<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Arun Shan on Medium]]></title>
        <description><![CDATA[Stories by Arun Shan on Medium]]></description>
        <link>https://medium.com/@arunshan04?source=rss-276e8fd793e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*JP6mYyjsh8M_G_csUlQdiw.jpeg</url>
            <title>Stories by Arun Shan on Medium</title>
            <link>https://medium.com/@arunshan04?source=rss-276e8fd793e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 14:57:17 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@arunshan04/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Why Agentic AI with LangChain is Easier Than You Think]]></title>
            <link>https://medium.com/@arunshan04/why-agentic-ai-with-langchain-is-easier-than-you-think-0086bb2f7c2a?source=rss-276e8fd793e------2</link>
            <guid isPermaLink="false">https://medium.com/p/0086bb2f7c2a</guid>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[Arun Shan]]></dc:creator>
            <pubDate>Sat, 19 Jul 2025 12:29:45 GMT</pubDate>
            <atom:updated>2025-07-19T12:29:45.266Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AYLwAkYFscuYMEdA" /></figure><p>Agentic AI represents a powerful new approach to artificial intelligence where large language models (LLMs) act autonomously using memory, tools, logic, and interaction. However, despite its sophisticated capabilities, building AI agents is more approachable than most developers realize.</p><p>When we talk about an AI agent, we’re referring to a system that uses an LLM to decide the control flow of an application. In fact, frameworks like LangChain have emerged specifically to simplify this process. As one of the most widely adopted AI agent frameworks today, LangChain acts as the essential glue between raw language models and structured, logical behaviors. Additionally, it allows developers to build intelligent applications by chaining components like memory, tools, and LLMs into cohesive workflows.</p><p>Throughout this article, we’ll explore why agentic AI with LangChain is surprisingly accessible, even if you’re just getting started with AI development. We’ll break down the fundamentals, walk through a step-by-step process for building your first agent, and showcase practical applications that demonstrate just how powerful these systems can be.</p><h3>What is Agentic AI and Why It Matters</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9P7hiT8gXDMAvH17" /></figure><p>Image Source: <a href="https://www.srajdev.com/">Swapan Rajdev</a></p><h3>Agentic AI definition and core traits</h3><p>The magic of agentic AI lies in its ability to operate with genuine autonomy, making decisions and taking actions without constant human supervision. Unlike basic AI systems, agentic AI combines four distinctive traits that make it uniquely powerful: autonomous decision-making, goal-driven behavior, adaptive learning, and advanced reasoning capabilities.</p><p>At its core, agentic AI doesn’t just follow static commands — it analyzes situations independently, formulates action plans, and executes them. Furthermore, it works toward specific objectives by breaking down complex tasks into manageable steps. This capacity for self-directed action is what transforms it from a simple tool into a true collaborator.</p><h3>How it differs from traditional LLMs</h3><p>Traditional large language models primarily generate content based on specific prompts. Essentially, they’re reactive systems waiting for user commands. Agentic AI, on the other hand, operates proactively with goal-driven autonomy.</p><p>Consider these key differences:</p><ul><li><strong>Control flow</strong>: Traditional LLMs are stateless and single-step; agentic AI is stateful (via memory) and multi-step</li><li><strong>Problem-solving approach</strong>: LLMs generate responses to prompts; agentic AI follows a four-step process: perceive, reason, act, and learn</li><li><strong>Autonomy level</strong>: LLMs require continuous human guidance; agentic AI makes independent decisions to achieve objectives</li></ul><p>Notably, by 2028, industry experts predict <a href="https://convergetp.com/2025/05/06/top-10-agentic-ai-examples-and-use-cases/">33% of enterprise software applications</a> will incorporate agentic AI, up dramatically from less than 1% in 2024.