AI for 108 Ambulance Emergency Triage: Fixing India's Pre-Hospital Crisis

Executive Summary

India's 108 emergency ambulance service is the world's largest pre-hospital emergency response network — and one of its most audited failures. Operated across 16 states under the National Health Mission's PPP model, the system answers approximately 86,000 calls daily out of 250,000 attempts, dispatches ambulances to 35,780 emergencies, and has cumulatively served 11.6 crore emergencies since inception. The performance record is more ambiguous. Karnataka's Comptroller and Auditor General (CAG) found that 44% of calls were non-emergency, 64% of responses were classified as ineffective, and only 3% of patients received follow-up callbacks. In Odisha, 59% of emergencies missed their response-time targets. In Maharashtra, average ambulance response time was 134.5 minutes. In Punjab, a January 2023 labour strike shut the service for six days.

Executive Callout The 108 ambulance service receives approximately 250,000 daily calls across 16 states but answers only 86,000. Karnataka's CAG audit (2020) found 44% of calls were non-emergency and 64% of responses were ineffective. Labour strikes disrupted service in Punjab (6 days), Rajasthan (21 days), and UP (10,000 workers terminated). Karnataka launched a fully government-owned command centre on C-DAC's NG-ERSS V2.0 platform in May 2026 — the most procurement-receptive state for an AI triage pilot. Bhashini's CONVERSE capability is already deployed for UP Police 112, providing a proven template for multilingual emergency voice response. The addressable market for 108 call centre operations alone is Rs 200–400 crore annually. (Aisewak Government Helpline Report, 2026, citing CAG Karnataka Audit 2014–19, CAG Odisha Audit 2013–14, GVK EMRI annual data, C-DAC ERSS Phase II documentation.)

The structural problem with 108 is not a resource gap — it is a signal-to-noise failure compounded by workforce fragility. Almost half of all calls should not reach a human emergency dispatcher. When they do, they consume capacity that should be available for genuine life-threatening emergencies. AI voice triage addresses this directly: classify calls before human dispatcher intervention, filter non-emergencies to appropriate services, and ensure genuine emergencies receive priority dispatch. This article sets out the evidence base, the AI solution architecture, the implementation roadmap, and the procurement pathways that health officials and state NHM directors need to act on.

Introduction

When a citizen dials 108 with a cardiac emergency, the next 90 seconds determine whether they survive. Pre-hospital care research consistently identifies the first hour after a medical emergency — the "golden hour" — as the period in which the difference between timely and delayed ambulance response is most consequential. Yet the system that governs those 90 seconds in India is, by its own auditors' assessment, functioning at a fraction of its intended capacity.

The 108 service is operated primarily by two private entities — GVK EMRI (13,000+ ambulances across 16 states) and Ziqitza Healthcare (3,300+ ambulances across 7 states) — under government PPP contracts funded by National Health Mission state programmes. The Karnataka government's decision in May 2026 to terminate GVK EMRI's contract and launch a fully government-owned 50-seat Central Command and Control Centre on C-DAC's NG-ERSS V2.0 platform marks a structural inflection point. For the first time, a major state is running its emergency medical response through government infrastructure — and that infrastructure explicitly supports AI integration. (Aisewak Government Helpline Report, 2026.)

The AI opportunity is not speculative. Bhashini's CONVERSE capability — multilingual voice AI — is already deployed in the UP Police 112 system. The technical template exists. What remains is adaptation for medical triage, procurement execution, and state-by-state deployment.

Current Challenges: A System Overwhelmed by Its Own Design

The Non-Emergency Call Flood

The single most documented operational problem in India's 108 ecosystem is call misclassification at the point of entry. Citizens call 108 for non-urgent transport requests, health information queries, follow-up on prior cases, and mistaken connections. This is not user error — it reflects a public that understands 108 as the single health contact number, not a strictly emergency line.

The consequence, documented by the CAG for Karnataka (covering 2014–2019), is systematic capacity diversion: 44% of calls received by the Karnataka 108 command centre were classified as non-emergency. (Aisewak Government Helpline Report, 2026, citing CAG Karnataka Audit 2014–19.) Of the remaining 56% that were classified as emergencies, 64% of responses were categorised as ineffective — representing wrong ambulance type, delayed arrival, inadequate care on scene, or failure to notify the receiving hospital. Only 3% of patients received any follow-up callback after their case was closed.

