1. What Is Samkhya AI

Samkhya AI is an AI-powered, cross-source financial intelligence terminal designed for investment professionals, Chartered Accountants, analysts, and management consultants. It reads two primary corporate documents — the Annual Report / MD&A and the Earnings Call Transcript — and produces a single, integrated intelligence report that surfaces the gap between what management says and what their filings actually show.

The profession does not have a data problem — it has a contradiction-detection problem. Samkhya AI is built to solve exactly that.

The name Samkhya is drawn from the classical Indian philosophical school (Sāṃkhya Darśana) — one of the six orthodox schools of Hindu philosophy — built around the principle of clear-eyed enumeration and discrimination between what is real (Purusha) and what is illusion (Prakriti). This maps directly to the tool’s core function: separating substance from narrative in corporate disclosures.

2. The Problem It Solves

Corporate analysis today relies on reading two disconnected streams: the legally filed documents (annual reports, MD&A, risk factors, financial footnotes) and the verbal narrative (earnings calls, investor presentations, media statements). Each stream runs to dozens of pages. The real intelligence lies in the gap between them — where management’s optimism in the call diverges from the cautious language in the filings.

This gap is invisible to any single-source analysis tool. Traditional screeners look at numbers. NLP tools score sentiment on one document at a time. No tool in common use cross-references both streams simultaneously and highlights the divergence.

Metric What It Means

40–80 pages Average length of an annual report MD&A section

60–90 minutes Typical earnings call transcript length

3–5 hours Time a senior analyst spends reading and cross-referencing both per company

Invisible The say-vs-file sentiment gap, without structured cross-source analysis

3. How It Was BuiltY

3.1 Technology Stack

Component Technology / Library

Frontend Framework React (JSX) — single-file component architecture, rendered as a Claude Artifact

UI Library Tailwind CSS utility classes for responsive, professional styling

Icon Set Lucide React — lightweight, consistent iconography (30+ icons used)

AI Engine Anthropic Claude API (claude-sonnet model) — in-artifact API calls for real-time intelligence synthesis

PDF Generation jsPDF + jsPDF-AutoTable — for the banker memo export

State Management React Hooks (useState, useEffect) — zero external state libraries

Charting / Data Viz Custom SVG-based gauge bars, sensitivity grids, and colour-coded matrices

Hosting Claude Artifact environment (browser-rendered, no separate server)

3.2 Build Philosophy

Samkhya AI was built entirely within the Claude AI platform using the “Claude-in-Claude” (Claudeception) pattern: the user-facing artifact itself makes API calls to Claude’s completion endpoint to perform the cross-source synthesis. There is no separate backend server, no database, and no deployment pipeline. The entire application — UI, business logic, AI orchestration, and PDF generation — lives in a single React JSX file.

This architecture was deliberate. The goal was to demonstrate that a Chartered Accountant, without a software engineering team, can build a production-grade intelligence tool using only AI-assisted development — embodying the ICAI’s “CA with AI” vision as a workflow, not a slogan.

3.3 Version History

Version What Changed

v1.0 — SENTINEL Initial build. Dark-themed terminal UI. Two input panes (Annual Report, Earnings Call). Single-button cross-source synthesis. Output: structured intelligence report with contradiction matrix, KPI extraction, sentiment spread, and risk flags.

v2.0 — SENTINEL Added the Banker Recommendation Engine: DCF valuation calculator, multiples-based valuation, blended fair-value output, investment thesis generator, bull/bear case framework, catalyst identification, and earnings-call question playbook.

v2.1 — SAMKHYA AI Rebranded from SENTINEL to Samkhya AI. Added the Scenario Saver (Bull / Base / Bear slots with full input capture and one-click reload). Added the DCF Sensitivity Grid (5×5 WACC × Terminal Growth matrix, colour-coded against current price).

v2.2 — SAMKHYA AI Added the Generate Banker Memo feature: one-click PDF export bundling cover page, executive summary, say-vs-file sentiment spread, saved scenarios comparison, sensitivity grid, contradiction matrix, deep-dive divergences, and earnings-call playbook into a downloadable deliverable.

