Agentic AI
Definition
AI systems that can autonomously plan, reason, and take actions to accomplish complex goals with minimal human intervention.
Use Cases
- Klarna: Customer service assistant that handles common support requests such as order issues, refunds/returns questions, and account inquiries. — Deployed an AI assistant in customer support workflows to answer questions and resolve routine issues, with escalation paths to human agents for complex cases. (Reported that the assistant handled a large share of customer service chats and reduced response times while improving operational efficiency.)
- Instacart: Shopping assistant that helps customers find products, get recommendations, and build carts based on preferences and constraints. — Integrated an LLM-powered assistant into the Instacart app experience, grounded in product catalog and shopping context to provide guided, multi-step help. (Improved product discovery and customer experience by enabling conversational, goal-driven shopping flows.)
- Duolingo: Conversational language practice that adapts to the learner and provides interactive role-play scenarios. — Used LLM-driven conversational features to simulate dialogues and provide feedback, creating an agent-like tutoring experience within the app. (Expanded premium feature offerings and increased engagement through more interactive practice experiences.)
Provider Equivalents
- AWS: Amazon Bedrock Agents
- Azure: Azure AI Agent Service
- GCP: Vertex AI Agent Builder
- OCI: OCI Generative AI Agents
Frequently Asked Questions
- What's the difference between AI agents and chatbots?
- A chatbot mainly answers questions in a conversation. An AI agent can also plan and take actions to achieve a goal—like calling APIs, updating a ticket, booking an appointment, or triggering a workflow—often across multiple steps, with guardrails and human escalation when needed.
- When should I use AI agents?
- Use AI agents when work involves repeatable goals that require multiple steps or tool use, such as customer support triage, IT helpdesk automation, order status and refunds, report generation, or internal knowledge search plus actions (create tickets, update CRM). Avoid them for high-risk decisions without strong controls, unclear ownership, or when a simple search/FAQ system is sufficient.
- How much do AI agents cost?
- Costs usually come from (1) model usage (tokens/requests), (2) orchestration/agent runtime charges (if the platform bills separately), (3) tool calls (API gateways, serverless functions, databases), (4) retrieval costs (vector database/storage, embeddings), and (5) monitoring and human-in-the-loop review. Pricing varies by provider, model size, traffic volume, context length, and how often the agent calls tools.
Category: ai-ml
Difficulty: advanced
Related Terms
See Also