</p><h3>Real-world examples of agentic AI agents</h3><p>Agentic AI is already transforming numerous industries through practical applications:</p><p>In customer service, AI agents can resolve issues by navigating multiple systems, updating records, and confirming resolutions — all autonomously. For businesses facing supply chain challenges (which typically <a href="https://www.fullstack.com/labs/resources/blog/agentic-ai-vs-traditional-ai-what-sets-ai-agents-apart">cost organizations 45% of one year’s profits</a> over a decade), agentic AI can dynamically reconfigure logistics in response to disruptions.</p><p>Healthcare providers are implementing AI agents that monitor patients with chronic conditions via wearable devices, alerting clinicians and scheduling follow-ups when detecting potential risks. Consequently, this leads to better illness management and fewer hospitalizations.</p><p>In cybersecurity, Darktrace’s AI agents autonomously identify and respond to previously unseen cyber-attacks in real-time by using algorithms modeled on the human immune system. This proactive approach represents a fundamental shift from defensive to preventative security strategies.</p><h3>How LangChain Makes Agentic AI Simple</h3><p>LangChain transforms the complex world of AI agents into manageable building blocks that developers can easily assemble. This framework operates as a bridge between raw language models and the structured, autonomous behavior that defines true agentic AI.</p><h3>LangChain agents vs traditional LLM prompts</h3><p>Traditional LLM prompts operate as one-shot interactions — you ask a question, you get an answer. LangChain agents, meanwhile, function as <a href="https://www.deepchecks.com/langchain-components-a-comprehensive-beginners-guide/">dynamic systems</a> that can reason through problems and divide them into smaller subtasks.</p><p>While chains consist of predefined sequences of actions, agents dynamically use the LLM’s analytical abilities to decide which actions to take and in what order. This fundamental shift from static to dynamic interaction allows agents to tackle complex tasks that simple prompts cannot handle.</p><p>The core idea behind LangChain agents is using a language model as a <a href="https://www.analyticsvidhya.com/blog/2024/07/building-smart-ai-agents-with-langchain/">reasoning engine</a> to determine which actions to take in sequence. Unlike traditional approaches that rely solely on completions, LangChain enables true reasoning by building intermediary stages and chaining complex commands together.</p><h3>Key components: chains, tools, memory, and agents</h3><p>LangChain’s architecture revolves around four essential components that work together seamlessly:</p><ul><li><strong>Chains</strong>: Sequences that process input, transform it, and produce output — allowing complex tasks to be broken down into manageable steps</li><li><strong>Tools</strong>: Functions called by agents, representing their capabilities and abstracting the interaction between LLMs and external resources</li><li><strong>Memory</strong>: Mechanisms that store conversation history and context over time, enabling coherent and contextually aware interactions</li><li><strong>Agents</strong>: High-level abstractions that use chains and tools to accomplish specific goals while making informed decisions</li></ul><h3>Why LangChain is beginner-friendly</h3><p>Initially designed with accessibility in mind, LangChain offers a high-level API that simplifies connecting language models to various data sources. Its intuitive design and comprehensive documentation make it accessible to developers with basic Python knowledge, even without prior machine learning experience.</p><p>LangChain particularly excels at combining LLM capabilities with external data sources and APIs, thereby enabling developers to create applications tailored to specific requirements without starting from scratch. For first-time AI developers, this modular approach means you can focus on what you want your agent to do rather than wrestling with complex implementation details.</p><p>Overall, LangChain serves as both the brain and the nervous system of agentic AI applications — organizing your code into functions, defining configuration parameters, and managing runtime behavior while giving developers control over crucial aspects like custom functions and agent behavior.</p><h3>Step-by-Step: Building Your First LangChain Agent</h3><p>Now that we understand what makes LangChain powerful, let’s build our first agentic AI assistant! This practical walkthrough will transform theory into working code you can expand upon later.