Upload Notebook

src/agentic_multiwriter/agents/agentic_ai_leadership_session.ipynb

@@ -0,0 +1,346 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "bfff9a65",
+   "metadata": {},
+   "source": [
+    "# Agentic AI as a Partner in Leadership Excellence\n",
+    "\n",
+    "Welcome to **Session #6** of the One Million Leaders Asia (OMLAS) Champion Training (Pakistan edition). In this hands‑on session we will explore how *Agentic AI* can support leadership excellence. You don't need to be a technical expert – this notebook is designed for beginners and decision‑makers who want to understand what agentic AI is, why it matters for leaders, and how we can build a simple multi‑agent system.\n",
+    "\n",
+    "**Speaker**: *Kheem Parkash Dharmani* – AI & Machine Learning Engineer with deep expertise in NLP, Generative AI, and LLM‑based solutions. Kheem has built multi‑modal AI applications for healthcare and automation and will guide us through the session.\n",
+    "\n",
+    "**Event details**:\n",
+    "\n",
+    "- **Date**: **7 December 2025** (Wednesday)\n",
+    "- **Time**: **08:00–09:00 PM PKT**\n",
+    "- **Location**: Online (Pakistan)\n",
+    "\n",
+    "Below is the event poster to set the context:\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "c5699c52",
+   "metadata": {
+    "hide_input": true
+   },
+   "outputs": [],
+   "source": [
+    "from IPython.display import Image, display\n",
+    "# Display the event poster stored in the shared directory\n",
+    "image_path = '/home/oai/share/1765357356721.jpeg'\n",
+    "display(Image(filename=image_path, width=600))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "0a239580",
+   "metadata": {},
+   "source": [
+    "## Session Agenda (1 hour)\n",
+    "\n",
+    "1. **Introduction & Motivation** – why agentic AI matters for leadership.\n",
+    "2. **Understanding AI, Agentic AI & Multi‑Agent Systems** – key definitions and simple analogies.\n",
+    "3. **Agentic AI for Leadership Excellence** – benefits, challenges and examples.\n",
+    "4. **Hands‑On: Build a simple multi‑agent workflow** – using Python and LangGraph (conceptual demonstration).\n",
+    "5. **Responsible AI & Ethics** – ensuring fairness, transparency, privacy and accountability.\n",
+    "6. **Q&A / Discussion** – your questions answered.\n",
+    "7. **Wrap‑up & Next Steps** – resources and recommendations.\n",
+    "\n",
+    "Each section contains easy‑to‑follow explanations and interactive examples. Feel free to run the code cells if you're comfortable using Python, or simply read along.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "53981b39",
+   "metadata": {},
+   "source": [
+    "## 1. Introduction & Motivation\n",
+    "\n",
+    "Artificial intelligence (AI) has progressed from simple automation to systems that can reason, learn and act in complex environments. **Agentic AI** refers to AI systems that take actions autonomously, coordinate with other agents, and adapt to new information. Instead of acting as a single monolithic model, **agentic systems are built like teams** – each agent specialises in a task and the system orchestrates them like a relay race.\n",
+    "\n",
+    "For leaders, this shift means AI can become a *partner* rather than just a tool. Agentic AI can help:\n",
+    "\n",
+    "- **Gather and synthesise information** quickly.\n",
+    "- **Take informed actions** (e.g., sending emails, analysing reports) with oversight.\n",
+    "- **Scale decision‑making** by handling routine tasks and freeing up leaders' time for strategic thinking.\n",
+    "- **Promote collaboration** between humans and AI, encouraging new roles such as **AI whisperers** and **Chief Agentics Officers**【644505837552892†L386-L436】【164762978622591†L63-L68】.\n",
+    "\n",
+    "In Pakistan and across Asia, institutions are exploring how to integrate AI ethically and responsibly. This session aims to demystify agentic AI so leaders can harness its potential while understanding its limitations.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "9d4892bf",
+   "metadata": {},
+   "source": [
+    "## 2. Understanding AI, Agentic AI & Multi‑Agent Systems\n",
+    "\n",
+    "Let's clarify some terminology:\n",
+    "\n",
+    "### 2.1 Artificial Intelligence (AI)\n",