State	CAG Finding	Severity
Karnataka	44% non-emergency calls; 64% ineffective responses; 3% callback rate	Critical
Odisha	59% of emergencies missed response-time targets (30 min rural, 20 min urban)	Critical
Maharashtra	Average response time 134.5 minutes — among the worst documented	Critical
Kerala	54.48% of emergencies exceeded the 10-minute response target	High
Punjab	86% vehicle shortfall; 6–7 day labour strike (January 2023); no SOPs after nine years	Critical

Source: Aisewak Government Helpline Report, 2026, citing CAG audit reports for respective states.

The Workforce Fragility Problem

India's 108 service has a second structural weakness that receives less policy attention than performance metrics: its dependency on a workforce operating under chronic labour stress.

In January 2023, Punjab's 108 service was shut for six days after staff struck work over unpaid wages. In September 2023, Rajasthan's service faced a 21-day strike. In Uttar Pradesh, the state government terminated 10,000 108 workers during a restructuring exercise, triggering violent protests. Kerala's government accumulated Rs 100 crore in arrears owed to GVK EMRI, threatening contract continuity. (Aisewak Government Helpline Report, 2026, citing state news archives.)

Each strike creates a window where the state has no emergency medical dispatch capability for millions of citizens. These disruptions are not freak events — they are predictable outcomes of a PPP model where cost pressures transfer to frontline salaries. AI voice triage does not replace paramedics or EMTs. But it does create a resilience layer for call dispatch operations: a 24/7 automated triage system that can handle classification, routing, and caller communication regardless of human workforce disruption.

The Response Time Gap and Its Consequences

India's national EMS standard sets a target of 20 minutes for urban emergencies and 30 minutes for rural. CAG data shows systematic failure against these targets across multiple states. Maharashtra's 134.5-minute average response time is not an outlier caused by geography — it is the product of a dispatch system that cannot distinguish between a cardiac arrest requiring immediate response and a non-emergency call requesting routine transport.

Pre-hospital care research consistently identifies early response as the dominant variable in trauma survival outcomes, particularly for road accidents (India's leading cause of injury-related deaths), cardiac events, and stroke. A dispatch system that spends 44% of its capacity on non-emergency calls will systematically delay response to genuine emergencies. The cost is not measured in service-level targets — it is measured in preventable deaths.

Why Traditional 108 Call Handling Fails

The current model places a human dispatcher at the first point of contact for every call. This is the correct design for a low-volume, high-accuracy emergency service. It is the wrong design for a system that receives 250,000 daily calls across 16 states, the majority of which are not emergencies.

The classification problem. A human dispatcher must make a triage decision within 60–90 seconds of answering a call. Under this pressure, with no decision-support tool, dispatchers either classify conservatively (send ambulance to every call, exhausting fleet capacity) or liberally (reject calls that may be genuine emergencies). Neither is acceptable at scale.

The language problem. India's 22 official languages and hundreds of regional dialects mean that a caller speaking Kannada in rural Karnataka, Bhojpuri in eastern Uttar Pradesh, or Odia in coastal Odisha may encounter a dispatcher who does not understand them fluently. Multilingual capacity across 16 states with regional language variation is structurally impossible to achieve with human-only workforce.

The surge problem. Emergency call volumes spike predictably during monsoons, festival seasons, and extreme heat events. These spikes are calendar-predictable, yet the current fixed-workforce model cannot scale to meet them. The only cost-effective response to a 3x surge in 108 calls during monsoon flooding is an AI layer that absorbs the volume increase without requiring proportional staffing increases.

How AI Voice Triage Solves the Problem

The Three-Tier Triage Architecture

An AI triage system for 108 performs three functions that human dispatchers cannot execute efficiently at scale:

Tier 1 — Immediate Emergency (life-threatening): Caller presents with symptoms indicating cardiac arrest, stroke, severe trauma, obstetric emergency, or loss of consciousness. AI instantly routes to human dispatcher for ambulance dispatch. Target handling time: under 30 seconds before transfer.

Tier 2 — Urgent Medical (non-life-threatening but time-sensitive): Caller presents with fractures, fever, moderate pain, controlled bleeding. AI registers the call, collects precise location, and queues for dispatch within 15 minutes. Target: 85% resolved without immediate human dispatcher involvement.

Tier 3 — Non-Emergency (information or routine transport): Caller seeks health guidance, follow-up on prior case, or non-urgent transport. AI redirects to the 104 National Health Helpline for health guidance or schedules a callback. Target: 40–44% of all calls, redirected without consuming dispatcher time.