</p><h3>1. Install LangChain and dependencies</h3><p>First, we need to install the necessary packages. Open your terminal and run:</p><pre>pip install langchain langgraph langchain-community langchain-core</pre><p>Depending on your use case, you might need additional packages. For example, to use OpenAI models or Tavily search:</p><pre>pip install langchain-openai langchain-tavily</pre><p>For memory persistence, add:</p><pre>pip install langgraph-checkpoint-sqlite</pre><h3>2. Set up your environment and API keys</h3><p>Next, store your API keys securely as environment variables rather than hardcoding them:</p><pre>import os<br>import getpass</pre><pre># Set environment variables (in a notebook)<br>os.environ[&quot;OPENAI_API_KEY&quot;] = getpass.getpass(&quot;OPENAI_API_KEY: &quot;)<br>os.environ[&quot;SERPAPI_API_KEY&quot;] = getpass.getpass(&quot;SERPAPI_API_KEY: &quot;)</pre><p>For tracking and debugging, consider enabling LangSmith:</p><pre>os.environ[&quot;LANGSMITH_TRACING&quot;] = &quot;true&quot;<br>os.environ[&quot;LANGSMITH_API_KEY&quot;] = getpass.getpass()</pre><h3>3. Define your agent’s tools and memory</h3><p>Afterward, create tools that give your agent useful capabilities:</p><pre>from langchain_community.utilities import SerpAPIWrapper<br>from langchain_tavily import TavilySearch</pre><pre># Web search tool<br>search = TavilySearch(max_results=2)<br>tools = [search]</pre><pre># Set up memory to persist conversations<br>from langgraph.checkpoint.memory import MemorySaver<br>memory = MemorySaver()</pre><h3>4. Create a simple workflow using LangGraph</h3><p>Subsequently, build your agent using LangGraph, which offers more control than traditional prompting:</p><pre>from langchain_openai import ChatOpenAI<br>from langgraph.prebuilt import create_react_agent</pre><pre># Initialize LLM and agent<br>model = ChatOpenAI(model=&quot;gpt-4o&quot;)<br>agent_executor = create_react_agent(model, tools, checkpointer=memory)</pre><h3>5. Run and test your AI agent</h3><p>Finally, test your agent with a query:</p><pre>input_message = {&quot;role&quot;: &quot;user&quot;, &quot;content&quot;: &quot;What is the weather in San Francisco?&quot;}<br>response = agent_executor.invoke({&quot;messages&quot;: [input_message]})</pre><pre># For conversation persistence, use thread_id<br>config = {&quot;configurable&quot;: {&quot;thread_id&quot;: &quot;conversation-123&quot;}}<br>response = agent_executor.invoke({&quot;messages&quot;: [input_message]}, config=config)</pre><p>Congratulations! You’ve built a basic agentic AI system that can make decisions, use tools, and retain information across conversations. The real power comes from customising tools and workflows to solve specific problems in your domain.</p><h3>Use Cases That Show the Power of LangChain Agents</h3><p>LangChain’s versatile framework is currently powering innovative AI applications across industries. Let’s explore real-world implementations demonstrating why agentic AI through LangChain is becoming increasingly essential.</p><h3>AI research assistants</h3><p>Research assistants built with LangChain dramatically accelerate information discovery and analysis. These agents handle everything from planning research outlines to data collection and publication. Indeed, some implementations feature specialized agents working in tandem — Editor agents plan research structure, Researcher agents conduct in-depth analysis, and Reviewer agents validate findings. This approach allows for parallel processing of different subtopics while maintaining consistency across the final output.</p><h3>Customer support bots</h3><p>Customer support presents a perfect use case for agentic AI. Klarna’s AI assistant has <a href="https://www.langchain.com/">reduced average customer query resolution time by 80%</a>, powered by LangSmith and LangGraph. Similarly, multi-agent helpdesk systems use specialized components — Query Router agents classify intent, retrieval agents search knowledge bases, and escalation agents involve human representatives when necessary. This layered approach has boosted resolution rates by 35–45% compared to single-agent alternatives.</p><h3>Coding copilots</h3><p>LangChain enables powerful coding assistants that break down developer requests into manageable steps. A Planner agent divides features into tasks, a Code Generator writes and debugs code, while a Tester agent validates the output. Primarily, this approach addresses concerns with proprietary solutions like GitHub Copilot, where enterprises risk sharing internal code with external platforms. Accordingly, open-source alternatives using StarCoder and RAG techniques allow businesses to maintain their intellectual property securely within their infrastructure.