4. What Is Inside — Core Modules

4.1 Cross-Source Synthesis Engine

The heart of Samkhya AI. The user pastes two text inputs — the filed document (MD&A, risk factors, financial footnotes) and the verbal narrative (earnings call transcript). The engine sends both to the Claude API with a structured system prompt that forces extraction of:

• Company identification and financial headline

• Four key performance indicators (KPIs) with values and context

• Say-vs-File Sentiment Spread — a three-axis gauge (verbal tone, filed tone, alignment score) that quantifies the narrative gap

• Contradiction Matrix — topic-by-topic comparison of verbal claims versus legal disclosures, with variance analysis and severity flags (SEVERE / HIGH / MINOR)

• Deep-Dive Divergences — numbered, tagged, paragraph-length analysis of each material discrepancy

• Investment thesis (bull and bear case), catalyst identification, and risk assessment

• Earnings Call Playbook — AI-generated questions an analyst should ask management based on the contradictions surfaced

4.2 Banker Recommendation Engine

A full-featured valuation calculator embedded within the same terminal. Two tabs:

• DCF (Discounted Cash Flow): Revenue, growth rate, EBIT margin, WACC, terminal growth rate, shares outstanding, net debt — computes enterprise value, equity value, and fair value per share.

• Multiples: P/E, EV/EBITDA, and P/S multiples with peer-derived ranges — computes a multiples-implied fair value.

• Blended Output: Weighted average of DCF and multiples with implied upside/downside against current price and an auto-generated investment recommendation (Strong Buy / Buy / Hold / Sell / Strong Sell).

4.3 Scenario Saver

Three pre-named slots — Bull, Base, Bear — that capture the complete set of DCF and Multiples inputs at the moment of saving. Each slot displays the resulting DCF per-share, Multiples per-share, and Blended fair value. Users can reload any saved scenario with one click, iterate the inputs, and re-save. A “scenario loaded” indicator appears and disappears when inputs drift from the saved state.

4.4 DCF Sensitivity Grid

A classic 5×5 investment-banking sensitivity matrix. X-axis: WACC (base ±1pp, ±2pp). Y-axis: Terminal Growth (base ±0.75pp, ±1.5pp). Each cell shows the implied per-share fair value. Cells are colour-coded against the current market price: emerald for significant upside, slate for neutral, red for downside. The base-case cell is outlined for easy reference.

4.5 Generate Banker Memo (PDF Export)

One-click generation of a downloadable PDF that bundles every output into a single deliverable. The memo includes: a branded cover page with colour-coded rating ribbon, executive summary with sentiment gauges, KPI cards, bull/bear thesis blocks, saved scenario comparison table, DCF sensitivity grid (colour-coded), full contradiction matrix, deep-dive divergences, earnings-call playbook, and a professional disclaimer footer on every page.

5. The Name — Why “Samkhya”

The Sāṃkhya school of Indian philosophy is one of the oldest systems of rational enumeration in human intellectual history. Its foundational method is viveka — the disciplined discrimination between the real and the apparent. In the context of corporate analysis, this maps precisely to the tool’s purpose:

• What management says (the apparent / Prakriti) versus what the filings show (the real / Purusha).

• Enumeration of contradictions, not just sentiment scoring.

• Structured, numbered analysis — echoing the tradition of systematic categorisation that defines Samkhya philosophy.

The name signals that this is an Indian-origin intelligence tool, rooted in an intellectual tradition of analytical rigour that predates modern finance by millennia — now applied to contemporary corporate disclosure analysis using AI.

6. Target Users and Use Cases

User Segment How They Use Samkhya AI

Investment Analysts / PMS Pre-earnings and post-earnings cross-referencing; detecting narrative shifts quarter-over-quarter; building investment memos from raw filings.

Chartered Accountants Due diligence on listed companies; forensic analysis of management representations versus filed accounts; IPO / DRHP advisory where the say-vs-file gap is a regulatory red flag.

Management Consultants Competitive intelligence; sector scanning; client-facing deliverables that synthesise public corporate data.