This architecture — if deployed with 92% triage accuracy — eliminates the 44% non-emergency call burden that Karnataka's CAG identified, restoring dispatcher capacity exclusively to genuine emergencies.

Multilingual Voice Response Using Bhashini

Bhashini's CONVERSE capability, already deployed in production for UP Police 112, provides the multilingual infrastructure for 108's emergency voice triage. The system supports 22 scheduled languages with speech recognition and text-to-speech synthesis. For 108, this means:

A Kannada-speaking farmer in rural Karnataka can describe symptoms in their own language and receive triage questions and instructions in Kannada
An Odia-speaking accident victim can give their location naturally, with the AI converting dialect-accented speech to standardised location data
A Hindi-speaking migrant in Kerala can access emergency services without a Malayam-speaking dispatcher

The Bhashini integration is not a future requirement — it is an available API with a government-to-government procurement pathway that bypasses commercial licensing complexity. (Aisewak Government Helpline Report, 2026, citing Bhashini API documentation and UP Police 112 deployment record.)

Predictive Ambulance Positioning

Beyond call triage, AI can improve the 108 system's dispatch efficiency through predictive deployment. Historical call data, combined with traffic patterns, festival calendars, and seasonal health event data (monsoon disease spikes, summer heat emergencies), can pre-position ambulances at predicted hotspot locations 2–4 hours before peak demand.

Haryana's AI-based auto-dispatch for its 112 emergency system reduced average police response time from 12 minutes 4 seconds to 7 minutes 3 seconds — a 42% reduction — and achieved 92.6% caller satisfaction. (Aisewak Government Helpline Report, 2026, citing Haryana 112 AI deployment data, July 2025.) The same principle applies to ambulance deployment: AI does not eliminate the need for ambulances, it positions them where emergencies are most likely to occur before the call comes in.

Implementation Roadmap

Phase 1: Karnataka Pilot (Months 1–3)

Karnataka's new government-owned 50-seat Central Command and Control Centre, running on C-DAC's NG-ERSS V2.0 platform, is the most technology-receptive deployment target in India. The pilot design:

Scope: Deploy AI voice triage for 9 PM–6 AM night shift (full AI) and 20% daytime overflow (AI + human). Target 3,000 calls over 30 days.
Focus languages: Kannada and Hindi
Target KPIs: Triage accuracy ≥92%; non-emergency filter rate ≥35%; average call handling time under 90 seconds; zero missed emergencies (100% of Tier 1 calls routed to human dispatchers); CSAT ≥75%
Integration: C-DAC NG-ERSS V2.0 API; Bhashini CONVERSE API; Karnataka GIS/AVLT ambulance tracking; 104 health helpline routing for non-emergencies

Procurement pathway: Engage Karnataka NHM Mission Director → Health Minister endorsement → Pilot MoU with NHM Karnataka → C-DAC partnership for platform integration.

Phase 2: Multi-State Replication (Months 4–9)

With a Karnataka reference deployment, scale to states with active procurement:

State	Procurement Signal	Daily Call Volume
Andhra Pradesh	Active integrated 108+104 RFP	~30,000
Himachal Pradesh	104 tender with AI requirements	~5,000
Meghalaya	QCBS tender open	~2,000
Chhattisgarh	Active procurement	~8,000

Source: Aisewak Government Helpline Report, 2026.

Each state deployment uses the Karnataka template as technical baseline, adapting for local language (Telugu in AP, tribal languages in Meghalaya) and operator (GVK EMRI or Ziqitza depending on state).

Phase 3: National Scale via NHSRC and GVK EMRI (Months 10–18)

NHSRC (National Health Systems Resource Centre) sets national EMS guidelines and influences state NHM procurement across all 36 states. A single NHSRC advisory recommending AI triage for 108 creates a policy-level mandate that converts 16 existing GVK EMRI state deployments into a pipeline.

GVK EMRI, with 13,000+ ambulances and 16-state operational footprint, represents an alternative scaling pathway: a technology partnership with GVK EMRI embeds AI triage into their existing command centre operations without requiring individual state procurement cycles.

ROI and Cost-Benefit Analysis

Cost reduction. The primary financial case for AI triage is call containment. If AI handles 44% of calls (the non-emergency volume identified by Karnataka's CAG) without human dispatcher involvement, it eliminates 44% of the dispatcher workload from that segment. At an estimated Rs 25 per human-handled call (fully loaded cost including salary, infrastructure, and management overhead), redirecting 110,000 daily non-emergency calls (44% of 250,000) to AI at Rs 3–5 per AI-handled call generates annual savings of Rs 73–82 crore across the current 16-state footprint. This is a conservative estimate that does not account for reduced fleet utilisation costs from better dispatch efficiency.

