Session 2: AI for Research, Analysis, and Decision Support

IIM Kozhikode Senior Management Programme - Batch 16

Aditya V. Jain

January 2026

Quick Recap: Session 1 Foundations

LLMs are 'Brilliant Interns' — encyclopedic knowledge, 100x speed, but no judgment
Three cognitive deficits: Hallucination, No Real-Time Knowledge, No Persistent Memory
Context engineering > Elaborate prompting
The skill isn't writing prompts — it's getting AI to extract context from you

Key Insight: Today: We solve the 'No Real-Time Knowledge' and 'No Persistent Memory' problems

Today's Journey: Two Superpowers

🔍 Deep Research

The World's Information → Your Intelligence

Research that would take days, done in minutes
Synthesizes hundreds of sources
Creates actionable reports, not just summaries

📚 NotebookLM

Your Documents → Queryable Knowledge

Your organization's memory, made intelligent
No more 'I know we discussed this somewhere...'
Transforms static PDFs into dynamic conversations

The 'Google Test' for AI Value

Question: If you Google the first 10 words of your query, would you get a useful answer?

Query	Verdict	AI Value
What is Quick Commerce?	❌ Google handles this fine	Low
How should Perfetti Van Melle India respond to Blinkit's dominance given their ₹1 price point dependency?	✅ Google can't answer this	High

Principle: AI earns its keep when it synthesizes, not when it retrieves

Part 1

Know Your Organization

Making Internal Documents Intelligent

The Organizational Memory Problem

Scenario: You've been at your organization for 15+ years. You've seen strategies evolve, pivots happen, lessons learned...

New leaders ask questions you answered 3 years ago
Strategic decisions get re-debated without institutional context
Annual reports become write-once-read-never artifacts
Contracts sit in folders until there's a dispute

Cost: Organizations don't learn — individuals do. And individuals leave.

NotebookLM: Your Organization's Memory

What it is: Upload documents → Ask questions → Get answers with citations

Grounds responses in YOUR documents only (no hallucination from training data)
Cites specific sources for every claim
Creates study guides, FAQs, timelines automatically
Generates 'Audio Overviews' — podcast-style discussions of your content

Crucial difference: Unlike ChatGPT: It won't make things up. It only knows what you upload.

Live Demo

Plan India NotebookLM

Organization History & Evolution Analysis

Demo 1: 17 Years of History

The Setup

Context: Plan India — child-rights NGO

Documents: 17 years of annual reports (2008-2025)

Challenge: New executive joining needs to understand organizational evolution, strategic pivots, program changes

What We'll Ask

How has the strategic focus evolved from 2008 to 2025?
What were the major programmatic pivots and why?
How has the funding mix changed over time?
What lessons from past initiatives inform current strategy?

Live Demo

EPC Contract NotebookLM

Risk Analysis & Clause Interpretation

Demo 2: Contract Intelligence

The Setup

Context: Infrastructure/EPC Contracts

Documents: MoRTH Model EPC Agreement (2012) + Risk Analysis

Challenge: Before signing, identify contractor-unfriendly clauses that could cost crores

What We'll Ask

What are the top 5 risks for contractors in this agreement?
Where does the 'foreseeability clause' create exposure?
What are the strict time-bars that could forfeit claims?
How does risk allocation compare to industry norms?

The NotebookLM Pattern

1. Gather: Collect documents that represent institutional knowledge (reports, contracts, research, transcripts)
2. Upload: Add to NotebookLM (up to 50 sources, 500K words total)
3. Query: Ask questions as if talking to someone who read everything
4. Verify: Every answer cites sources — click to verify
5. Share: Generate Audio Overviews, study guides, or FAQs for team distribution

Part 2

Know Your Market

External Intelligence at Scale

The Research Problem

Scenario: You need to understand a market, a competitor, a prospect, or a regulatory landscape...

Old Way: Google search → 10 blue links → open tabs → read → synthesize → repeat
Or: Commission expensive consulting reports
Or: Ask a junior analyst to 'do some research'

Time Cost: A thorough research brief = 2-4 hours minimum, often days

Quality Issue: Results depend heavily on researcher's domain knowledge

Gemini Deep Research: Superhuman Speed

Research that would take days, done in minutes.

Takes your research question
Creates a multi-step research plan
Searches hundreds of sources (news, reports, filings, articles)
Reads and synthesizes across sources
Produces a structured report with citations

Output: A 5-15 page research report in 5-10 minutes

Live Demo

HDFC Bank Proposal Research

Deep Research for Sales Intelligence

Demo 4: Sales Proposal Intelligence

The Setup

Context: IT Services / Consulting

Challenge: Atos Eviden pursuing HDFC Bank for AI/GenAI services

Need: Understand HDFC's IT strategy, pain points, and create tailored proposal

The Research Question

"Research HDFC Bank's technology strategy, digital transformation initiatives, and IT challenges from their annual reports and investor calls. Then match Atos Eviden's AI/GenAI capabilities to their specific needs and create a proposal framework."

Live Demo

Labor Solutions Strategy Research

Market Dynamics & Regulatory Analysis

Demo 5: Strategic Positioning

The Setup

Context: Social Impact Tech / B2B

Company: Labor Solutions — worker voice technology

Challenge: New EU regulations (CSDDD) changing market dynamics; need strategic pivot analysis

The Research Output

Identified EcoVadis-Ulula merger as primary threat
Recommended pivot from 'worker voice' to 'market access insurance'
Mapped Tier 1 mega-supplier targets
Defined differentiation strategy vs. ESG aggregators

The Deep Research Pattern

Prospect/client research before important meetings
Competitive intelligence and market analysis
Regulatory landscape understanding
Due diligence and background research
Strategic planning inputs

Good vs Bad

Weak: "What is HDFC Bank's revenue?"

