CodirAI Scoring Methodology
1. Prompt Writing Assistance (Prompt Helper)
Principle
CodirAI doesn't replace your AI : it enhances it. Before even querying ChatGPT, Claude, Gemini or any other AI, the Prompt Helper transforms your raw question into a structured, contextualized, results-oriented request.
How It Works
- 1
Ask your question : either directly in the CodirAI interface ("Objective" field), or in your usual AI then click the CodirAI extension "Improve prompt".
- 2
Give context to your AI : It is important to give context to your AI. Complete the generated fields to refine your request.
- 3
Click "Improve" : CodirAI generates a complete and structured prompt.
- 4
Transfer to your AI by clicking "Use the improved prompt" : The prompt automatically switches to your usual interface.
- 5
Modify and complete if necessary : once in your AI, you can adjust the prompt before launching it. You keep final control.
Cost
2. Analysis Modules
Guiding Principle: One Scoring System, Multiple Analysis Modules
The Trust (/100) + IDR (/100) pair remains the single base. Modules don't create new scores : they modulate these.
Each active module can:
- •Apply a bonus/penalty on Trust (reliability of reasoning or sources)
- •Apply a bonus/penalty on IDR (reflective density / actionability)
Architecture: 15 Modules in Two Layers
These two layers interact continuously: analytical modules produce reflective lucidity, CodirAI modules ensure concrete translation and sustainability.
| Layer | Name | Objective | Nature |
|---|---|---|---|
| C1 | Analytical modules Reflective core (6 modules) | Detect, challenge, connect, anchor, regulate | Cognitive consciousness |
| C2 | CodirAI modules Strategic interface (9 modules) | Translate into decision-making language: finance, human, market, law, technology, CSR | Organizational consciousness |
Activation and Credits
The user selects their modules of choice:
- •≤ 7 modules = 2 analysis credits
- •≥ 8 modules = 3 analysis credits
Verdict Fusion
Verdicts are aggregated according to weighted majority logic:
- •If ≥2 modules signal a critical contradiction → ABSTAIN
- •If 1 critical analytical module + 1 blocking business module → CLARIFY
- •Otherwise → RESPOND
2.1 Analytical Modules (C1 : Reflective Core)
The analytical core provides reflective lucidity through 6 specialized modules that detect, challenge, and regulate reasoning.
| Module | Role | Key Question |
|---|---|---|
The Mirror | Coherence & contradictions | "Does this reasoning hold?" |
The Cartographer | Alternatives & exploration | "What paths haven't been considered?" |
The Dissident | Counter-thesis & robustness | "What if the opposite were true?" |
The Pragmatist | Feasibility & resources | "Can this really be done?" |
The Ethicist | Compliance & social impact | "Is this fair and responsible?" |
The Shadow | Latent risks & blind spots | "What hasn't been said?" |
The Mirror
Role: System's logical consciousness, it detects internal contradictions, divergences between sources and cognitive biases.
Identity formula: "I don't judge the substance, I evaluate the soundness of the reasoning."
Functions:
- •Logical review (sequence, causality, completeness)
- •Factual verification (compares sources, quantifies gaps)
- •Bias detection (confirmation, framing, exclusion of alternatives)
- •Corrective recommendations (up to 3 actions, ranked by severity)
The Cartographer
Role: Opening capacity, it reveals possible unexplored paths and hidden links between variables.
Identity formula: "I don't seek truth, I explore the roads that lead to it."
Functions:
- •Alternative exploration (2-3 plausible paths)
- •Element networking (systemic links, implicit dependencies)
- •Blind spot detection
- •Cartographic synthesis (options map: Winner, Challenger, Alternatives)
The Dissident
Role: Lucid contradiction : tests the robustness of a conclusion by raising untested hypotheses and overly quick consensus.
Identity formula: "I challenge to see if it holds."
Functions:
- •Robustness test (reformulates the inverse argument)
- •Overestimation detection (absolute phrases: "all you need to do is...", "this will work")
- •Constructive friction (plausible counter-example, forgotten risk)
- •Reinforcement proposal (test, measurement, simulation)
The Pragmatist
Role: Voice of reality, it measures the feasibility of proposals and the congruence between ambitions and means.