PE / VC Analysts Target screening; management credibility assessment prior to investment committee presentation.

CA Students / CFA Candidates Learning to read filings critically; understanding how narrative and legal disclosure diverge in practice.

7. What We Are Planning to Build Next

7.1 Near-Term (Next 3 Months)

• Multi-Quarter Comparison: Upload 2–4 quarters of earnings call transcripts and track how management narrative shifts over time. Surface topics where the tone changed materially quarter-over-quarter.

• Peer Comparison Mode: Run Samkhya on two competing companies simultaneously. Output: a head-to-head credibility matrix showing which company’s narrative is more closely aligned with its filings.

• Indian Listed Company Integration: Pre-loaded templates for NSE/BSE annual reports and SEBI-format MD&A sections; auto-extraction of KPIs specific to Indian regulatory filings (including segment reporting under Ind AS 108).

• Voice Input: Allow users to paste or dictate the earnings call audio directly. Integrate Whisper-based transcription so that even non-transcribed calls can be analysed.

7.2 Medium-Term (6–12 Months)

• Regulatory Filing Parser: Direct upload of XBRL filings, DRHP PDFs, and SEBI annual return formats. Auto-extraction into the structured input panes without manual copy-paste.

• Historical Contradiction Database: Build a searchable archive of past analyses. Allow users to query: “Show me all companies where the SEVERE contradiction was in revenue recognition in the last 12 months.”

• Sector Dashboards: Pre-built templates for banking (NPA recognition vs. call commentary), pharma (pipeline claims vs. regulatory filings), and infrastructure (order book claims vs. segment revenue).

• Integration with Lex AI: Cross-link Samkhya’s corporate intelligence output with Lex AI’s judgment analysis. If a company’s disclosure contradiction involves a legal issue (e.g., contingent liability, pending litigation), auto-fetch relevant tribunal/court judgments for context.

• Team Collaboration: Multi-user workspace where analysts can share analyses, annotate contradictions, and build institutional knowledge.

7.3 Long-Term Vision

• Samkhya Intelligence Suite: A unified platform combining Samkhya AI (corporate disclosure analysis), Lex AI (legal judgment analysis), LexTax Advisory Engine (tax research and opinion generation), and NewsIntel (financial news intelligence). One login, one workspace, four specialised AI engines — designed for the Indian CA and legal professional.

• API for Institutional Use: Offer Samkhya as an API that institutional investors, brokerages, and research houses can embed into their existing workflows.

• SEBI / Regulatory Body Pilot: Propose Samkhya as a regulatory technology tool for market surveillance — helping regulators detect narrative-filing divergence at scale across the listed universe.

8. ICAI AI Hackathon Context

Samkhya AI was developed and presented at the ICAI AI Hackathon 2025. The hackathon invited Chartered Accountants across India to demonstrate practical AI use cases relevant to the profession. Samkhya AI was submitted as a working prototype demonstrating how a single CA, using Claude AI as both the development environment and the intelligence engine, can build a tool that would traditionally require a team of software engineers, data scientists, and domain experts.

The demonstration used Nexus Corp (a dummy company profile) to showcase the full workflow: pasting filed documents and call transcripts, running cross-source synthesis, navigating the contradiction matrix, building DCF and multiples valuations, saving Bull/Base/Bear scenarios, reviewing the sensitivity grid, and generating a downloadable banker memo — all within a single browser session.

The ICAI vision of “CA with AI” cannot remain a slogan — it must become a workflow. Samkhya AI is that workflow.

9. Closing Note

Samkhya AI represents a new category of AI-native professional tool: built by a practitioner, for practitioners, using AI as both the builder and the engine. It is not a theoretical prototype — it is a working terminal that produces deliverables (banker memos, contradiction matrices, valuation scenarios) that are immediately usable in professional practice.

The tool embodies the conviction that the next generation of professional tools in India will be built not by technology companies alone, but by domain experts — Chartered Accountants, lawyers, consultants — who understand the problem deeply enough to architect the right solution. Samkhya AI is proof of concept for that thesis.