+    "AI encompasses techniques that enable machines to perform tasks that normally require human intelligence, such as learning, problem‑solving and language understanding. Traditional AI often focuses on *narrow tasks* (e.g., image classification) and requires human direction.\n",
+    "\n",
+    "### 2.2 Agentic AI\n",
+    "According to the AlphaBOLD discussion, **agentic AI** refers to **autonomous systems that can reason, plan and act** across multiple platforms with minimal human supervision【644505837552892†L379-L383】. Agentic AI differs from standard automation in several ways:\n",
+    "\n",
+    "| Feature | Traditional Automation | Agentic AI |\n",
+    "| --- | --- | --- |\n",
+    "| **Behaviour** | Follows a fixed script; limited adaptability | Can reason, plan and adapt to new information |\n",
+    "| **Interactions** | Single agent or tool | Multiple agents collaborating via a shared state |\n",
+    "| **Human oversight** | High; requires explicit instruction | Lower; humans provide goals and guardrails |\n",
+    "| **Outcome** | Efficiency in repetitive tasks | Flexibility and creativity in complex tasks |\n",
+    "\n",
+    "A helpful analogy is a **relay race**: Runner A (an agent) completes a lap (task) and passes the baton (data) to Runner B. Each runner has a specific role, and the race depends on smooth handovers.\n",
+    "\n",
+    "### 2.3 Multi‑Agent Systems (MAS)\n",
+    "A **multi‑agent system** is a computational system where multiple agents interact with each other and their environment to achieve individual or collective goals【82730657746531†L104-L110】. Key points:\n",
+    "\n",
+    "- Each agent has its own knowledge and goals.\n",
+    "- Agents may **cooperate**, **compete** or **coordinate**.\n",
+    "- MAS are used to tackle problems too complex for a single agent (e.g., logistics, robotics, smart grids).\n",
+    "\n",
+    "In our context, an agent can be a piece of code (function) that performs a specific task (e.g., research or writing). We coordinate these agents using a workflow engine like LangGraph.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "89ab0c0a",
+   "metadata": {},
+   "source": [
+    "## 3. Multi‑Agent Workflow & LangGraph\n",
+    "\n",
+    "**LangGraph** is a lightweight Python framework for orchestrating a graph of tasks or agents. Instead of writing one giant prompt, you break your logic into smaller steps. In LangGraph:\n",
+    "\n",
+    "- **Nodes** represent the agents or functions (the runners).\n",
+    "- **Edges** represent the rules of who goes next (the track).\n",
+    "- **State** is the shared memory accessible to all agents (the baton).\n",
+    "\n",
+    "Below is a simple diagram illustrating the flow between a *Researcher* agent and a *Writer* agent. The Researcher looks up information, updates the shared state with the findings, and hands control to the Writer, who drafts content based on the research.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "2189ef3d",
+   "metadata": {
+    "hide_input": false
+   },
+   "outputs": [],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "\n",
+    "# Define positions for nodes\n",
+    "positions = {\n",
+    "    'Start': (0.5, 0.5),\n",
+    "    'Researcher': (2.0, 1.0),\n",
+    "    'Writer': (3.5, 0.0),\n",
+    "    'End': (5.0, 0.5)\n",
+    "}\n",
+    "\n",
+    "# Plot nodes\n",
+    "fig, ax = plt.subplots(figsize=(6, 3))\n",
+    "for node, (x, y) in positions.items():\n",
+    "    circle = plt.Circle((x, y), 0.3, color='#c8e6c9', ec='black')\n",
+    "    ax.add_patch(circle)\n",
+    "    ax.text(x, y, node, horizontalalignment='center', verticalalignment='center')\n",
+    "\n",
+    "# Draw arrows between nodes\n",
+    "ax.annotate('', positions['Researcher'], positions['Start'], arrowprops=dict(arrowstyle='->', lw=2))\n",
+    "ax.annotate('', positions['Writer'], positions['Researcher'], arrowprops=dict(arrowstyle='->', lw=2))\n",
+    "ax.annotate('', positions['End'], positions['Writer'], arrowprops=dict(arrowstyle='->', lw=2))\n",
+    "\n",
+    "ax.set_xlim(0, 5.5)\n",
+    "ax.set_ylim(-1, 2)\n",
+    "ax.axis('off')\n",
+    "ax.set_title('Simple LangGraph Workflow')\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "8328fc0a",
+   "metadata": {},
+   "source": [
+    "## 4. Agentic AI for Leadership Excellence\n",
+    "\n",