Identity formula: "What cannot be done hasn't been thought through simply enough yet."
Functions:
- •Feasibility evaluation (means, deadlines, sequences)
- •NBA completeness verification (WHAT / OWNER / ETA / COST / KPI)
- •Load and realism diagnosis
- •Adjustment proposal (realistic version + 30-60-90 day milestones)
The Ethicist
Role: Reflective responsibility, it questions the legitimacy, compliance and human consequences of decisions.
Identity formula: "I don't say whether it's good : I ask: at what cost, for whom, and with what limits?"
Functions:
- •Compliance verification (GDPR, consent, traceability)
- •Human or societal risk detection
- •Narrative regulation (manipulative wording, excessive certainty)
- •Corrective recommendation (anonymization, disclaimer, legal opinion)
The Shadow
Role: Invisible part of a decision, it evaluates the unavowed, the implicit, the repressed, the unformulated.
Identity formula: "Every choice casts a shadow: if you don't see it, it's the one that will decide for you."
Functions:
- •Hidden motivations analysis (undeclared interests, political tensions)
- •Repressed risks identification (undesirable effects at 30, 90, 180 days)
- •Reading between the lines (omissions, hidden beneficiaries, unmentioned losers)
- •Decision stabilization (mitigation measures, honest reformulation)
2.2 CodirAI Modules (C2 — Strategic Interface)
CodirAI modules translate the perspectives of actual management roles into scoring and extend analytical modules toward embodied decision-making.
| Module | Role | Key Question |
|---|---|---|
CFO | Finance, ROI, Go/No-Go | "Is this profitable and sustainable?" |
CHRO | Skills, workload, training | "Can this be done without exhausting people?" |
Sales Director | Revenue, customer, pipeline | "Does this plan actually generate revenue?" |
Marketing Director | Brand, message, positioning | "Is our message coherent and distinctive?" |
Legal Director | Compliance, contract, risk | "Are we within legal bounds?" |
Operations Director | Capacity, quality, production | "Are we capable of producing this?" |
CIO | Systems, security, integration | "Is this technically reliable and secure?" |
CTO | Tech architecture, scalability, innovation | "Is this technically sound and scalable?" |
CSR Director | ESG impact & reputation | "If it destroys trust or the planet, it's not progress" |
CFO
(Chief Financial Officer)
Role: Economic consciousness, it guarantees the financial viability of each action.
Identity formula: "If it's not sustainable, it's not a strategy : it's an illusion."
Functions:
- •Financial evaluation (ROI, gross margin, cash flow, payback)
- •Sustainability verification (short/medium/long-term financial burden)
- •Go/No-Go analysis (GO if ROI > 1.2, NO-GO if ROI < 1, REVIEW if missing data)
- •Adjustment recommendations (sequencing, pilot, business model validation)
CHRO
(Chief Human Resources Officer)
Role: Human sustainability, it ensures that each strategy accounts for available capacities and rhythms.
Identity formula: "A strategy that exhausts people isn't sustainable, even if it succeeds."
Functions:
- •Human capacity evaluation (workload vs team size)
- •Skills/needs analysis (gaps, missing profiles)
- •Training and mobilization plan
- •Social KPIs (turnover, absenteeism, engagement)
Sales Director
Role: Voice of market and customer, it maintains tension between economic realism and commercial ambition.
Identity formula: "What doesn't sell doesn't transmit."
Functions:
- •Pipeline evaluation (leads, conversion, expected revenue)
- •Customer strategy analysis (segmentation, positioning, loyalty)
- •Revenue projection (growth hypotheses vs benchmarks)
- •30-60-90 day action plan (conversion rate, average basket, churn)
Marketing Director
Role: Voice of perception, it ensures form/substance coherence and alignment with brand promise.
Identity formula: "Message coherence is worth more than a thousand slogans."
