये। एजेंटिक ए आई प्रणालियों से सम्बद्ध लाभों, जोखिमों और चुनौतियों की चर्चा कीजिये।
Read this question in English
What 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.
🧭 How to think about this question · directive: “what / explain / describe / discuss”
Decode the demand
- Four directives, four parts: 'what is' (define), 'explain its working' (the mechanism step by step), 'describe applications with examples' (named, sector-wise), 'discuss advantages, risks and challenges' (three separate lists, each with reasons). Each gets its own heading.
- The issue in one line: AI that acts on a goal, not AI that answers a prompt; so the policy question shifts from speech to action.
- The trap: writing one long paragraph on 'AI in general'. The examiner wants the word agentic explained: autonomy, planning, tool use, feedback.
- Carry specifics: one verified scheme number (IndiaAI Mission), the date and stance of India's AI Governance Guidelines, the Summit theme.
Think in a structure
- Open with a two-line definition that contrasts a chatbot and an agent, then turn to the demand.
- Working as a flowchart (goal → plan → tools → observe → revise → act), then one point per component; applications as a table by sector with an Indian example in each row.
- Advantages, risks, challenges as three short headings, each point an ethical or technical keyword plus a reason; keep 'risk' (what can go wrong) separate from 'challenge' (what is hard to govern).
- End from the administrator's seat: human in the loop, risk-based rules already issued, testing, sovereign stack; anchor the conclusion in the AI Impact Summit 2026 theme.
Facts checked at source (5)· open
- 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.”
- Cabinet approved the IndiaAI Mission on 7 March 2024 with a budget outlay of ₹10,371.92 crore.PMO (pmindia.gov.in) — Cabinet approves IndiaAI Mission (7 March 2024) ↗“the Cabinet chaired by the Prime Minister Shri Narendra Modi has approved the comprehensive national-level IndiaAI mission with a budget outlay of Rs.10,371.92 crore.”
- The India AI Governance Guidelines were released on 5 November 2025 as a national framework for safe, responsible and inclusive AI.PIB — India AI Governance Guidelines do not allow unrestricted deployment of high-risk AI systems (MeitY, 19 December 2025) ↗“The India AI Governance Guidelines were released on 5 November 2025. It provides a comprehensive national framework to ensure the safe, responsible, and inclusive development of Artificial Intelligence in the country.”
- The Guidelines create no new statutory audits, appeals or oversight bodies; they rely on existing laws, the IT Act, the DPDP Act and sector-specific regulation, with sectoral regulators enforcing within their mandate.PIB — India AI Governance Guidelines (MeitY, 19 December 2025) ↗“They do not introduce new statutory mechanisms such as independent audits, appeals, or new oversight bodies. Instead, the Guidelines rely on existing laws. These include the Information Technology Act, the Digital Personal Data Protection Act, and sector-specific regulations.”
- PMO: the Prime Minister inaugurated the India AI Impact Summit 2026 on 19 February 2026; theme 'Welfare for All, Happiness of All'; three pillars People, Planet and Progress.PMO (pmindia.gov.in) — PM to inaugurate India AI Impact Summit 2026 on 19 February (18 February 2026) ↗“Three key pillars of the Summit: People, Planet, and Progress”
Question: UPSC's CS (Main) 2026, GS Paper III — paper ↗. Guidance: Minimalist IAS, checked 8 Nov 2026 (how we verify). ·
Model answer· 672words · Minimalist IAS
Agentic AI is software that takes a goal, plans the steps, uses tools and keeps acting on its own until the goal is met, where a chatbot only answers one prompt. The policy question therefore shifts from what AI says to what AI does.
What agentic AI is
- Goal-directed autonomy: it decides which step comes next and executes it without a fresh human instruction at every turn.
- A chatbot drafts a travel plan; an agent compares fares, books the ticket and files the reimbursement claim.
- Autonomy is a spectrum: from a tool that suggests, to a co-pilot that acts on approval, to an agent that acts and then reports.
How it works
- User sets a goal
- LLM plans sub-tasks
- Calls tools (APIs, browser, code)
- Observes result
- Revises plan
- Acts or reports
- Reasoning core: a large language model splits the goal into sub-tasks and chooses their order, the planning stage.
- Memory: short-term context of the task and long-term memory of past interactions keep it coherent across many steps.
- Tool use: it calls databases, payment and e-mail APIs, browsers and code interpreters, so it acts in the real world.
- Feedback loop: each result is checked against the goal and the plan revised, the cybernetic idea Norbert Wiener described.
- Multi-agent systems: a planner, an executor and a verifier agent divide roles, mirroring a human team.
Applications with suitable examples
| Sector | What the agent does | Example |
|---|---|---|
| Business | Resolves customer cases end to end, writes and tests code | Coding agents that open, fix and close software tickets |
| Governance | Routes grievances, follows up, fills welfare forms | CPGRAMS-style redressal with automated reminders to the officer |
| Agriculture and language | Voice advisory in Indian languages | Bhashini-based agents for weather, mandi prices and scheme eligibility |
| Finance and health | Flags fraud, triages patients | Banks blocking anomalous transactions; symptom-to-appointment triage |
Advantages
- Scale and speed: round-the-clock service that finishes multi-step work no chatbot can complete.
- Inclusion: one agent can serve lakhs of citizens in their own tongue, narrowing the gap between entitlement and delivery.
- Public capacity: the IndiaAI Mission supplies the compute and datasets a domestic agent ecosystem needs.
- Cabinet approved the IndiaAI Mission on 7 March 2024 with an outlay of ₹10,371.92 crore.
Risks
- Compounding errors: a wrong early step propagates through later actions with no human to catch it.
- Prompt injection: hidden instructions planted in a webpage or e-mail the agent reads can hijack its goal.
- Credential misuse and privacy: broad access to accounts and personal data widens the breach surface the DPDP Act, 2023 guards.
- Irreversible actions: payments made, files deleted and messages sent cannot be recalled by a later correction.
Challenges
- Accountability gap: when an agent acts, who answers, the developer, the deployer or the user; the IT Act's Section 79 safe harbour assumed passive intermediaries.
- Bias at scale: a skewed model decides thousands of welfare or credit cases before anyone notices the pattern.
- Jobs and energy: back-office roles shrink while data-centre power demand grows as agents spread.
- Concentration and borders: a few firms own frontier models; an agent acting across jurisdictions escapes any single regulator.
Way forward: what a government can do
- Human in the loop: mandatory human approval for high-stakes actions, a kill switch, and audit logs of every tool call.
- Risk-based rules: the India AI Governance Guidelines (5 November 2025) already bar unrestricted deployment of high-risk systems.
- They rely on existing laws, the IT Act, the DPDP Act and sectoral regulators, rather than new oversight bodies.
- Test before deploy: an AI Safety Institute to red-team agents, as the EU AI Act's risk tiers demand for high-risk uses.
- Sovereign stack: Indian-language models on IndiaAI compute, so public agents are not hostage to foreign providers.
- Skilling: reskilling of displaced back-office workers under Skill India, so productivity gains are shared.
Autonomy must add capacity without diluting accountability. The India AI Impact Summit 2026, inaugurated on 19 February 2026, framed the goal as 'Welfare for All, Happiness of All' on the pillars of People, Planet and Progress; agents that act under human command and audit are how India can keep that promise.
Prepared by Minimalist IAS to the answer standard of Shivam Singh (mentor, 7 years) · facts checked against official sources (how we verify) · updated 8 Nov 2026. UPSC publishes no model answers: compare your structure and coverage with this, then write your own.
