Methodology & Framework Overview
The AI Risk Pulse is built on Purna Medha's SIRA (Synthetic Intelligence Risk Assessment) framework — a 7-layer model for measuring AI adoption risk across sectors. Below is a qualitative overview of the framework components.
1. Medha Grade Scale
7-tier rating system from AAA (equilibrium) to CCC (failure). Each grade reflects the balance between AI value creation and value destruction for a given sector.
| Grade | Label | Description | Color |
|---|---|---|---|
| Medha-AAA | Equilibrium | AI augments without dependency. Minimal value destruction. | #22c55e |
| Medha-AA | Healthy | Minor dependency risks. Value creation far exceeds destruction. | #4ade80 |
| Medha-A | Adequate | Monitor trends. Some emerging risks in 1-2 SIRA layers. | #86efac |
| Medha-BBB | Warning | Restructure needed. Value destruction measurable across layers. | #facc15 |
| Medha-BB | Stressed | Intervention required. Multiple SIRA layers under active stress. | #f97316 |
| Medha-B | Critical | Dissolve non-essential AI. Destruction outpacing creation. | #ef4444 |
| Medha-CCC | Failure | AI destroying more value than it creates. Immediate action. | #dc2626 |
2. Live Signal Event Types
Live RSS signals are classified by an LLM into event types. Each type adds upward pressure to the instrument's severity — signals can only escalate risk, never reduce it.
| Event Type | Label | Relative Impact |
|---|---|---|
| ⚠safety_departure | Safety Exit | Very high — institutional safety erosion |
| ◉outage | Outage | Moderate — infrastructure stress |
| △market | Market | Lowest — market noise (most common) |
| □layoff | Layoff | High — workforce destruction signal |
| ✕incident | Incident | Highest — direct harm or failure |
| ↺reversal | Reversal | Lower — corrective action (less alarming) |
Exact severity weights per event type are proprietary.
3. Tracked Instruments
9 sector-specific AI risk instruments, each with analyst-curated baselines covering grade, severity, layer exposure, and financial parameters. Live signals can only escalate from these baselines.
| Ticker | Name | Description |
|---|---|---|
| GLBL.AI | Global AI Ecosystem | Aggregate risk across all sectors of AI adoption worldwide. The broadest measure of systemic AI risk. |
| SAAS.AI | Enterprise SaaS AI | Companies adopting AI copilots and agents into productivity software — Copilot, Gemini Workspace, Claude for Work. |
| FIN.AI | Financial Services AI | Banks, trading firms, and fintechs deploying AI for risk modelling, algorithmic trading, fraud detection, and client servicing. |
| MED.AI | Medical & Healthcare AI | Clinical decision support, diagnostic imaging, drug discovery, EHR automation, surgical robotics, and mental health chatbots. AI where errors can kill. |
| CXOP.AI | Customer Operations AI | Support, success, and service teams replaced or augmented by AI agents. The Klarna pattern — replace humans, measure what breaks. |
| AUTO.AI | Autonomous Systems | Self-driving vehicles, warehouse robotics, drone delivery — AI operating in the physical world with safety-critical decisions. |
| DEVX.AI | Developer Productivity AI | Coding copilots, AI-assisted CI/CD, automated testing, and code review tools used by engineering teams. |
| LAW.AI | Legal & Compliance AI | Contract review, legal research, due diligence automation, compliance monitoring, e-discovery, and AI-drafted filings. Where hallucination is malpractice. |
| IND.AI | Heavy Industry & Manufacturing AI | Predictive maintenance, quality inspection, supply chain optimisation, digital twins, and industrial robotics across manufacturing, mining, energy, and construction. |
| GAME.AI | Entertainment & Gaming AI | AI in game development, procedural content generation, NPC behavior, streaming recommendation, music/video generation, and interactive media. Where creativity meets automation. |
Base grades, severity values, SPE scores, AI spend/employee, and per-layer weights are proprietary.
4. SIRA Layer Framework
The 7-layer Synthetic Intelligence Risk Assessment framework. Each instrument has unique exposure weights across all layers, driving the heatmap visualization on the Pulse dashboard.
| Layer | Name | Risk Surface |
|---|---|---|
| L7 | Human | Cognitive atrophy, deskilling, safety culture erosion |
| L6 | Workforce | Layoffs, role displacement, institutional knowledge loss |
| L5 | Market | Stock impact, competitive distortion, sector contagion |
| L4 | Vendor | Concentration risk, dependency lock-in, supply chain |
| L3 | Application | Hallucination, data leakage, integration failure |
| L2 | Infrastructure | GPU chip monopoly (Nvidia/TSMC), outages, compute dependency, latency spikes |
| L1 | Energy | AI data center power demand, grid strain, cooling costs, sustainability |
Per-instrument layer exposure weights and severity-to-grade boundaries are proprietary.
5. Regional Currency Coverage
Value at Risk (VaR) is computed in 8 currencies with region-specific salary benchmarks, exchange rates, and regional grade offsets that reflect local AI adoption maturity.
| Currency | Symbol | Region |
|---|---|---|
| USD | $ | United States |
| EUR | € | EU-27 median |
| GBP | £ | United Kingdom |
| INR | ₹ | India (Tier 1 cities) |
| JPY | ¥ | Japan |
| CHF | CHF | Switzerland |
| SGD | S$ | Singapore |
| AED | AED | UAE |
Salary benchmarks, exchange rates, and regional grade offsets are proprietary.
Proprietary Calibration Data
Exact numerical parameters — grade boundaries, VaR percentages, displacement rates, instrument severity baselines, SIRA layer weights, event severity weights, salary benchmarks, exchange rates, regional offsets, and SPE scoring formulas — are proprietary to Purna Medha.
These values are analyst-curated and periodically recalibrated. They represent the core intellectual property that drives the Medha Grade and Value at Risk computations on the Pulse dashboard.
Request a Methodology BriefingFramework overview last updated: Feb 2026 · Return to Pulse · Full Methodology