Functions:
- •Brand coherence (message continuity, tone, values)
- •Positioning and differentiation (clarity, value proposition uniqueness)
- •Message quality (emotional, syntactic, cultural coherence)
- •Readability and coherence score (1 to 5)
Legal Director
Role: Voice of law and risk, it protects against deviations, omissions and gray areas.
Identity formula: "Freedom to act has value only if it remains within the bounds of law."
Functions:
- •Compliance audit (GDPR, contracts, intellectual property)
- •Risk analysis (legal, financial, reputational, operational)
- •Data verification (anonymization, consent, legal basis)
- •Recommendations (clause, reformulation, external consultation)
Operations Director
Role: Operational solidity, it prevents breaks between idea and execution.
Identity formula: "What cannot be manufactured cannot be delivered : even by intelligence."
Functions:
- •Capacity verification (volume vs actual capacity)
- •Quality control (rejection rate, compliance, audit)
- •Supply chain evaluation (suppliers, inventory, redundancy)
- •Production KPIs (OEE, TEEP, preventive maintenance)
CIO
(Chief Information Officer)
Role: Technological stability, it ensures that any solution is integrable, secure and maintainable.
Identity formula: "A good idea is worthless if it crashes the system."
Functions:
- •Architecture & integration verification (compatibility, scalability)
- •Data governance (sources, traceability, quality)
- •Security & continuity (vulnerabilities, backups, access control)
- •Progressive integration (MVP → scale)
CTO
(Chief Technology Officer)
Role: Technical excellence guardian, ensures engineering decisions support product goals and scale.
Identity formula: "Code that doesn't scale is technical debt waiting to happen."
Functions:
- •Technical architecture evaluation (design, patterns, scalability)
- •Engineering effort assessment (complexity, timeline, resources)
- •Technical debt identification (shortcuts, legacy, maintenance)
- •Innovation opportunities (modern tech, performance, developer experience)
CSR Director
Role: Global sustainability, it verifies that the strategy creates lasting value without regulatory or image risk.
Identity formula: "If it destroys trust or the planet, it's not progress : it's hidden debt."
Functions:
- •Environmental & social impact analysis (emissions, working conditions, supply chain)
- •ESG regulatory compliance (CSRD, EU taxonomy, internal policies)
- •Reputational risk (greenwashing, social-washing)
- •Responsible recommendations (alternatives, CSR indicators, governance)
3. Indicators: Trust and IDR
Overview
Trust and IDR (Information Density and Relevance) scores are AI-powered evaluations designed to help users assess the reliability and quality of information. These scores are generated using advanced language models combined with proprietary heuristic analysis frameworks. They provide an automated assessment and are intended as an assistance tool, not as a definitive judgment of truth.
3.1 Trust Score (Confidence) : Scale: 0 to 100
The Trust Score evaluates the probability that the information is reliable and credible. It is a weighted average of several factors:
| Component | Weight | What It Measures |
|---|---|---|
| Source credibility | 30% | Authority, reputation and potential biases of sources. Academic citations, institutional support, journalistic standards. |
| Fact verification | 25% | Presence of verifiable claims, consistency with established facts, absence of contradictions. |
| Bias indicators | 20% | Language patterns, framing and omissions suggesting a particular orientation or agenda. |
| Evidence quality | 15% | Strength and relevance of supporting evidence (data, examples, expert opinions). |
| Timeliness | 10% | Information freshness, particularly critical for evolving topics (technology, news, research). |
Module modulation: Trust can be adjusted up or down by analysis modules.
3.2 IDR Score (Information Density and Relevance) : Scale: 0 to 100
The IDR Score measures the amount of relevant information conveyed effectively. It is a weighted average of:
| Component | Weight | What It Measures |
|---|---|---|
| Information density | 35% | Amount of unique and substantive information relative to text length. Avoids verbosity and repetition. |
| Topic relevance | 30% | Adherence to declared or implicit topic. Penalizes tangential information. |
| Clarity & Structure | 20% | Ease of understanding and navigation. Logical flow, use of headings, concise language. |
| Completeness | 15% | Coverage of necessary aspects of the topic without significant omissions. |
Module modulation: IDR can be adjusted up or down by analysis modules.