+    "Traditional leadership focuses on **vision**, **strategy**, **communication** and **decision‑making**. Agentic AI can augment these skills by:\n",
+    "\n",
+    "- **Speed & Scalability**: AI agents can handle repetitive tasks and process large volumes of data, allowing leaders to focus on high‑value decisions【644505837552892†L386-L436】.\n",
+    "- **Enhanced Insight**: Research agents can gather information from various sources and summarise key points, enabling informed decisions.\n",
+    "- **Continuous Learning**: Agentic systems learn from feedback, improving over time and adapting to organisational needs.\n",
+    "- **New Roles & Skills**: Organisations may create roles such as **AI whisperer** to work alongside agentic systems【644505837552892†L386-L436】【164762978622591†L63-L68】.\n",
+    "\n",
+    "### Pakistan Context\n",
+    "\n",
+    "In Pakistan, many industries are exploring AI adoption – from healthcare to education. Agentic AI can help:\n",
+    "\n",
+    "- **Streamline workflows** in organisations with limited resources.\n",
+    "- **Improve transparency and accountability** by automating record‑keeping and providing audit trails.\n",
+    "- **Enhance collaboration** between departments through shared knowledge bases.\n",
+    "- **Develop local expertise** in AI and data science, enabling innovation across sectors.\n",
+    "\n",
+    "**Key takeaway**: Agentic AI should complement leadership rather than replace it. Leaders remain responsible for ethical decision‑making, guiding the system and ensuring fairness.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "42a57221",
+   "metadata": {},
+   "source": [
+    "## 5. Hands‑On: Build a Simple Multi‑Agent Workflow with LangGraph\n",
+    "\n",
+    "Let's walk through a simple example using Python. We'll build a two‑agent system:\n",
+    "\n",
+    "1. **Researcher** – searches the web for key facts about a topic.\n",
+    "2. **Writer** – drafts a short blog post using the research.\n",
+    "\n",
+    "We'll orchestrate these agents using LangGraph. The code below demonstrates how to define the agents, connect them in a workflow, and run the system. Note that executing this cell requires `langgraph`, `langchain` and `duckduckgo-search` packages, and a local Ollama model for Llama 3. If these packages are not installed, you can still read through the code and understand how the system works.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "3007dec8",
+   "metadata": {
+    "hide_input": false
+   },
+   "outputs": [],
+   "source": [
+    "\n",
+    "# This code defines the shared state, agents, and workflow.\n",
+    "# You can run it if you have the required libraries installed.\n",
+    "\n",
+    "from typing import TypedDict, List\n",
+    "from langgraph.graph import StateGraph, END\n",
+    "from langchain_community.tools.ddg_search import DuckDuckGoSearchRun\n",
+    "from langchain_ollama import ChatOllama\n",
+    "from langchain_core.prompts import ChatPromptTemplate\n",
+    "\n",
+    "# ----- Shared State -----\n",
+    "class AgentState(TypedDict):\n",
+    "    topic: str\n",
+    "    research_data: List[str]\n",
+    "    blog_post: str\n",
+    "\n",
+    "# ----- Researcher Agent -----\n",
+    "def researcher_node(state: AgentState):\n",
+    "    topic = state[\"topic\"]\n",
+    "    search = DuckDuckGoSearchRun()\n",
+    "    try:\n",
+    "        results = search.run(f\"key facts and latest news about {topic}\")\n",
+    "    except Exception as e:\n",
+    "        results = f\"Could not find data: {e}\"\n",
+    "    # Update state with research\n",
+    "    return {\"research_data\": state.get(\"research_data\", []) + [results]}\n",
+    "\n",
+    "# ----- Writer Agent -----\n",
+    "def writer_node(state: AgentState):\n",
+    "    topic = state[\"topic\"]\n",
+    "    data = state[\"research_data\"][-1] if state[\"research_data\"] else \"\"\n",
+    "    llm = ChatOllama(model=\"llama3\", temperature=0.7)\n",
+    "    prompt = ChatPromptTemplate.from_template(\n",
+    "        \"\"\"You are a tech blog writer.\n",
+    "\"\n",
+    "        \"Write a short, engaging blog post about \"{topic}\"\n",
+    "\"\n",
+    "        \"based ONLY on the following research data:\n",
+    "\n",
+    "\"\n",
+    "        \"{data}\n",
+    "\n",
+    "\"\n",
+    "        \"Return just the blog post content.\"\"\"\n",
+    "    )\n",
