We believe that key modalities of healthcare delivery and organization can, will and should change in the next 10 years. We are building the infrastructure that drives this change. A healthcare for every Indian that is data and evidence based, abundant, preventative and customer-centric.
For a 90-day sprint · 21 April 2026
Covers every material decision: Proposition, Build Posture, Commercial Model, Structure, and Strategic Boundaries. Each decision is tagged for timing: Sprint (close by end-July 2026 to make the plan fundable), Pilot (close by end-October 2026 with pilot design in hand), Post-pilot (close after six months of pilot signal), or Principal (MDA-level call made during the sprint, not inside it).
Timing legend: Sprint (July 2026) · Pilot (October 2026) · Post-pilot (H1 2027) · Principal (MDA call)
Who is the member, what are we selling, where do we start, and what clinical scope is inside the bundle.
| Decision | Options | Timing | Informing activities |
|---|---|---|---|
| Beachhead segment | Urban cardiometabolic 45-65 (₹3-4L) / Urban longevity 30-45 (₹75K-1.5L) / Employer group / Jamnagar captive / Sequenced two-step | Sprint | - Willingness-to-pay survey n=300+ across candidate segments - CAC estimation via Jio App enrolment test in two cities - Actuarial baseline for each segment from Jio Allianz - Retention and engagement benchmarks from comparable international programs (Oak Street, ChenMed etc) |
| Price architecture | Family floater bundled into Allianz premium / Standalone subscription / Employer capitation / Tiered (basic + premium) | Sprint | - Actuarial pricing co-developed with Jio Allianz - Conjoint analysis on bundle vs standalone preference - Competitive pricing scan (Niva Bupa, HDFC Ergo, Apollo 24/7 etc.) - IRDAI view on bundled add-ons to insurance premium |
| Product scope - what services sit inside the bundle | Core primary + chronic disease only / + mental health / + maternity / + fertility / + oncology navigation / full-service | Pilot | - Claims distribution analysis by specialty from Allianz and comparables - Clinical-risk assessment on out-of-scope handoffs with medical advisory - Bundle-preference conjoint at member level - Partner-availability review (oncology, fertility specialists) |
| Geographic sequence | Mumbai-first / Jamnagar-first + city-two in Y2 / Three-city parallel (Mum/Blr/Hyd) / Tier-2 test city | Sprint | - Claims density and NCD prevalence mapping by city - Jio user density and Jio Allianz GTM by city - Reliance retail footprint and hospital-partner quality by city - Operational readiness assessment (CHW hiring, clinical leadership pipeline) by city |
For each component of the operating system, what we own, what we partner for, what we refuse to do.
| Decision | Options | Timing | Informing activities |
|---|---|---|---|
| Physical footprint | Thin own (<15% of capex) / Franchise / Partner existing clinic-chain networks / Zero physical | Sprint | - Economic model for each option — capex, opex, per-member cost at 100K and 1M scale - Footfall and utilization economics from Jamnagar pilot pods - Franchise and partner-model benchmarks (Apollo, 1mg) |
| CHW cadre model | Salaried own-payroll / Gig platform / Third-party outsourced cadre / Hybrid · Ratio 1:500 to 1:2,000 | Sprint | - Clinical-touch requirement modelled against CHW skills and home-visit density - CHW or Clinical Concierge/Navigator? - Retention and protocol-adherence data from Jan Swasthya Sahyog, Piramal Swasthya, MAS India - Time-and-motion study with 20 CHWs - Cost modelling across employment models at three density ratios - Training cycle and certification pathway design - Identification of ideal profile and background for recruiting |
| Clinical AI stack | Jio Brain exclusive / Model-agnostic (Anthropic, OpenAI, open-weights) / Both in parallel | Sprint | - Safety benchmarking across providers - Fine tuning data needs and planning - Cost-per-inference modelling at 100K, 1M, 10M member scale - India data-residency and DPDPA review for each provider - Latency and availability SLAs across providers |
| Biobank infrastructure | Onshore centralized / Onshore distributed / Strand-hosted / Third-party biobank specialist | Sprint | - Vendor evaluation (Strand, Eurofins, purpose-built biobank operators) - LIMS benchmark and consent-platform compliance review under DPDPA - Sample chain-of-custody and retrospective-assay protocol design - International biobank architecture study (UK Biobank, FinnGen, All of Us) |
| Hospital network model | Preferred network at negotiated rates / Exclusive carve-outs / Agnostic navigation | Pilot | - 10-hospital MoU negotiations in pilot city to test rate-card leverage - Claims-leakage analysis by hospital tier - Legal review of exclusive-arrangement risk for navigation neutrality - Quality-signal data (outcomes, complications, readmissions) by partner hospital |
Where margin accrues, how money flows between insurer and care entity, how the Data Dividend is structured.
