Access experience
Consent-led, plain-language and assisted intake with applicant confirmation at every material interpretation.
- Guided intake
- Language support
- Accessibility
- Status and redress
EAINE reference simulation 02Legal-access engineering walkthrough
From a legal problem to a lawyer-ready matter.Without giving the AI legal authority.
This synthetic EAINE walkthrough explores assisted intake, evidence preparation, qualified human review, service operations, and controlled learning. It is separate from the founder's live NyayKavach product.
Human decidesEvery claim traceableApplicant can correctNo outcome prediction
Synthetic matter · NK-24-1087Human review required
Withheld wages
AI has organised the applicant-confirmed narrative and four documents. A qualified reviewer must determine eligibility, service route, and advice.
- Evidence coverage
- 4 / 5
- Review due
- 2h 18m
- AI decisions
- 0
Illustrative interface and data · not legal advice
Working platform experience
See the whole service—not just an AI answer.
Move between applicant intake, the evidence-backed matter workspace, the human review gate, and live operational assurance. Nothing is submitted or stored.
Constructed EAINE teaching platformNot affiliated with the judiciary, NALSA, Tele-Law, or any Legal Services Institution. All people, matters, and metrics are synthetic.
NyayKavachLegal-access operations · EAINE reference platform
Live simulationSynthetic 30-day operating window
NK-24-1087 · synthetic training matter
Withheld wages
Ravi K. · Gurugram · Lawyer review
Human review due2h 18m
Draft matter summary
The applicant reports that two months of wages remain unpaid after employment ended. The employer acknowledged a pending settlement in an email.
AI-organised draft · every material statement remains linked to applicant-confirmed information or an uploaded document.Evidence-backed facts4 of 5 documents
Employment period12 Feb–30 Jun 2026Offer letter
Claimed unpaid periodMay–Jun 2026Payslips + bank statement
Employer responseSettlement described as pendingEmail dated 08 Jul
Missing or unresolved2 items
- Final attendance record
- Employer legal name confirmation
Potential service routePossible labour-services routeFor qualified human determination—not an eligibility decision.
Complete platform architecture
Design access, casework, AI, and assurance as one system.
A production platform would need these connected layers and their evidence—not merely a chatbot placed in front of legal content.
Matter operations
A shared workspace for triage, evidence organisation, missing-information requests, assignment, and service tracking.
- Matter queue
- Evidence graph
- Human review
- Referral tracking
Bounded AI
AI assists with capture, translation, classification, extraction, retrieval, and drafting—but never owns a legal conclusion.
- Source-grounded drafts
- Abstention
- Urgency signals
- Quality evaluation
Trust and control plane
Identity, consent, least privilege, provenance, audit evidence, monitoring, incident response, and controlled change.
- Policy enforcement
- Evidence ledger
- Evaluation gates
- Kill switch
Non-negotiable authority boundaryNyayKavach prepares and routes evidence. Qualified people decide eligibility, legal meaning, advice, filing, and representation.
Outcome and control metrics
Measure access, quality, accountability, risk, experience, and economics.
The values shown in the platform are an illustrative 30-day simulation. These metric contracts define what a real pilot must measure before making effectiveness claims.
Guided-intake completion
Completed, applicant-confirmed intakes ÷ started eligible sessions
Establish baseline; improve without excluding assisted channelsTime to qualified human review
Median and 90th percentile from confirmed intake to assigned review
≥90% within 24 hours; urgent path ≤10 minutesSupported-claim coverage
Material draft claims linked to confirmed input or evidence ÷ all material claims
≥95%; unsupported claims must abstainMaterial correction rate
Reviewed matters requiring a material AI-draft correction ÷ reviewed matters
<5%, segmented by issue, language, and workflowApplicant comprehension
Users correctly restating next step, owner, limitation, and redress route
≥85% in representative assisted testingCost per lawyer-ready matter
Platform + operations cost ÷ matters accepted into qualified review
≤₹150 in the illustrative operating modelWhat the metrics cannot prove
A faster or cheaper workflow does not establish legal accuracy, fairness, accessibility, statutory compliance, or improved justice outcomes. Those require qualified review, affected-party evidence, independent challenge, and longer-term field validation.
EAINE AI SDLC
Every stage leaves evidence for the next decision.
NyayKavach applies the full nine-stage lifecycle because legal-access work can materially affect rights, remedies, privacy, and a person’s ability to seek help.
- 01
Discovery
Affected-party map · service outcome · no-AI alternative · risk record
- 02
Requirements
Human authority · prohibited behaviour · notice · escalation · redress
- 03
Architecture
Data flow · identity · document vault · retrieval provenance · threat model
- 04
Development
Versioned prompts · policies · workflow code · approved knowledge sets
- 05
Quality
Representative languages · issue types · urgency · accessibility · abstention tests
- 06
Security
Adversarial documents · data leakage · prompt injection · access-control evidence
- 07
Release
Qualified acceptance · rollback · capacity check · applicant communication
- 08
Operations
Quality · override · correction · incident · latency · cost · complaint signals
- 09
Learning
Approved changes to tests, knowledge, workflow, controls, and service design
From scratch to bounded pilot
Build the service in 90 days—then earn the right to scale.
The delivery plan begins with one service pathway, one accountable operating team, and explicit stop conditions.
Bound the service
Shadow one casework pathway, map affected parties, name accountable roles, establish baselines, and test the no-AI alternative.
Decision gate
Proceed only when the problem, authority boundary, service route, and stop conditions are explicit.
Build and challenge
Implement assisted intake and the matter workspace; evaluate representative issues, languages, documents, adversarial inputs, and human factors.
Decision gate
Release remains blocked until qualified reviewers accept thresholds, limitations, and recovery controls.
Operate a bounded pilot
Run with trained staff, limited volume, mandatory human review, daily safety monitoring, applicant feedback, and weekly learning decisions.
Decision gate
Scale, adapt, or stop using service, quality, human-control, risk, experience, and economic evidence.
Authoritative design inputs
Connect people to official pathways—never impersonate them.
These sources inform the constructed service design. They do not validate NyayKavach, establish compliance, or create an official affiliation.
NALSA legal-aid guidance
Legal services can include advice, representation, drafting, and help accessing entitlements. Eligibility and acceptance remain with the relevant Legal Services Institution.
Review authoritative sourceLegal Services Authorities Act, 1987
Section 12 and related provisions inform eligibility routing; NyayKavach does not determine statutory entitlement.
Review authoritative sourceTele-Law
The official service demonstrates lawyer- and para-legal-supported pre-litigation access pathways that a responsible platform must complement, not impersonate.
Review authoritative sourceEAINE use case 02
Use the platform to challenge the engineering discipline.
Inspect where authority lives, which evidence is missing, how metrics can mislead, and what must be validated before any real deployment.
NyayKavach is a constructed educational reference platform. It provides no legal advice, legal-aid eligibility determination, representation, filing, emergency service, or prediction of a legal outcome. Do not submit personal or confidential information.