Skip to main content
    Open Source Methodology

    The AI Oracle

    Your Macroscope for Hidden Business Intelligence

    Transform vast organizational data into actionable insights. Surface the overlooked risks and emergent opportunities that traditional analysis misses entirely.

    🔮 The Macroscope Promise

    In 60-90 minutes, discover what's been hiding in plain sight: the critical risk no one saw coming and the game-changing opportunity everyone missed.

    The Macroscope: A New Lens for Business Intelligence

    Microscope

    Traditional analytics zoom in on specific metrics and KPIs, missing the forest for the trees.

    Telescope

    Market research looks at distant trends and competitors, but can't see what's happening internally.

    Macroscope

    AI Oracle sessions reveal system-wide patterns across all your data, making the invisible visible.

    The Power of Legibility

    Your organization generates massive amounts of "digital exhaust" - customer emails, support tickets, meeting transcripts, Slack conversations. This data contains critical insights, but they remain unactivated because they're not legible to decision-makers.

    The Oracle methodology makes these patterns visible, affording you new options and fundamentally better decisions. Once you can see what was always there, you can act on it.

    What Success Looks Like

    Every Oracle session delivers concrete, high-impact outcomes with convexity

    One Overlooked Risk

    Previously invisible threat

    • Subtle shifts in customer sentiment signaling churn
    • Recurring bottlenecks indicating future failure points
    • Competitive threats emerging in the periphery

    Asymmetric Downside

    Mitigating prevents massive losses

    One Emergent Opportunity

    High-upside potential

    • New customer persona showing intense demand
    • Repeated feature requests pointing to new product lines
    • Internal processes ripe for high-ROI automation

    Asymmetric Upside

    Seizing could lead to exponential gains

    Beyond Individual Insights: Building Your Transformation Pipeline

    The session naturally uncovers a pipeline of AI implementation opportunities - from automating rote processes to creating new customer-facing tools. It becomes a catalyst for broader transformation.

    Workflow Opportunities

    When Oracle sessions reveal specific processes ripe for 10x improvement

    Systematic Transformation

    When insights reveal need for organization-wide AI adoption strategy

    Success = Clear, committed next steps with ownership and accountability.

    Understanding Convexity in Oracle Insights

    Why we focus on asymmetric risk/reward profiles

    Overlooked Risks

    📉

    Small Probability

    Often <5% chance

    💥

    Massive Impact

    Could destroy value

    🛡️

    Cheap to Mitigate

    Small investment prevents disaster

    Hidden Opportunities

    🎯

    High Probability

    Often >50% success rate

    🚀

    Exponential Upside

    10-100x returns possible

    First-Mover Advantage

    Competition hasn't seen it yet

    The Oracle focuses on insights with asymmetric payoffs - where the potential gain far outweighs the cost of action

    The Three-Phase Oracle Process

    Oracle Session Timeline

    Click each phase to explore the activities

    Phase 1: Data Preparation

    1-2 weeks
    • Inventory data sources
    • Capture executive context
    • Process & refine
    1

    Phase 1: Preparation & Data Curation

    1-2 weeks prior • Gathering the raw materials

    Step 1: Create Comprehensive Data Inventory

    🎯 Value Hierarchy Principle

    Private data from customer interactions yields the highest Oracle insights. Start here first - recorded calls, support tickets, customer emails contain unfiltered truth.

    🔥

    Private External (★★★★★)

    Customer calls, support tickets, emails - your proprietary gold mine

    📊

    Private Quantitative (★★★)

    ERP data, metrics, analytics - reliable ground truth for sizing

    🏢

    Private Internal (★★)

    Employee surveys, Slack, meetings - organizational context

    🌐

    Public Data (★)

    Reviews, industry reports - supplementary validation

    Note on Bias: Every data source has inherent bias. Support tickets skew negative, sales calls skew optimistic. Cross-functional experts contextualize these biases during analysis.

    Data Source Value Hierarchy

    Organized by value for Oracle insights - start with highest value sources

    Key Insight: Private data from customer interactions yields the highest-value insights. Each category builds on the others, but prioritize your time accordingly.