Strong: "What are HDFC Bank's stated technology priorities and where do they face implementation challenges?"

30 Minutes

☕ Break

When we return: Technical ROI, Structured Thinking, and Your Action Plan

Part 3

Accelerate Your Team

Technical Decisions & Structured Thinking

Live Demo

PostgreSQL Upgrade NotebookLM

Technical Analysis & Business Case Generation

Demo 6: Technical ROI

The Setup

Context: Engineering / IT Leadership

Challenge: PostgreSQL 16 → 18 upgrade decision

Documents: PostgreSQL 16, 17, 18 documentation + release notes

What We'll Ask

What are the breaking changes between versions?
What performance improvements justify the migration cost?
What's the business case I can present to the CFO?
What are the risks and rollback considerations?

Beyond Features: Shaping Behavior

The demos so far used AI tools as-is. But you can SHAPE how AI thinks.

Concept: Custom instructions (Gemini Gems, GPTs, Claude Projects) let you embed methodologies.

Example: Instead of asking AI to brainstorm, you can create an AI that follows a specific brainstorming methodology.

Live Demo

Adaptive Inquiry Gem

Structured Brainstorming & Idea Stress-Testing

Demo 7: The Adaptive Inquiry

The Methodology

Gödelian Inquiry: Every idea has unstated assumptions — surface them
Darwinian Ideation: Ideas must survive in an ecosystem — test their fitness
Stage 1: Axiomatic Deconstruction
Stage 2: Ecosystem & Fitness Scan

The Implementation

Tool: Gemini Gem with custom instructions

Behavior: AI doesn't answer "How do I build X?" — it guides you through rigorous idea refinement

Output: A much more robust, stress-tested project plan

Part 4

Choosing the Right Tool

A Decision Framework

The Tool Selection Matrix

📚 NotebookLM

Use when:

You have specific documents
You need answers grounded in YOUR content
You want citations you can verify

Examples:

Contract analysis, Policy interpretation, Historical research

🔍 Deep Research

Use when:

You need external intelligence
Research would take hours manually
You need synthesis across many sources

Examples:

Market analysis, Competitor intelligence, Prospect research

🧠 Gems/Custom GPTs

Use when:

You have a repeatable methodology
You want consistent AI behavior
You're embedding expertise

Examples:

Structured brainstorming, Interview frameworks, Coaching

The Combination Play

Deep Research → NotebookLM

Use Deep Research to gather external intelligence
Download/save the research output
Upload to NotebookLM with your internal documents
Query across both external research AND internal context

Example: Market research + strategy docs

NotebookLM → Audio Overview

Upload complex documents to NotebookLM
Generate Audio Overview with specific framing
Share podcast with team members who won't read the documents
Use as onboarding or briefing material

Example: "Hidden Traps" podcast for team

Part 5

Your Action Plan

Starting Tomorrow

This Week: Three Experiments

Tomorrow: NotebookLM Experiment
Upload one document you reference frequently (a policy, a contract, a strategy doc). Ask it 5 questions. See what it surfaces that you'd forgotten.
Day 2-3: Deep Research Experiment
Think of a prospect meeting, strategic decision, or competitive question. Run Deep Research. Compare output to what you would have done manually.
Day 4-5: Audio Overview Experiment
Create a NotebookLM notebook. Generate an Audio Overview. Share with one colleague and get their reaction.

Building Your AI Toolkit

Free Tools (Start Here)

NotebookLM notebooklm.google.com (Free, generous limits)

Gemini gemini.google.com (Free tier available)

ChatGPT chatgpt.com (Free tier available)

Claude claude.ai (Free tier available)

Paid Upgrades (When Ready)

Gemini Advanced Deep Research, 1M token context (~₹1,650/month)

ChatGPT Plus GPT-4, DALL-E, Advanced Data Analysis (~₹1,650/month)

Claude Pro Higher limits, Projects (~₹1,650/month)

The Mindset Shift

From AI as search engine → To AI as research analyst
From Documents as storage → To Documents as queryable knowledge
From One-off queries → To Accumulated context
From Generic AI → To AI shaped for your methodology

The value isn't in the AI — it's in the context you bring to it.

Session 2 Summary

NotebookLM: Makes your documents intelligent (Demos: Plan India, EPC Contract, PostgreSQL)
Gemini Deep Research: Research that would take days, done in minutes (Demos: HDFC proposal, Labor Solutions)
Gemini Gems: Shape AI to follow your methodology (Demo: Adaptive Inquiry)

You now have tools to solve the 'No Real-Time Knowledge' and 'No Persistent Memory' problems.

Q&A

Questions, Clarifications, and Discussion

15-20 minutes

Looking Ahead: Session 3

AI and the Future of Work

The economic impact of AI on jobs and industries
The 'Reshuffle' — how work gets reorganized, not eliminated
Building AI-first workflows in your organization
Managing teams in the age of AI augmentation

Preparation: Think about: What tasks in your role could be augmented (not replaced) by AI?

Thank You

Aditya V. Jain

Links to tools and further reading will be shared

"The best time to start was yesterday. The second best time is tomorrow morning with NotebookLM."