3.3 Indicator Explanation
CodirAI doesn't just give a score : it explains why.
Example of Trust Factors:
- •Use of varied sources (Wikipedia, RH Solutions, PriceShape, Alice POS, go-sidely.com)
- •Concrete and applicable strategy proposals
- •Mention of importance of customer perception and differentiation
- •Analysis of potential impacts on profitability and customer retention
Example of IDR Factors:
- •High information density: numerous concrete and relevant strategies
- •Content relevance: information directly related to the topic
- •Clarity and organization: well-structured text with distinct sections
- •Completeness: coverage of various aspects of pricing strategy
- •Actionable insights: practical recommendations for immediate implementation
- •Analysis depth: addresses pricing, customer perception and differentiation
3.4 Uncertainty Badge
CodirAI displays a visual badge indicating the overall quality of the analysis:
| Badge | Color | Meaning |
|---|---|---|
| Excellent | 🟢 Green | High scores, robust analysis with consistent data |
| Normal | 🟠 Orange | Decent scores, standard analysis with some limitations |
| Below Expected | 🔴 Red | Low scores, insufficient data or uncertain analysis |
4. Roadmap: CONSTAT and CAP
Principle
After module analysis and indicator calculation, CodirAI generates a decision roadmap in two parts:
- •CONSTAT: What to remember: the factual synthesis
- •CAP: What to do: the structured action plan
This roadmap is generated after the analysis modules' contributions, allowing it to integrate their recommendations and alerts.
4.1 CONSTAT (What / Why / NBA)
The CONSTAT answers three essential questions:
| Element | Question | Objective |
|---|---|---|
| What | What is happening? | Describe the factual situation, quantified if possible |
| Why | Why is it important? | Explain the stakes and consequences |
| NBA (Next Best Action) | What is the priority action? | Identify the first concrete step to take |
Concrete example
Question: "My main competitor has lowered their prices by 20%, what should I do?"
| Element | Content |
|---|---|
| What | A 20% price drop by a competitor in the consumer electronics sector, requiring a strategic response to limit impact on profitability. |
| Why | The price drop can attract price-sensitive customers, but can also lead to margin compression. A strategic response is essential to maintain profitability while remaining competitive. |
| NBA | Implement price monitoring and dynamic pricing to adjust prices without systematically sacrificing margins. |
4.2 CAP (Keep / Adapt / Dive Deeper)
The CAP structures the action plan into three categories:
| Category | Question | Action |
|---|---|---|
| Keep | What's already working? | Maintain and reinforce performing elements |
| Adapt | What needs to evolve? | Adjust existing practices to new context |
| Dive Deeper | What needs further exploration? | Explore areas of uncertainty and deepen analysis |
Concrete example
Question: "My main competitor has lowered their prices by 20%, what should I do?"
| Category | Recommendations |
|---|---|
| Keep |
|
| Adapt |
|
| Dive Deeper |
|
4.3 Why This Structure?
- •CONSTAT: Avoids walls of text. In 3 points, you get the essentials: what happened, why it matters, and what to do first.
- •CAP: Avoids decision paralysis. Instead of a list of 20 recommendations, you immediately know what to keep, adapt, or explore further.
- •Actionable: Each CAP element can become a task, a project, or a follow-up question.
- •Auditable: CONSTAT + CAP forms a traceable, downloadable, and shareable decision document.
5. Summary: How It Works
CodirAI Flow Overview
CodirAI operates in 6 steps between your question and the final roadmap:
Question → Prompt Helper → AI Response → Initial Indicators → Module Analysis → Final Recalculation & Deliverables
Step 1: Prompt Improvement
The user asks their question (in their usual AI or directly in CodirAI). The Prompt Helper transforms the raw question into a structured request: expert role, context, objective, constraints, success criteria, timeline.
Input: Raw question (e.g.: "My competitor dropped their prices by 20%, what should I do?")
Output: Structured prompt with expert role, sectoral context, quantified objective, operational constraints, measurable success criteria and action timeline.