| Decision | Options | Timing | Informing activities |
|---|---|---|---|
| Which two P&Ls underwrite the plan | Insurer + Research / Insurer + Subscription / Research + Subscription / Insurer + Group-synergy | Sprint | - Bottom-up financial model for each P&L across three scenarios - Claims-savings sensitivity on the insurer P&L (MLR delta by lever) - Data Trust revenue modelled against UK Biobank, deCODE, 23andMe, Grail precedent shapes - 10-15 pharma R&D head conversations to size Indian willingness-to-pay |
| Insurer-Care money flow | PMPM management fee / Share of claims savings / Intervention-budget pass-through / Capitation | Sprint | - IRDAI precedent and regulatory review for each structure - International analogue review (Oak Street, ChenMed, Iora Health, Oscar) - Financial modelling of three variants at pilot and scale - Co-design sessions with Jio Allianz actuarial and finance teams |
| Data Dividend structure | Cash/credits to member / Premium discount / Equity-like unit in Data Trust / Mixed | Pilot | - Legal review of each structure under DPDPA and SEBI - Member-trust research across candidate mechanisms - Precedent review (All of Us return-of-value, Sage Bionetworks, 23andMe research credits) - Tax and accounting treatment analysis |
Legal entities, governance, leadership, and reporting lines inside Reliance.
| Decision | Options | Timing | Informing activities |
|---|---|---|---|
| Legal entity structure | Two (insurance + care) / Three (+ Data Trust) / Four (+ platform co) | Sprint | - Tax and regulatory review with external counsel - DPDPA data-fiduciary implications of each structure - Exit-optionality modelling (strategic, PE, IPO pathways) - Intercompany pricing and data-sharing framework design |
| Data Trust governance | Reliance-controlled board / Majority-independent board / Section 8 with Scientific Advisory Committee | Sprint | - Comparable governance study (UK Biobank, FinnGen, All of Us, deCODE) - Reputation and narrative-risk review with external communications counsel - Legal counsel drafting of trust deed and access policy - Advisory board composition plan |
| Jio Allianz relationship | Exclusive wrap / Anchor plus open to other insurers / Horizontal platform across insurers | Principal | - Allianz leadership dialogue — exclusivity, pricing, data access - Platform-vs-vertical strategic analysis with TAM and moat implications - IRDAI view on platform structure - Competitive analysis of other insurers’ positioning |
| OS owner inside Reliance | MDA office direct report / JFS / Jio Platforms / Standalone new vertical | Principal | - Reliance governance and cross-entity precedent review - Accountability mapping across insurance, care, data, research - MDA office consultation on reporting line and authority |
The explicit exclusions that sharpen the plan and pre-empt objection.
| Decision | Options | Timing | Informing activities |
|---|---|---|---|
| Data commercialization red line | Never sell identified data / De-identified only / Consented and aggregated only | Sprint | - Legal review under DPDPA of each position - Member-trust research on each position - Precedent review (23andMe policy, UK Biobank access rules, All of Us) - Policy document drafted for board and external publication |
| Insurance risk-bearing | Never / Selective capitation arrangements / Full insurer licence over time | Sprint | - IRDAI position confirmation - Capital-intensity modelling of risk-bearing - Strategic analysis — care entity as risk-taker vs risk-reducer |
| Rural entry timing | Never in five-year plan / Y4 onwards / Y2 pilot / Day-one urban-plus-rural | Pilot | - Urban unit-economics proof first, to be established in Jamnagar pilot - PMJAY partnership option review with state government(s) - Rural CHW operations assessment against existing models JSS etc. - Narrative risk review — Reliance Healthcare brand expectation |
| Pharma partner exclusivity | No exclusive in Y1-3 / Selective exclusive by asset class / Open to anchor exclusive | Post-pilot | - Y1 LOI range and exclusivity-premium data from pharma conversations - Legal review - Precedent review (deCODE-Amgen exclusive vs UK Biobank non-exclusive) |
Specification of the five systems — Underwriting*, Access, Escalation, Data, Science.