Response time improvement. Removing 44% of non-emergency calls from dispatcher queues is the direct lever for response time improvement. If Karnataka's CAG data is representative, the current dispatcher sees 1.8 calls for every genuine emergency. Eliminating the non-emergency volume means each genuine emergency call reaches a dispatcher faster. Even a 15% improvement in average response time — from 134.5 minutes in Maharashtra or 41 minutes in Madhya Pradesh — translates to measurable improvement in trauma survival rates, though we will not speculate on specific survival rate changes without clinical data.

Pilot investment. A 30-day Karnataka pilot is estimated at Rs 40 lakh, covering voice AI platform, Bhashini Kannada integration, C-DAC NG-ERSS API integration, night-shift pilot operations, and analytics dashboard. (Aisewak Government Helpline Report, 2026.) The ROI demonstration from a single 30-day pilot — showing triage accuracy, non-emergency filter rates, and CSAT data — provides the evidence base for state-level procurement at Rs 50 lakh to Rs 2 crore annual maintenance per state.

Risks and Mitigation

Risk	Likelihood	Mitigation
AI misclassifies genuine emergency as non-emergency	Low	Zero-miss threshold design: any caller expressing pain, distress keywords, or uncertainty defaults to Tier 1 (human dispatcher)
Kannada/regional dialect speech recognition accuracy insufficient	Medium	Bhashini CONVERSE tested for Kannada in production (UP Police 112 reference); pilot data sets confidence threshold before scale
C-DAC NG-ERSS V2.0 integration complexity	Medium	C-DAC as technical partner from pilot stage; no replacement of platform, only overlay integration
GVK EMRI/Ziqitza resistance to AI overlay	Low	Technology presented as operator support tool, not replacement; labour arbitrage argument removed by framing as surge-handling
State NHM procurement bureaucracy (18–36 month cycles)	High	Karnataka pilot designed as MoU-based proof of concept, not formal procurement; scale through NHSRC national advisory to compress state cycles
Budget austerity in state health departments	Medium	AI triage priced as cost-neutral from day one: savings on non-emergency call handling offset pilot costs within 60–90 days

Future Outlook

Karnataka's transition to a government-owned 108 command centre is the beginning of a national trend, not an isolated event. The PPP model that has governed 108 for 15 years is under structural pressure: labour disputes, contract renewals, and the CAG's documented performance failures are pushing states toward direct government operation. As states take direct control, they also gain direct procurement authority for technology — removing the intermediary operator layer that previously controlled call centre technology decisions.

The NHSRC's national EMS guidelines are due for review in 2026–27. A technology modernisation mandate in those guidelines — specifically requiring AI-assisted triage and multilingual voice capability — would create a national compliance requirement that converts every state 108 deployment into a procurement window. The IndiaAI Mission's Rs 10,372 crore GPU compute investment is already reducing the cost of voice AI inference, making AI triage economically viable at the per-call pricing required for government contracts.

By 2030, the question for India's 108 system will not be whether AI triage should be adopted — it will be which states adopted it first and have the performance data to prove its value.

Key Takeaways

India's 108 ambulance service receives approximately 250,000 daily calls but answers only 86,000, with CAG audits documenting 44% non-emergency volume and 64% ineffective responses in Karnataka
Labour strikes in Punjab, Rajasthan, and UP reveal systemic workforce fragility that AI triage can mitigate through 24/7 automated call classification
Bhashini's CONVERSE API, already production-deployed for UP Police 112, provides the multilingual infrastructure needed for 108's regional language diversity
Karnataka's new government-owned command centre on C-DAC NG-ERSS V2.0 is the optimal pilot target: technology-ready, procurement-receptive, and politically motivated
Active tenders in Andhra Pradesh, Himachal Pradesh, Meghalaya, and Chhattisgarh provide immediate multi-state scaling opportunities after a Karnataka pilot reference
The ROI case rests on call containment economics: redirecting 44% of non-emergency calls to AI at Rs 3–5 per interaction generates material savings against the Rs 200–400 crore annual 108 call centre operations market

Conclusion

India's 108 ambulance service has operated for 15 years without AI-assisted triage at the point of first contact. The result is documented by the CAG: nearly half of all calls misclassified, two-thirds of responses ineffective, and a workforce so underpaid and overworked that four states have experienced operational shutdowns through strikes. The AI solution is architecturally clear, technically available through Bhashini and C-DAC infrastructure, and economically justified at the per-call pricing that government procurement requires.