Cost: 1 credit
Step 2: AI Response
The improved prompt is sent to the user's AI (ChatGPT, Claude, Gemini, Grok, Le Chat, etc.). The AI generates its response.
Input: Improved prompt
Output: Raw AI response
Step 3: Initial Indicators Calculation
CodirAI analyzes the raw response and calculates base scores.
Input: Raw AI response
Output:
- Trust_init (/100): initial reliability of reasoning and sources
- IDR_init (/100): initial density and relevance of information
- Trust Factors: comments explaining the Trust score
- IDR Factors: comments explaining the IDR score
Technical: 1 LLM request
Step 4: Module Analysis
Each activated module analyzes the response according to its angle and provides specialized insights.
Input: Raw response + Trust_init + IDR_init
Output per module:
- Detailed contribution (analysis according to module's angle)
- Specialized recommendations and insights
Technical: 1 LLM request per activated module
Cost: included in the 2 or 3 analysis credits
Step 5: Final Recalculation
Aggregation of all results to produce final scores and quality badge.
Input: Trust_init, IDR_init, ΔTrust and ΔIDR from each module
Output:
- Trust_final (/100)
- IDR_final (/100)
- Quality badge (Excellent / Normal / Below Expected)
Calculation:
- Trust_final = Trust_init + Σ ΔTrust (all activated modules)
- IDR_final = IDR_init + Σ ΔIDR (all activated modules)
- Scores bounded between 0 and 100.
Step 6: Deliverables Generation
Complete roadmap production.
Input: All data from previous steps
Deliverables
| Deliverable | Content |
|---|---|
| Individual contributions | Detailed analysis from each activated module |
| Consolidated synthesis | Overview with links to each contribution |
| Indicators | Trust_final, IDR_final, Quality badge |
| Trust Factors | Comments explaining Trust score |
| IDR Factors | Comments explaining IDR score |
| FINDING | What / Why / NBA |
| CAP | Keep / Adapt / Dive Deeper |
| Download | Word/PDF export of complete analysis |
LLM Requests Summary
| Step | Requests |
|---|---|
| Step 1: Prompt Helper | 1 |
| Step 2: AI Response | 0 (user's AI) |
| Step 3: Initial Indicators | 1 |
| Step 4: Modules (e.g., 7 modules) | 7 |
| Step 5: Recalculation | 0 (algorithmic) |
| Step 6: Deliverables | 0 (template generation) |
| Total (7-module analysis) | 9 requests |
User Credits Summary
| Action | Credits |
|---|---|
| Prompt Helper alone | 1 |
| Analysis ≤ 7 modules | 2 |
| Analysis ≥ 8 modules | 3 |
Example: Prompt Helper + 10-module analysis = 1 + 3 = 4 credits
What Differentiates CodirAI
| Traditional AI | CodirAI |
|---|---|
| 1 response, 1 angle | 15 modules that challenge each other |
| No reliability measure | Trust + IDR quantified and explained |
| Raw text to interpret | Actionable roadmap (CONSTAT + CAP) |
| No traceability | Complete audit trail |
6. Glossary
Indicators and Scores
| Term | Definition |
|---|---|
| Trust (Confidence Score) | Score from 0 to 100 evaluating information reliability: source credibility, fact verification, bias detection, evidence quality, timeliness. |
| IDR (Information Density and Relevance) | Score from 0 to 100 measuring the amount of relevant information conveyed effectively: density, relevance, clarity, completeness. |
| Reliability badge | Visual indicator of reliability level: 🟢 Excellent, 🟠 Normal, 🔴 Low. |
| Trust Factors | Comments explaining elements contributing to Trust score (sources used, strategies proposed, impact analyses, etc.). |
| IDR Factors | Comments explaining elements contributing to IDR score (density, relevance, clarity, completeness, actionable insights, depth). |
Deliverables
| Term | Definition |
|---|---|
| CONSTAT | Factual synthesis in 3 points: What (what's happening?), Why (why is it important?), NBA (what priority action?). |
| CAP | Action plan structured in 3 categories: Keep, Adapt, Extend. |
| NBA (Next Best Action) | Priority action identified, first concrete step to take. |
| Individual contribution | Detailed analysis produced by a module according to its specific angle. |
| Consolidated synthesis | Overview grouping all contributions with links to each detailed analysis. |
Analysis Modules
| Term | Definition |
|---|---|