This document specifies Jio Arogya as a register of components. Each component is defined by a specification table describing its function, scope, inputs, outputs, exclusions, and build requirements. Summary matrices show how the components relate to one another. The Access, Underwriting, and Science systems follow a uniform specification form. The Escalation system uses a cross-tier matrix because the controlling question is how four contract types substitute for hospital ownership. The Data system uses instrumentation and signal registers because the controlling question is what is captured, how often, and what longitudinal record accumulates.
*Would need to be workshopped with the Jio Alliance team in collaboration.
| UNDERWRITING | ACCESS (90%) | NAVIGATION (10%) | DATA | SCIENCE |
|---|---|---|---|---|
| GTM through Insurance | How care reaches the member proactively | How member reaches care when needed | What we measure | What the data pays back |
| 1 · Embedded within insurance | 2 · Direct, on-demand at Home Care | 3 · Four-tier hospital partner network | 4 · Instrumented member-year | 5 · Data Trust dividend |
Underwriting:
Access:
Navigation:
Data:
Science:
Unit economics covered within Insurance Product
Investments repaid by Data and Science Assets
Bundled Insurance Product → Proactive, at home, on-demand care → Supported escalation journey → Instruments measure → Science compounds
Jio Allianz carries the risk. Jio Arogya delivers the care. Plans are co-designed.
Premium, reserves, claims, and grievance redressal sit on the Jio Allianz balance sheet under the IRDAI licence. Clinical care sits in Jio Arogya under its protocols and ethics board. Plans are priced jointly: the intervention budget for each plan is computed from the actuarial baseline and released to care delivery as the operating envelope. The two entities are legally separated. Data flow between them is DPDP-compliant, consent-ledger-tracked, and purpose-limited.
| Plan tier | Annual premium | Target loss ratio | Zero-marginal service floor | Intervention budget / member / year | Served cohort |
|---|---|---|---|---|---|
| Basic | ₹7,000 | 70% | ₹300 | ₹680 | Low-risk single adult |
| Core | ₹15,000 | 70% | ₹300 | ₹1,500 | Mass-market family floater |
| Standard | ₹30,000 | 70% | ₹300 | ₹3,800 | Urban middle-class individual or couple |
| Premium | ₹50,000 | 70% | ₹300 | ₹6,500 | Cardiometabolic or high-utilisation cohort |
| Comprehensive | ₹80,000 | 70% | ₹300 | ₹10,900 | Chronic-disease and research cohort |
Entity
Jio Allianz General Insurance Company Limited. IRDAI-licensed insurer.
Function
Relationship to care
Every plan is co-designed with Jio Arogya Health Services prior to filing with IRDAI. Neither entity prices a product the other has not agreed to deliver.
Risk
Absorbs actuarial and catastrophic risk on the insurer balance sheet.
Exclusions
Content
Register of at least thirty priced interventions. Each intervention tagged with a priced unit, expected outcome effect, and expected claims-cost elasticity.
Evidence standard
Published peer-reviewed evidence at launch. Internal claims-linked outcome data replaces external evidence within 24 months of launch.
Zero-marginal interventions (every plan)
Priced interventions (allocator-selected)
Exclusions
Build requirements
Function
Quarterly measurement cycle. Claims experience flows from Jio Allianz into the Intervention Budget Engine via a linked claims–care data feed. Each active intervention is repriced against observed claims-cost effect in the served cohort.
Decision rule
No intervention remains in the catalogue without demonstrated claims-cost effect within 18 months of launch. Interventions failing the rule are repriced once; a second failure removes the intervention from the catalogue.
Governance
Independent review by the Jio Allianz appointed actuary. Quarterly report submitted to the Jio Arogya board.
Exclusions
Build requirements
Entity structure
Jio Allianz (IRDAI-licensed insurer) and Jio Arogya (care-delivery entity) are legally separate. A master services agreement governs the care-delivery relationship.