    Highest Value: Private External

    Private data from customer interactions

    Value:

    Your proprietary window into customer reality - this is where the gold is

    Data Sources:

    Recorded sales calls & demos
    Support tickets & chat logs
    Customer email threads
    Client meeting transcripts
    User onboarding sessions
    Customer success check-ins

    Why This Matters:

    High-fidelity, qualitative data that reveals what customers actually think, need, and struggle with

    Step 2: Capture Executive Context with Voice AI

    Deploy an AI voice assistant (e.g., Vapi) to interview each participant before the session:

    The Crystal Ball Question:

    "If you had a crystal ball that saw everything the company did for the past 60 days, what three questions would you ask it?"

    The Attention Question:

    "What's the most important thing happening that no one is paying enough attention to?"

    The Pre-Mortem:

    "It's one year from now and we've failed to hit our goals. What's the most likely reason?"

    Step 3: The Data Refinery Process

    🛢️

    The Crude Oil Analogy

    Raw data exports are like crude oil from the ground - valuable but unusable. You need to refine them into dense, high-octane fuel that AI can efficiently process.

    1. Consolidate

    Gather all raw exports into a single local directory

    • • Customer service exports
    • • Sales call transcripts
    • • Email threads
    • • Meeting recordings
    2. Sequence

    Create chronologically-ordered unified dataset

    • • Timestamp normalization
    • • Oldest to newest ordering
    • • Single JSON/CSV/TSV output
    • • Preserve interaction context
    3. Compress

    Maximize signal density for context windows

    • • Remove redundant metadata
    • • Preserve core facts
    • • Handle massive datasets
    • • Iterative compression
    Tactical Implementation with Claude Code
    Use Claude Code to inspect raw export structure and understand data schema
    Give plain-language instructions: "Take all data in this directory and normalize it"
    Iterate until you have one giant textual output ordered chronologically
    Handle context window limits through intelligent compression

    "Every system is different - this is always a bit annoying, but you just have to iterate until it's right."

    2

    Phase 2: Initial Analysis

    Asynchronous • Multi-agent reasoning

    Step 4: Conduct Multi-Agent Analysis

    Use Claude Code or similar to instantiate AI agents with distinct personalities:

    🔍The Explorer

    Insatiably curious, always looking for novel connections and patterns

    🤔The Skeptic

    Demands hard evidence for every claim, pushes back on assumptions

    🎯The Strategist

    Focused on "So what?" and "What can we do about this?"

    Setup: Create a shared "scratchpad" for agents to debate. Task them to identify the single most critical risk and opportunity. Let them use code for quantitative analysis where needed. Give them Gemini API key for additional tool calling if sophisticated.

    Step 5: Generate Initial Intelligence Report

    • Clean up multi-agent output into structured markdown report
    • Share with stakeholders ahead of time for comments
    • This becomes the anchor for live discussion
    3

    Phase 3: The Live Oracle Session

    60-90 minutes • Interactive discovery

    Step 6: Set the Stage

    Amazon-Style Reading: Set a timer for 5-20 minutes for silent reading if participants haven't pre-read the document. This ensures everyone has the same baseline understanding.

    Pro Tip:

    "The initial report is a starting point for debate, not a final verdict. This reading time is critical for productive discussion."

    Step 7: Facilitate Interactive Discussion

    Facilitation is an Art

    Critical Reality: Managing executives with strong personalities requires skill. This is very much an art, especially when insights challenge existing beliefs.

    • • Keep everyone feeling heard (executive context from Vapi helps)
    • • Manage time strictly - executives' time is precious
    • • Navigate disagreements professionally
    • • Push for concrete commitments, not just discussion
    Your Role as Facilitator
    • • Guide conversation flow
    • • Manage time and energy
    • • Ensure all voices heard
    • • Push for concrete actions
    • • Navigate executive dynamics
    Participant Role
    • • React to findings
    • • Challenge assumptions
    • • Provide missing context
    • • Commit to specific actions
    • • Battle-test insights

    Step 8: Use Large Context Window AI for Real-Time Inquiry

    The Live Oracle Toolkit

    1

    Google AI Studio with Gemini 2.5 Pro

    Pre-load entire dataset into context window

    2

    Fireflies.ai Live Transcript

    Capture questions and discussions in real-time

    3

    Interactive Loop

    Copy questions from transcript → Paste to Gemini 2.5 Pro in AI Studio → Share insights

    This highly interactive loop allows the team to battle-test insights in real-time, ask follow-ups, and dig deeper until satisfied.