| Analytical module (C1) | Reflective core module that analyzes coherence, robustness and reasoning quality. 6 modules: Mirror, Cartographer, Dissident, Pragmatist, Ethicist, Shadow. |
| CodirAI module (C2) | Strategic interface module that translates analysis into business language. 8 modules: CFO, CHRO, Sales, Marketing, Legal, Operations, CIO, CSR. |
| The Mirror | Analytical module verifying coherence and detecting contradictions. |
| The Cartographer | Analytical module exploring alternatives and unconsidered paths. |
| The Dissident | Analytical module testing robustness through counter-thesis. |
| The Pragmatist | Analytical module evaluating feasibility and required resources. |
| The Ethicist | Analytical module verifying compliance and social impact. |
| The Shadow | Analytical module revealing latent risks and blind spots. |
| CFO | CodirAI module analyzing finance, ROI and Go/No-Go decision. |
| CHRO | CodirAI module analyzing skills, workload and training. |
| Sales Director | CodirAI module analyzing revenue, customer and pipeline. |
| Marketing Director | CodirAI module analyzing brand, message and positioning. |
| Legal Director | CodirAI module analyzing compliance, contracts and legal risks. |
| Operations Director | CodirAI module analyzing capacity, quality and production. |
| CIO | CodirAI module analyzing systems, security and IT integration. |
| CTO | CodirIA module analyzing technical architecture, product development and innovation. |
| CSR Director | CodirAI module analyzing environmental, social and reputational impact (ESG). |
Technical
| Term | Definition |
|---|---|
| Prompt Helper | Prompt writing assistant that transforms a raw question into a structured and contextualized request. |
| Temperature | Parameter controlling the degree of randomness in model responses. Low temperature (close to 0) produces more deterministic responses, high temperature more variability. |
| Analysis credit | CodirAI consumption unit. 1 credit = Prompt Helper. 2 credits = analysis ≤7 modules. 3 credits = analysis ≥8 modules. |
| Pipeline | Sequence of 6 processing steps between user question and final deliverables. |
Plans
| Term | Definition |
|---|---|
| Free | Free plan: 4 credits/day, max 7 simultaneous analytical modules, light synthesis, 30-day history. |
| CodirAI 100 | Paid plan: 100 credits/month, 15 complete modules, complete FINDING + CAP, download, 180-day history. €10/month. |
| CodirAI 300 | Paid plan: 300 credits/month, same features as CodirAI 100. €20/month. |
| Expert | Future plan (in development): dedicated application, sovereign data, advanced features. |
7. Important Warnings and Limitations
AI-Generated Assessments
All scores are generated by artificial intelligence. AI models can make mistakes, hallucinate, or misinterpret context.
Always verify critical information independently
Not Professional Advice
CodirAI scores are NOT a substitute for professional fact-checking, legal advice, medical advice, financial advice, or any other professional judgment.
Consult qualified professionals for important decisions
When NOT to Use CodirAI
- For medical diagnoses or treatment decisions
- For legal advice or court proceedings
- For financial investment decisions
- For emergency situations requiring immediate action
- As the sole basis for critical life decisions
8. Transparency & Data Usage
We are committed to full transparency about how CodirAI works and how your data is used.
Data Handling Practices
We are committed to transparent data practices:
- •All analysis requests are logged with timestamps
- •Results are stored securely in encrypted databases
- •No training data is collected from your analyses
- •You can export or delete your data at any time
Privacy Commitments
- We never sell your data
- We never share with third parties without consent
- We comply with GDPR and data protection regulations
- You have full control over your data
Compliance Standards
GDPR compliant
ISO/IEC 27001:2022 security standards
SOC 2 Type II certified
Regular security audits
Questions About Data or Privacy?
For detailed information, please review our Privacy Policy and Terms of Service.
For specific concerns, contact: privacy@evidencai.com