Regulated flows
Data flows
Inter-entity data flows are DPDP-compliant, consent-ledger-tracked, and purpose-limited. Every cross-entity request carries a linked consent token.
Oversight
Exclusions
Three components reach the member. Engagement is ordered.
The access system contains three operating components: a per-member conversational interface running on the member device, a stateless central clinical inference service, and a distributed cadre of salaried Community Health Workers. Each component has a defined scope, deployment footprint, and handoff rule. Cross-component flow is governed by the Access Rule (Component 1.4).
| Component | Location | Service ratio | Holds PII | Clinical reasoning | Physical presence |
|---|---|---|---|---|---|
| A · Personal Agent | Member device + member cloud locker | 1 per member | Yes | No | No |
| B · Clinical AI | Jio Brain (centralised service) | Shared | No | Yes | No |
| C · Frontline Worker | Field / member home | 1 per 1000 members | No (routed via Agent) | Assisted by Clinical AI | Yes |
Function
Runs a persistent conversational interface on the member’s mobile device. It runs within the member’s trust boundary - this is the only component of the entire system that may touch identifiable data. Connected devices stream raw data to this agent, not to the cloud. Operates as the system-of-record for member identity, language preference, medication list, appointment history, consent ledger, and prior interaction log. Orchestrates downstream layers on the member’s behalf: issues queries to the Clinical AI, dispatches the Frontline Worker, files claims with Jio Allianz, retrieves records from hospitals and labs, and manages consent grants.
Deployment
Native integration on JioPhone. Jio MyHealth app on Android. One instance per enrolled member. Eight Indian languages, voice and text.
Inputs
Outputs
PII scope
Holds all member identifiable data. PII does not leave the device except to an encrypted per-member cloud locker under the member’s own key. Every cross-boundary request carries a linked consent token.
Exclusions
Build requirements
Function
Stateless clinical inference service. Returns differentials, protocol recommendations, screening flags, risk scores, prescription suggestions, and escalation triggers across the top fifty Indian condition categories. Services queries from the Personal Agent, the Frontline Worker’s copilot, and the physician CoPilot deployed at Tier-1 and Tier-2 hospital partners.
Deployment
Central service on Jio Brain inference infrastructure. Accessible via authenticated API.
Inputs
De-identified, consent-scoped queries containing structured member history and the presenting question.
Outputs
Availability
24×7. Eight Indian languages. Voice and text.
Autonomy posture
Human-in-the-loop for all prescriptions and all clinical escalations through Year 2 (?). Autonomous operation on a restricted task list thereafter, subject to published safety evaluation with HTA-In and ICMR and regulator concurrence.
Exclusions
Build requirements
Function
Salaried Front line Health Worker executing a defined protocol at each home visit. Operates as the primary physical layer in the access stack. Performs measurements, sample collection, medication reconciliation, dispensing where licensed, dietary recall, first-line counselling, and escalation to the Clinical AI.
Deployment
1 CHW per 1000 enrolled members. Approximately 100,000 CHWs at 100M member scale. ASHA cadre integrated under MoU with State Health Missions where desirable.
Visit cadence
Visits are at the family level, not just the individual level, which creates efficiencies, captures more data and better represents the Indian social dynamic and interconnected relationships
Visit protocol
Equipment
Point-of-care kit: pulse oximeter, otoscope, handheld ECG, peak-flow meter; portable ultrasound from Year 3. Estimated unit cost ₹12,000. Tablet with CHW copilot.
Inputs
Outputs
Exclusions
Build requirements
Statement
Layer 1 (Personal Agent) answers first. Layer 2 (Clinical AI) answers the subset of queries that Layer 1 escalates. Layer 3 (Frontline Worker) intervenes only on Layer 2 escalation or on a scheduled protocol visit. Member self-dispatch of the CHW is not permitted.
Logging
Every escalation carries a structured reason code. Every physical visit carries a clinical trigger recorded against the member’s longitudinal record.
Rationale
Audit
Rule enforcement is audited weekly by the actuarial loop (Component 3.4).
Four contract tiers substitute for hospital and clinic ownership.