The procurement moment has arrived. Karnataka has built the government-owned platform. Andhra Pradesh, Himachal Pradesh, Meghalaya, and Chhattisgarh have active tenders. NHSRC has the national policy lever. What is needed now is a 30-day proof of concept in Karnataka that demonstrates triage accuracy, call containment rates, and citizen satisfaction — the three data points that convert departmental interest into procurement action.

Government leaders exploring AI-powered citizen engagement can begin with a focused pilot in one department or constituency to validate impact before scaling statewide. Aisewak helps public institutions deploy multilingual Voice AI solutions designed specifically for Indian governance.

FAQ

Q: What percentage of 108 calls are genuine emergencies that require ambulance dispatch? A: Karnataka's CAG audit (covering 2014–2019) found that 44% of calls received by the 108 command centre were non-emergency. If this pattern holds nationally, approximately 56% of the 86,000 daily answered calls represent genuine emergencies — though this figure varies by state and urban-rural mix. (Aisewak Government Helpline Report, 2026, citing CAG Karnataka Audit 2014–19.)

Q: Can AI reliably distinguish a genuine cardiac emergency from a non-emergency call? A: AI triage systems designed for emergency services use conservative classification thresholds: any caller expressing pain, distress, or uncertainty is routed to a human dispatcher (Tier 1). The AI only classifies calls as non-emergency when the caller explicitly describes a non-urgent situation (routine transport, health information, follow-up). Well-designed systems target zero missed genuine emergencies, with accuracy validated against historical call data before deployment.

Q: How does Bhashini enable multilingual triage for 108? A: Bhashini's CONVERSE capability supports 22 scheduled languages with speech recognition and text-to-speech synthesis. For 108, this means a caller can describe symptoms and provide location in Kannada, Telugu, Odia, or Hindi, and the AI processes the response in that language. The UP Police 112 deployment provides a production reference for the same infrastructure in an emergency services context. (Aisewak Government Helpline Report, 2026.)

Q: What is Karnataka's specific advantage as a pilot state? A: Karnataka launched a fully government-owned 50-seat Central Command and Control Centre on C-DAC's NG-ERSS V2.0 platform in May 2026 — terminating its previous GVK EMRI contract. This makes Karnataka the only state where the government directly controls the 108 command centre infrastructure, enabling technology deployment without commercial operator intermediaries. C-DAC's NG-ERSS V2.0 explicitly supports AI integration. (Aisewak Government Helpline Report, 2026.)

Q: How does AI triage reduce ambulance response times? A: By filtering non-emergency calls from dispatcher queues, AI triage ensures every human dispatcher interaction is with a genuine emergency. Fewer non-emergency calls competing for dispatcher attention means genuine emergencies reach dispatch faster. Additionally, predictive ambulance positioning — pre-deploying vehicles to high-probability locations based on historical patterns — reduces travel time from dispatch to scene. Haryana achieved a 42% reduction in police response time using AI auto-dispatch on its 112 system. (Aisewak Government Helpline Report, 2026, citing Haryana 112 AI deployment data.)

Q: What are the active 108 tenders that offer immediate procurement opportunities? A: As of mid-2026, Andhra Pradesh has an integrated 108+104 RFP open, Himachal Pradesh has a 104 tender with AI requirements, Meghalaya has issued a QCBS tender, and Chhattisgarh has active procurement underway. (Aisewak Government Helpline Report, 2026.) State-level procurement timelines vary; NHM Mission Directors in each state are the primary decision-makers for tender scope.

Q: How does AI triage address the 108 labour strike problem? A: AI triage creates an operational resilience layer for call classification and routing that is not dependent on human workforce continuity. During a labour dispute, an AI system can continue to accept calls, triage them, redirect non-emergencies, and maintain a queue for emergency dispatch — even with reduced human dispatcher staffing. This does not replace paramedics or ambulance crews, but it prevents the total operational collapse that characterised the Punjab and Rajasthan strikes.

