Knowledge Management & Decision Support: Systems, DSS, AI, and Knowledge Workflows

TAKEAWAYS:

• Knowledge Management Systems (KMS) are digital platforms that organize, share, and reuse organizational knowledge (e.g., wikis, databases, AI chatbots).
• Decision Support Systems (DSS) combine data, models, and user input to help managers solve semi-structured problems (e.g., sales forecasting, supply chain optimization).
• Expert Systems mimic human expertise using rule-based AI (e.g., medical diagnosis tools, loan approval bots in banks like Nabil).
• AI and Analytics (e.g., predictive modeling, NLP) enhance DSS by uncovering patterns in big data (e.g., Daraz’s demand forecasting, Pathao’s route optimization).
• Knowledge Workflows (e.g., communities of practice, crowdsourcing) turn tacit knowledge into explicit assets (e.g., eSewa’s customer service training, Himalayan Java’s coffee quality control).
• Ethical challenges (e.g., data privacy, bias in AI) must be addressed in KMS/DSS design (e.g., NTC’s network planning transparency).

1. Knowledge Management Systems (KMS): The Digital Brain of Organizations

Knowledge is the most valuable asset in the digital age. KMS are IT-based systems that help organizations create, store, share, and apply knowledge to improve performance. Unlike traditional databases (which store structured data), KMS handle tacit knowledge (experience, expertise) and explicit knowledge (documents, reports).

How KMS Works: The Knowledge Cycle

mindmap
  root((Knowledge Management Cycle))
    Create
      Brainstorming
      Documenting
      AI/NLP (e.g., WhatsApp Business auto-replies)
    Store
      Databases (SQL/NoSQL)
      Wikis (e.g., internal Nabil Bank wikis)
      Cloud (Google Drive, eSewa’s knowledge base)
    Share
      Intranets
      Social networks (e.g., Slack for Pathao teams)
      Mobile apps (e.g., Daraz seller forums)
    Apply
      Decision-making (DSS integration)
      Training (e.g., NTC technician manuals)
      Innovation (e.g., Himalayan Java’s flavor profiles)

Types of KMS

Worked Example: eSewa’s Knowledge Management

Problem: eSewa receives 50,000+ customer queries daily, many repetitive (e.g., "How to recharge?"). Solution:

1. Capture: AI transcribes voice calls (IVR) and chat logs.
2. Store: Structured in a knowledge base (SQL database + NLP tags).
3. Share: Chatbot routes queries to FAQs or human agents.
4. Apply: Agents use shortcut links (e.g., "Recharge → Step 1: Open App") to resolve issues faster. Result: 30% reduction in call handling time, 24/7 support.

2. Decision Support Systems (DSS): Turning Data into Decisions

DSS are interactive systems that help managers make semi-structured decisions (e.g., "Should we expand to Pokhara?"). They combine:

• Data (historical sales, market trends),
• Models (forecasting, optimization),
• User interface (dashboards, what-if analysis).

Components of a DSS

flowchart TD
  A["User Input"] --> B["Data Management"]
  B --> C["Model Management"]
  C --> D["Knowledge Management"]
  D --> E["User Interface"]
  E -->|"Feedback"| A

Types of DSS

Worked Example: Daraz’s Demand Forecasting DSS

Problem: Daraz’s warehouses in Nepal face stockouts or overstock due to unpredictable demand (e.g., Diwali sales spikes). Solution: A data-driven DSS with:

1. Data Sources:
   • Past 3 years of sales data (SQL database).
   • External data: Weather (monsoon delays deliveries), festivals (Tihar boosts ghee sales).
2. Model: Machine learning (Python’s scikit-learn) predicts demand with 85% accuracy.
3. Output: Dashboard shows:
   • Stock levels (red = low, green = optimal).
   • Reorder points (e.g., "Order 500 kg rice by Oct 15").
4. Action: Warehouse managers adjust orders via the system.

Advantages of DSS:

• Reduces uncertainty (e.g., "Will this marketing campaign work?").
• Saves time (e.g., NTC’s network planning DSS cuts analysis time by 40%).
• Enables what-if analysis (e.g., "What if we raise prices by 10%?").

Disadvantages:

• High cost (requires skilled data scientists).
• Over-reliance on data (may ignore human intuition).
• Bias in models (e.g., if trained only on urban data, rural trends may be missed).

3. Expert Systems: AI That Mimics Human Experts

Expert systems are a subset of AI that emulate human expertise in a specific domain using:

• Knowledge base (rules, facts),
• Inference engine (logic to apply rules),
• User interface (questions/answers).

How Expert Systems Work

flowchart LR
  A["User Input:<br/>'Patient has fever + cough'"] --> B["Inference Engine"]
  B --> C["Knowledge Base:<br/>'If fever + cough + age >5 → likely flu'"]
  C --> D["Conclusion:<br/>'Prescribe rest + paracetamol'"]
  D --> E["User"]

Real-World Expert Systems in Nepal

Worked Example: Nabil Bank’s Loan Approval Expert System

Problem: Manual loan approval takes 5–7 days; many applications are rejected due to human error. Solution: An expert system with:

1. Knowledge Base:
   • Rules like:
     • "If credit score > 650 AND income > Rs. 50,000 → Approve."
     • "If loan amount > Rs. 5M → Require collateral."
2. Inference Engine: Applies rules in real-time.
3. Output: Instant approval/rejection with reasoning (e.g., "Rejected: Income too low").

Result:

• 80% faster approvals.
• Reduced bias (no favoritism based on branch manager’s mood).