Hospital and clinic ownership would solve four problems: control over the episode, cost of the stay, visibility into outcomes, and owning the patient relationship. However, owning and running hospitals is asset heavy, creates the wrong (volume) incentives and distracts from the real value add work. So we propose that instead of owning hospitals and large clinic chains, we substitute for four contract types. Almost every clinic and hospital in India is placed in one of four tiers. The matrix below specifies how each tier addresses each of the four problems; the specification tables that follow give contract-level detail.
| Mechanism | Tier 1 · Preferred | Tier 2 · Standard | Tier 3 · Small Hospital and Partner Clinics | Tier 4 · Non-Network |
|---|---|---|---|---|
| Share of network | 5% | 10% | 10% | Remainder |
| Share of volume | ≥60% | ~25% | ~10% | ≤5% |
| Contract form | Capitation or outcomes-based payment (within 24 months) | Case-rate with outcome adjustments | Pre-paid annual volume envelope | None — tariff at point of service |
| Control mechanism | Embedded clinical leadership; direct-discharge protocol | On-site member-services desk; contractually required protocols for top 15 conditions | None — influence via technology adoption (to be proven) | None — 24×7 member liaison only |
| Cost mechanism | Capitation fixes cost per member-year | Readmission penalties; never-event exclusions | Pre-payment priced below Tier-2 case-rate average | Settled at tariff through standard claims process |
| Visibility / Data mechanism | Two-way EHR integration; real-time clinical dashboards to CMO | HCX claims data; discharge summaries via on-site desk | Technology licensing writes structured data directly to Jio Arogya record | Liaison collects and uploads discharge documentation |
| Patient relationship | Personal Agent before, during, after; CHW on discharge | Personal Agent + on-site desk + post-discharge CHW | Personal Agent through stay; CHW post-discharge | Personal Agent + 24×7 liaison; CHW within 48 hours of discharge |
Scope
Deeply contracted anchor hospitals in each geography. HN Reliance, Seven Hills, Karkinos, plus one or two chosen peers per metro and tier-one city.
Contract form
Migration from fee-for-service to capitation or outcomes-based payment within 24 months of onboarding. Fee-for-service closed thereafter.
Integration
Instrumentation
Joint investments
Centres of Excellence — cardiometabolic, oncology, maternal-fetal, liver — jointly operated. CoEs double as clinical-trial and biobank sites. Research consent and sample collection are built into the admission workflow.
Volume commitment
Majority of the elective volume within each geography. Research enrolment at admission for consenting members.
Exit trigger
Two consecutive quarters below contractually defined outcome thresholds. Contract is renegotiated or the partner is down-tiered to Tier 2.
Scope
Mid-sized hospitals in every district.
Contract form
Case-rate contracts. Outcomes-based adjustments layered on top: readmission penalties, never-event exclusions, documented protocol adherence required for the top fifteen conditions.
Integration
Instrumentation
Volume commitment
15–30% of the partner’s admitted volume, rising with performance in successive contract cycles.
Exit trigger
Failure to meet clinical evidence thresholds over two review cycles. Status drops to Tier 3 and volume is routed elsewhere.
Scope
Long-tail 50–150-bed local hospitals serving Bharat. Chronic insurer-receivable cash-flow stress — 60 to 120 days of outstanding claims.
Contract form
Pre-paid annual volume envelope, advanced monthly, settled quarterly against actual utilisation. Economically equivalent to a prompt-payment discount, priced below the Tier-2 case-rate average.
Integration
Data flow
Every encounter writes structured data into the Jio Arogya record through the licensed technology stack. No clinical-protocol intrusion by the contract.
Volume commitment
Pre-paid envelope sized against historic throughput. Surplus utilisation reconciled quarterly.
Exit trigger
Scope
All Indian hospitals without a Jio Arogya contract. Reached via emergency admissions in unfamiliar locations, family preference, unreferred specialists, and small-town facilities outside the network.
Contract form
None. Claims settled at tariff through the standard claims process.
Patient-side services
Exclusions
No clinical control. No cost negotiation. No instrumentation.
Exit trigger
Not applicable. This tier is the permanent catchment for uncontracted episodes.