Q: What is the procurement pathway for deploying AI triage on 108? A: For Karnataka (government-owned CCC): Engage Karnataka NHM Mission Director → Health Minister endorsement → MoU-based pilot with NHM Karnataka → C-DAC technical partnership for NG-ERSS integration. For GVK EMRI states: Technology partnership with GVK EMRI at the operator level, embedding AI triage into their existing command centre operations. For national scale: NHSRC advisory recommending AI triage creates a policy mandate that drives state-level procurement across all 16 108 states.

Q: What does a 108 AI triage pilot cost? A: A 30-day Karnataka pilot covering voice AI platform, Bhashini Kannada integration, C-DAC NG-ERSS API integration, night-shift pilot operations, and analytics dashboard is estimated at Rs 40 lakh. Full state deployment ranges from Rs 50 lakh to Rs 2 crore annual maintenance depending on state call volume and language requirements. (Aisewak Government Helpline Report, 2026.)

Q: How should state Health Secretaries evaluate AI triage vendors? A: Three criteria matter most: (1) triage accuracy — specifically, the system's false-negative rate (genuine emergencies classified as non-emergency); this should be near zero in any acceptable system; (2) language capability — the vendor should demonstrate production-grade speech recognition in the state's primary language, not just Hindi and English; (3) integration experience — C-DAC NG-ERSS integration or GVK EMRI system integration requires specific technical capability that not all vendors possess.

Schema Markup Suggestions

Article — headline, description, author (Aisewak), datePublished, dateModified, publisher
FAQPage — all 10 Q&A pairs; high probability of Google Featured Snippet / AI Overview extraction
GovernmentService — serviceType: "Emergency Medical Services", areaServed: India, provider: NHM/GVK EMRI/Ziqitza
MedicalOrganization — for NHSRC and NHM references
HowTo — for the 3-phase implementation roadmap

Suggested External References

CAG Report on Performance Audit of 108 Emergency Ambulance Services — Karnataka (2014–19)
CAG Report on Odisha Emergency Ambulance Services (2013–14)
CAG Report No. 7 of 2025 — Punjab Emergency Services
National Health Systems Resource Centre (NHSRC) — EMS Guidelines
GVK EMRI Annual Operations Report
Ministry of Health and Family Welfare — National Health Mission EMS Programme Documentation
Bhashini (DIBD, MeitY) — CONVERSE API Documentation
C-DAC — NG-ERSS V2.0 Platform Documentation

India's 108 ambulance service answers only 86,000 of 250,000 daily calls — and Karnataka's CAG found 44% of those were non-emergency. AI voice triage can fix this: filter non-emergencies, support 22 languages, and ensure genuine emergencies reach dispatchers instantly. Karnataka's new C-DAC command centre is the pilot target. The technology is available. The procurement window is now. #GovTech #India #AIGovernance #EmergencyServices

LinkedIn Executive Summary

India's 108 ambulance service is the world's largest pre-hospital emergency network — and one of its most documented failures. CAG audits across Karnataka, Odisha, Maharashtra, and Kerala reveal systematic response-time failures, 44% non-emergency call flooding, and workforce strikes that left millions without emergency medical access for days at a time.

The fix is not more call centre agents. It is AI voice triage at the point of first contact: classify calls into emergency, urgent, and non-emergency tiers before a human dispatcher is involved. Bhashini's multilingual voice capability — already deployed for UP Police 112 — provides the infrastructure. Karnataka's new government-owned command centre on C-DAC's NG-ERSS V2.0 platform is the optimal pilot site.

The business case is straightforward: redirect 44% of calls to automated handling at Rs 3–5 per interaction instead of Rs 25 per human-handled call, and use the savings to fund a better emergency response for the 56% that are genuine. State Health Secretaries and NHM Mission Directors have the procurement authority to act. The question is sequencing: Karnataka first, then the four states with active tenders, then NHSRC national advisory for the rest.

AI Search Optimization Summary

Entities: 108 Ambulance Service, National Health Mission (NHM), GVK EMRI, Ziqitza Healthcare, C-DAC, NG-ERSS V2.0, NHSRC, Bhashini, CAG India, Karnataka Health Department, Andhra Pradesh NHM

Topics: Emergency medical triage AI, pre-hospital care technology, government helpline automation, multilingual voice AI for healthcare, ambulance dispatch optimisation, India EMS modernisation, PPP emergency services technology

Semantic keywords: 108 AI triage, ambulance dispatch AI India, non-emergency call filtering, Bhashini emergency voice, Karnataka 108 command centre, C-DAC ERSS ambulance, NHM voice AI procurement, EMS AI India 2026, government ambulance AI vendor, NHSRC EMS guidelines AI