</p><h3>Enterprise knowledge retrieval</h3><p>Financial institutions leverage LangChain agents for loan decisions by examining customer financial history and applying standard loan policies. A global logistics provider <a href="https://www.langchain.com/">saves 600 hours daily</a> using an automated order system built on LangGraph. Additionally, Trellix, a cybersecurity firm with over 40,000 customers, reduced log parsing time from days to minutes using LangChain’s frameworks.</p><h3>Multi-agent collaboration</h3><p>The true strength of LangChain emerges through multi-agent systems. LangGraph excels at connecting specialized agents as nodes in a workflow graph, enabling sophisticated collaboration patterns. These collaborative designs can follow different architectures — from agents sharing a common scratchpad to hierarchical teams where agents operate as independent LangChain objects. Interestingly, this approach divides complex problems into manageable units handled by specialized agents, creating more robust and intelligent systems.</p><h3>Conclusion</h3><p>Agentic AI with LangChain represents a genuine paradigm shift in how we approach artificial intelligence development. Throughout this article, we’ve seen how LangChain transforms complex AI agent creation into an accessible process through its modular components and intuitive design. Most importantly, this framework bridges the gap between raw language models and the autonomous, goal-driven behavior that defines true agentic systems.</p><p>The step-by-step process we outlined demonstrates just how quickly you can build your first functional AI agent. Rather than wrestling with low-level implementation details, LangChain allows us to focus on defining what we want our agents to accomplish. This accessibility opens doors for developers at all skill levels to experiment with autonomous AI systems.</p><p>Real-world applications clearly showcase the transformative potential of these technologies. From research assistants that accelerate information discovery to customer support bots reducing resolution times by 80%, LangChain-powered agents solve genuine business problems across industries. Additionally, the multi-agent collaboration patterns enabled by LangGraph take these capabilities even further, allowing specialized agents to work together on complex tasks.</p><p>The transition from traditional LLM prompting to agentic AI represents a fundamental shift from static interactions to dynamic, reasoning-based systems. Therefore, we encourage you to start experimenting with the code examples provided. Even simple implementations can yield powerful results when you customize tools and workflows for your specific domain.</p><p>After all, the true power of agentic AI lies not just in its technical capabilities but in how it extends our own abilities. We now stand at the beginning of a new era where AI doesn’t just respond to our prompts but actively helps us achieve our goals through autonomous, intelligent action. The barrier to entry has never been lower — so why not build your first LangChain agent today?</p><h3>Key Takeaways</h3><p>Building agentic AI systems is more accessible than most developers realize, thanks to frameworks like LangChain that simplify complex AI agent development into manageable components.</p><p>• <strong>Agentic AI operates autonomously</strong> — Unlike traditional LLMs that react to prompts, agentic AI makes independent decisions, uses tools, and maintains memory across interactions.</p><p>• <strong>LangChain provides four key building blocks</strong> — Chains, tools, memory, and agents work together to create dynamic systems that reason through problems and execute multi-step workflows.</p><p>• <strong>Getting started requires just five simple steps</strong> — Install dependencies, set up API keys, define tools and memory, create workflows with LangGraph, and test your agent.</p><p>• <strong>Real-world applications show dramatic results</strong> — Companies like Klarna reduced customer query resolution time by 80%, while enterprises save hundreds of hours daily through automated systems.</p><p>• <strong>Multi-agent collaboration unlocks advanced capabilities</strong> — Specialized agents working together can tackle complex problems more effectively than single-agent systems, enabling sophisticated enterprise solutions.</p><p>The barrier to entry for agentic AI has never been lower. With LangChain’s intuitive framework, developers can focus on defining what they want their agents to accomplish rather than wrestling with complex implementation details, making autonomous AI accessible to developers at all skill levels.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0086bb2f7c2a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>