Operating principle. Jio Arogya does not operate hospitals. Jio Arogya operates the layer that knows the member, orchestrates the stay, captures the data, governs the outcome, and retains the relationship before, during, and after the admission. Hospital ownership is not the unit of control; contract design plus data integration plus patient-side services is.
What we capture, through what instruments, at what frequency, and what longitudinal record accumulates.
Intelligence has a shelf life. Static datasets devalue; trained models age; continuous instrumentation compounds. The data system is therefore specified in two registers. The Instrumentation Register (Component 4.1) enumerates every sensor and feed, what it captures, at what frequency, and whether it is existing or built. The Longitudinal Record (Components 4.2 and 4.3) specifies the dataset that accumulates per member per year and across the full five-year enrolment. Architecture, consent, and the Data Dividend are specified in Components 4.4 and 4.5.
| Instrument | Data captured | Frequency | Status |
|---|---|---|---|
| Jio MyHealth app on phone | Activity, steps, sleep pattern, voice cadence, medication reminders, consent-scoped location | Continuous, passive | Existing |
| Jio partner wearable | HR, HRV, SpO2, skin temperature, movement, sleep staging | Continuous | Existing; included in Premium and Comprehensive tiers |
| Continuous Glucose Monitor | Glucose trajectory, meal response, overnight pattern | Continuous, 14-day cycles | Existing; diabetic and pre-diabetic members |
| JioFiber router (ambient sensing) | Contact-free respiration, heart rate, fall detection, household rhythm | Continuous, ambient | Existing hardware; new software layer (scientific advancement needed) |
| Netmeds + Reliance Retail Pharmacy | Dispensing record, adherence, refill cadence | Per dispensation event | Existing |
| Reliance MedLab | Standard annual panels; specialty panels on indication | ≥1 per year | Existing |
| Strand multi-omic module | Whole-genome 30× depth; baseline proteomics; baseline metabolomics; 16S microbiome | Baseline at enrolment; 3-year refresh | Existing; research-scoped consent |
| Home air-quality sensor | PM2.5, PM10, CO, CO2, humidity | Continuous, passive | Build or Buy (~₹1,800 per unit; Premium and Comprehensive) |
| Home water-quality kit | Hardness, microbial load, heavy metals | Quarterly mail-in | Build or Buy (lab-processed) |
| Cellular BP cuff | Systolic, diastolic, pulse | Daily | Build or Buy (~₹2,400 per unit; hypertensive members) |
| On-device mental-health signal | Typing cadence, voice sentiment, sleep regularity, activity rhythm | Continuous, on-device inference | Build (only the risk score leaves the device) |
| Dietary recall tool | Voice-driven food diary scored against South Asian food database | Daily prompts; weekly minimum | Build (six regional languages) |
| CHW point-of-care kit | Pulse oximeter, otoscope, handheld ECG, peak-flow meter; portable ultrasound from Year 3 | Every CHW visit | Build (~₹12,000 per CHW) |
| EHR / HCX feed from hospital network | Admissions, diagnoses, prescriptions, procedures, discharge | Per encounter | Existing rails; live via tier integration |
The table below specifies what a single enrolled member produces in a single year of enrolment — the essential deliverable of the instrumentation stack.
| Signal class | Annual volume | Source instrument | Consent scope |
|---|---|---|---|
| Continuous physiology (HR, HRV, SpO2, respiration, activity) | ~8,760 hours | Wearable + JioFiber ambient | Member-standard |
| Sleep staging | ~365 nights | Wearable | Member-standard |
| Indoor air-quality exposure | ~525,000 minutes | Home air-quality sensor | Member-standard |
| Glucose (diabetic / pre-diabetic members) | ~35,000 readings | Continuous Glucose Monitor | Member-standard |
| Blood pressure (hypertensive members) | ≥365 readings | Cellular BP cuff | Member-standard |
| Mental-health score | ≥52 weekly scores | On-device signal | On-device, score-only |
| Dietary recall | ≥52 structured records | Voice recall tool | Member-standard |
| Water-quality readings | 4 quarterly samples | Mail-in kit | Member-standard |
| Lab panels | 1 standard + 1–3 specialty | Reliance MedLab | Member-standard |
| Pharmacy dispensations | Per refill event | Netmeds + Reliance Retail | Member-standard |
| Clinical encounters | Per encounter | EHR / HCX feed | Member-standard |
| CHW visit records | 2–12 visits | CHW point-of-care kit | Member-standard |
| Omics panels | 1 at enrolment; refresh on 3-year cycle | Strand multi-omic module | Research-scoped |
Scope
The cumulative dataset held on a Core member after five years of continuous enrolment.