+    "    chain = prompt | llm\n",
+    "    response = chain.invoke({\"topic\": topic, \"data\": data})\n",
+    "    return {\"blog_post\": response.content}\n",
+    "\n",
+    "# ----- Build the LangGraph -----\n",
+    "workflow = StateGraph(AgentState)\n",
+    "workflow.add_node(\"Researcher\", researcher_node)\n",
+    "workflow.add_node(\"Writer\", writer_node)\n",
+    "workflow.set_entry_point(\"Researcher\")\n",
+    "workflow.add_edge(\"Researcher\", \"Writer\")\n",
+    "workflow.add_edge(\"Writer\", END)\n",
+    "\n",
+    "# Compile the workflow\n",
+    "app = workflow.compile()\n",
+    "\n",
+    "# ----- Run the system -----\n",
+    "inputs: AgentState = {\n",
+    "    \"topic\": \"The future of AI Agents\",\n",
+    "    \"research_data\": [],\n",
+    "    \"blog_post\": \"\",\n",
+    "}\n",
+    "result = app.invoke(inputs)\n",
+    "print(\"\n",
+    "--- Final Output ---\n",
+    "\")\n",
+    "print(result[\"blog_post\"])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e6dfebd2",
+   "metadata": {},
+   "source": [
+    "## 6. Responsible AI & Ethics\n",
+    "\n",
+    "When deploying agentic AI in leadership contexts, ethics must be front and centre. Key considerations include:\n",
+    "\n",
+    "- **Fairness & Bias** – AI systems must be monitored for bias and discrimination. Detecting and mitigating bias is crucial【644505837552892†L501-L512】.\n",
+    "- **Transparency** – Leaders should understand how decisions are made. Explainable AI helps build trust and accountability【644505837552892†L513-L515】.\n",
+    "- **Privacy & Consent** – Collecting data requires clear consent and responsible use【644505837552892†L519-L525】.\n",
+    "- **Accountability & Audit Trails** – Maintain logs and audit trails so actions taken by AI agents can be reviewed and audited【644505837552892†L501-L512】.\n",
+    "\n",
+    "Building agentic systems responsibly means establishing governance processes, involving diverse teams in design, and continuously monitoring outcomes.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "8ab068fc",
+   "metadata": {},
+   "source": [
+    "## 7. Conclusion & Next Steps\n",
+    "\n",
+    "In this notebook you learned:\n",
+    "\n",
+    "- The difference between traditional automation and agentic AI.\n",
+    "- How multi‑agent systems allow specialised agents to collaborate using a shared state.\n",
+    "- Why agentic AI can enhance leadership excellence by speeding up decision‑making, scaling knowledge gathering, and promoting collaboration.\n",
+    "- How to design a simple two‑agent workflow with LangGraph (Researcher → Writer).\n",
+    "- The importance of responsible and ethical AI adoption.\n",
+    "\n",
+    "### Next Steps\n",
+    "\n",
+    "1. **Explore LangGraph & LangChain** – experiment with more complex workflows, add a *critic* agent or a *fact‑checker*.\n",
+    "2. **Integrate with your organisation** – identify repetitive decision processes that could benefit from agentic AI support.\n",
+    "3. **Stay informed** – follow developments in AI ethics and regulation.\n",
+    "4. **Educate your team** – encourage cross‑functional learning so teams can collaborate with AI systems effectively.\n",
+    "\n",
+    "Thank you for participating! Feel free to ask questions or propose discussion topics.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "ad30748e",
+   "metadata": {},
+   "source": [
+    "## 8. References\n",
+    "\n",
+    "1. **Agentic AI capabilities and leadership roles** – *AlphaBOLD article: 'Agentic AI and the Rise of AI Whisperers in Leadership'*【644505837552892†L379-L383】【644505837552892†L386-L436】.\n",
+    "2. **Ethical and regulatory considerations** – *AlphaBOLD article*【644505837552892†L501-L512】【644505837552892†L519-L525】.\n",
+    "3. **Definition of Multi‑Agent Systems (MAS)** – *GeeksforGeeks: 'Multi Agent System in AI'*【82730657746531†L104-L110】.\n",
+    "4. **Agentic organisation paradigm** – *McKinsey: 'The agentic organization: Contours of the next paradigm for the AI era'*【378999429249588†L20-L25】【378999429249588†L77-L96】.\n",
+    "5. **Agentics and leadership roles** – *Tony Wood blog: 'Implementing Agentics and AI for Strategic Leadership Excellence'*【164762978622591†L34-L41】【164762978622591†L63-L68】.\n"
+   ]
+  }
+ ],
+ "metadata": {},
+ "nbformat": 4,
+ "nbformat_minor": 5
+}