    The Reality:

    "There's a lot of copying and pasting, there's a lot of swishing around between different windows" - This is the practical mechanics. You're constantly moving between Fireflies transcript, Google AI Studio, and your screen share. It's active, dynamic facilitation.

    Best Practices & Advanced Techniques

    The Art of Insightful Questioning

    To Surface the Unknown:

    "What is the most important thing in this data that no one is paying attention to?"

    To Make Insights Actionable:

    "What is one action that [Executive Name] can take in the next 60 seconds to get the ball rolling?"

    To Uncover Social Dynamics (Use with Care):

    "Perform a 'Straussian reading' of the internal communications. What are the esoteric power dynamics?"

    ⚠️ Can yield powerful but sometimes conspiratorial-sounding insights

    To Quantify Findings:

    "Can you size this opportunity? Based on the financial data, what's the potential quarterly impact?"

    Oracle Question Bank

    Click to copy powerful questions for your Oracle sessions

    To Surface the Unknown:

    "What is the most important thing in this data that no one is paying attention to?"

    To Make Insights Actionable:

    "What is one action that [Executive Name] can take in the next 60 seconds to get the ball rolling?"

    To Uncover Social Dynamics (Use with Care):

    "Perform a 'Straussian reading' of the internal communications. What are the esoteric power dynamics?"

    To Quantify Findings:

    "Can you size this opportunity? Based on the financial data, what's the potential quarterly impact?"

    To Find the Most Interesting:

    "What's the most interesting thing you see here that no one's paying attention to?"

    To Enable Tool Calling:

    "Use code to calculate this more rigorously - what's the quantified potential upside?"

    Technical & Operational Best Practices

    Note: This toolkit is always evolving. New MCP servers, connectors, and integration options appear regularly. Adapt these techniques as the landscape develops.

    Data Security

    • Always use .gitignore for sensitive data
    • Never upload client data to remote repos
    • Use local processing whenever possible

    Repository Sharing & Transparency

    • Create private repo with your multi-agent configurations and methodology
    • Share repo access with technical clients or transfer ownership
    • Upload agent configurations, Claude Code prompts, analysis workflows
    • Build trust through radical transparency of your methods

    Critical: Always ensure sensitive client data is in .gitignore before uploading anything to repositories. Share methodology, not data.

    Ensuring Lasting Impact

    Step 1: Close the Feedback Loop Immediately

    Ask each participant for a satisfaction score (0-5) and use counterfactual questions:

    For scores < 5:

    "What would have had to be true for this to be a [score + 1]?"

    For score = 5:

    "What was the most valuable aspect for you?"

    Step 2: Send Comprehensive Recap & Establish Accountability

    • Use AI to summarize Fireflies transcript within hours
    • Highlight key insights, decisions, and commitments (who/what/when)
    • Request permission to follow up on action items
    • Schedule check-in emails immediately (use Shortwave, Gmail, or your preferred client)

    Step 3: Transition from One-Off to Recurring Process

    Manual (0→1)

    Oracle analysis becomes standing input for weekly leadership meetings

    Refine (1→10)

    Run manually 2-3 times, gather feedback, perfect the format

    Automate (10→100)

    APIs and SDKs run analysis as recurring "cron job"

    This transforms the AI Oracle from a one-time intervention into a continuous source of strategic advantage, permanently integrated into decision-making.

    Reality Check:

    Sometimes insights get "explained away" by data that wasn't available to the model, or context that was missing. That's why follow-up is critical - you can't know the true value unless you close the loop.

    Ready to See What You've Been Missing?

    This methodology is open source because insight shouldn't be proprietary. Master it yourself, or work with our certified practitioners.

    Study the Method

    Everything you need is documented here. Start with your own data.

    Join the Community

    Connect with practitioners applying Oracle methodology worldwide.

    Book a Session

    Work with certified experts to surface your hidden insights.