Omics
Collected and biobanked samples allow for at least:
Continuous signals
Discrete readings
Clinical record
Human touch
Five years of structured CHW visit records, 10–60 total visits depending on risk tier.
Comparators
Privacy is enforced in infrastructure rather than in policy. The architecture consists of three layers across three legal entities, with three boundaries. Every cross-boundary event is logged and audited. The architecture is ABDM-native: ABHA ID is the member identity, the ABDM Consent Manager is the legal spine, the Health Claims Exchange is the claims rail, and Health Locker is the member’s portable record.
| Layer | Entity | Holds | Emits | Cannot |
|---|---|---|---|---|
| Layer 1 | Personal Agent on member device + per-member cloud locker | All member PII; consent ledger; medication list; appointment history | De-identified, consent-scoped queries to Layer 2; work orders to Layer 3 | Leave the member’s control without a logged consent event |
| Layer 2 | Jio Brain Clinical AI (stateless service) | No member state | Clinical recommendations and escalation triggers | Re-identify a query to a member |
| Layer 3 | Data Trust (separate legal entity, independent ethics board) | Research-consented, de-identified, linked longitudinal dataset | Aggregate study results; licensed research access under approved protocols | Sell raw data; re-link to identity; share data outside Indian jurisdiction |
Function
Returns research-licensing revenue to contributing members on a weighted basis. Each licensing event generates a distribution pool released to members whose data contributed to the licensed study.
Weighting factors
Economic impact
Exclusions
Build requirements
Continuous instrumentation generates the substrate. Science converts it into new medicine.
The Data Trust holds the research-consented dataset and operates the scientific programme. Four components define the programme: an autonomous research engine (5.1), a maturity-graded study ladder (5.2), a biobank of physical samples (5.3), and a governance structure holding the ethics, access, and licensing rules (5.4).
Function
Autonomous research system operating on the Data Trust’s de-identified dataset. Generates hypotheses, designs cohort studies, and runs primary analyses. Human scientists review, validate, replicate, and publish.
Operating principle
Compresses the research loop from the conventional five-year design-collect-publish cycle to a six-week query-validate-publish cycle against the live cohort. This is the operational answer to the shelf-life-of-intelligence thesis: research must run at the pace at which intelligence itself ages.
Inputs
Outputs
Exclusions
Build requirements
Function
Defines which class of research runs when, as the cohort matures. Observational work runs at 1M members. Pragmatic trials require 10M+. Biomarker and drug co-development requires multi-timepoint sample collection and linked omics.
Year 1–2 — Observational, pre-disease trajectory
Year 2–3 — Embedded pragmatic trials
Year 3–4 — Pharmacogenomic and biomarker studies
Year 4+ — Target discovery and drug co-development
Exclusions
No randomised output in Year 1. No drug-discovery promises against the Year-2 cohort. No pharma-sponsored work outside Trust governance.
Function
Physical-sample asset underpinning the research programme. Every enrolling member provides a baseline set of blood, urine, stool, and saliva samples under explicit research consent. Residual volume is cryo-preserved at −80°C for retrospective assays.
Sample collection
Standard assay panel
Storage
Two biobank sites — West India and South India — operated with Strand Life Sciences. Residual cryo-preservation at −80°C. Cold-chain collection network through CHWs and Tier-1 hospitals.
Target inventory
By Year 5: 250,000 multi-timepoint sample sets from a longitudinally characterised population.
Operational rationale
Collect low cost samples through the journey - store and label properly. Assay selected patients with clinical signal or outcomes under study. Any future assay not yet invented — single-cell transcriptomics, spatial proteomics, immune repertoire profiling, novel biomarker panels — can be retrospectively executed on stored samples.
Exclusions
Build requirements
Entity
Separate legal entity with its own profit-and-loss, ethics board, and public reporting.
Ethics board
Independent board with veto over all research activity. Public quarterly safety and access report. Member representation on the protocol committee.