4. AI and Analytics in Decision Support

Modern DSS leverage AI/ML to handle unstructured data (e.g., customer reviews, social media) and predictive analytics.

Key AI Techniques in DSS

Worked Example: Pathao’s Route Optimization with AI

Problem: Pathao’s drivers in Kathmandu traffic waste time due to inefficient routes. Solution: A DSS with AI that:

1. Collects data: Traffic cameras, GPS, historical routes.
2. Uses ML: Predicts congestion (e.g., "Avoid Thapathali at 5 PM").
3. Optimizes routes: Suggests fastest path via real-time updates. Result:

• 20% faster deliveries.
• Lower fuel costs for drivers.

5. Knowledge Workflows: Turning Tacit Knowledge into Action

Tacit knowledge (experience, skills) is hard to document, but KMS can capture it through:

• Communities of Practice (CoP): Groups where experts share knowledge (e.g., NTC engineers).
• Crowdsourcing: Leveraging users to solve problems (e.g., Daraz’s seller forums).
• After-Action Reviews (AAR): Teams reflect on projects to extract lessons (e.g., Himalayan Java’s quality control meetings).

Example: Himalayan Java’s Coffee Quality Control

Process:

1. Tacit Knowledge: Farmers know which beans taste best, but this knowledge is not written down.
2. Capture: Experienced tasters document flavor profiles (e.g., "Batch X has citrus notes").
3. Share: Digital logs are stored in a KMS and shared with roasters.
4. Apply: Roasters use this data to blend beans for consistent taste.

6. Ethical and Social Issues in KMS/DSS

Solution: Organizations must:

• Follow data protection laws (e.g., Nepal’s Electronic Transactions Act).
• Audit AI models for bias (e.g., Nabil Bank tests loan bots on rural data).
• Train employees to use KMS ethically.

In the Real World

1. eSewa’s Knowledge Base

   • Idea Used: Document-driven KMS + AI chatbot.
   • How: eSewa’s internal wiki stores solutions to common issues (e.g., "How to fix a failed transaction?"). The chatbot uses NLP to match user queries to these documents, reducing call center workload by 30%.

2. Nabil Bank’s Loan DSS

   • Idea Used: Expert system + predictive analytics.
   • How: The bank’s AI-powered loan approval system checks credit scores, income, and past behavior. It rejects 15% more high-risk applicants than manual reviews, saving Rs. 200M/year in defaults.

3. Daraz’s Demand Forecasting

   • Idea Used: Data-driven DSS + machine learning.
   • How: Daraz’s AI predicts stockouts 3 months in advance using sales data, weather, and festival calendars. This reduced out-of-stock items by 40% during Diwali 2023.

4. Pathao’s Dynamic Pricing

   • Idea Used: Reinforcement learning in DSS.
   • How: Pathao’s AI adjusts ride prices in real-time based on demand (e.g., 3x surge pricing during Dashain). This maximizes driver earnings while keeping users satisfied.

5. NTC’s Network Planning

   • Idea Used: Model-driven DSS.
   • How: NTC uses simulation models to predict where to lay fiber optic cables. For example, before expanding to Dharan, they ran scenarios like:
     • "If 50,000 users join, will current towers handle it?"
     • "What’s the ROI of adding a new tower?"
   • Result: 10% cost savings on infrastructure.

Exam Tip

This unit is highly conceptual but application-heavy. Expect:

1. Definitions: Be ready to explain KMS vs. DSS vs. expert systems clearly.

   • Example Question: "Differentiate between a data-driven DSS and a model-driven DSS with examples from Nepali companies."
   • Answer: Use the Daraz forecasting (data-driven) vs. Nabil Bank loan rules (model-driven) examples.

2. Diagrams: Draw knowledge cycles, DSS components, or expert system workflows in exams.

   • Example: Sketch the 5-step KMS cycle (create → store → share → apply → feedback).

3. Case Studies: Describe how a Nepali company uses KMS/DSS.

   • Example: "Explain how eSewa’s chatbot reduces customer service costs."
   • Structure:
     • Problem: High call volume.
     • Solution: KMS + AI chatbot.
     • Impact: 30% faster resolutions.

4. Ethical Scenarios: Discuss privacy, bias, or job displacement in KMS.

   • Example: "Should Nabil Bank replace loan officers with AI? Discuss ethical concerns."
   • Answer: Cover job loss, bias in data, and transparency (e.g., "Users should know if AI rejected their loan").

5. Worked Examples: Solve hypothetical DSS problems.

   • Example: "A Daraz warehouse has 100 units of Product X. Demand is predicted at 120 units next month. What action should the DSS suggest?"
   • Answer: "Order 20 more units (stockout risk) OR raise prices by 10% to reduce demand (if elasticity data is available)."

Common Mistakes to Avoid:

• Confusing KMS (knowledge storage) with DSS (decision aid).
• Ignoring real-world examples in answers (examiners love Nepali cases!).
• Overcomplicating diagrams (stick to 3–5 key steps).

Final Visual Summary:

mindmap
  root((Knowledge Management & Decision Support))
    KMS
      Document Management
      Expert Systems
      Collaboration Tools
    DSS
      Model-Driven
      Data-Driven
      AI-Enhanced
    AI in DSS
      ML Forecasting
      NLP Chatbots
      Computer Vision
    Ethical Issues
      Privacy
      Bias
      Job Impact
    Nepali Examples
      eSewa KMS
      Nabil Bank DSS
      Daraz AI Forecasting