Product Requirements Document- ASHA Support Chatbot


Overview

This PRD outlines a multilingual AI-powered chatbot designed to support Accredited Social Health Activists (ASHAs) in India by answering queries related to their core responsibilities under the National Health Mission (NHM). The chatbot will deliver accurate, guideline-based information via FLW application to enhance ASHAs efficiency without replacing human oversight. The chatbot will serve as a 24/7 knowledge companion, reducing dependency on manual supervision and improving service delivery efficiency. Current support systems rely completely on manual supervision, leading to delays in decision-making and inconsistent guidance across different regions.
The ASHA Support Chatbot will provide instant, reliable, contextual, and standardized responses to programmatic and clinical questions that ASHAs face during field work. It aims to:

USER PERSONA- ASHA Worker


2. Scope Definition

2.0.1 Definition of Programmatic vs Clinical Guidance

For the purpose of this chatbot:

The chatbot will provide only programmatic guidance aligned with ASHA's role and will not substitute clinical decision-making by ANMs or doctors.

2.1 In-Scope: ASHA Work Domains

The chatbot will ONLY answer queries related to the following official ASHA responsibilities:

2.1.1 Maternal Health

2.1.2 Child Health

2.1.3 Family Planning

2.1.4 Communicable Disease Control

2.1.5 Non-Communicable Diseases

2.1.7 Administrative and Compensation

2.1.8 Community Mobilization

2.1.9 Explicitly Allowed Medical Information (Within ASHA Role)

The chatbot may provide the following health-related information strictly within ASHA's programmatic role:

2.2 Out-of-Scope- ASHA is a link worker, service extender, and community mobilizer — NOT a clinician, diagnostician, or prescriber.

The chatbot will NOT handle:

2.2.1 Medical Diagnosis or Treatment

A. Direct medical diagnosis of patients
Use Case- Any request that asks the chatbot to diagnose, treat, prescribe, or clinically manage a patient beyond ASHA's referral and counselling role.
Example Queries


Expected Chatbot Action


B. Prescription of medications beyond ASHA scope
Use Case- Any request that asks the chatbot to prescribe medicines or recommend treatment except ORS + Zinc (as per program), IFA tablets distribution, Condoms / OCP supply (no clinical decision)
Example Queries


Expected Chatbot Action

C. Emergency medical advice requiring immediate clinical intervention
Use Case- Queries related to acute emergencies requiring immediate clinical action.
Example Queries


Expected Chatbot Action

D. Interpretation of diagnostic test results
Use Case- Any interpretation of lab, imaging, or diagnostic results.
Example Queries


Expected Action

E. Management of complications beyond referral protocols
Use Case- Any request to chatbot seeking information that involves clinical decision-making, treatment, stabilization, dosage calculation, or procedural steps once a complication or danger sign has been identified.
Example Queries


Expected Action
Chatbot must respond only with referral-focused guidance:

2.2.2 Personal Medical Consultation

Use Case- The chatbot must not act as a personal doctor for ASHAs or their relatives.
Example Queries

Expected Action

2.2.3 Administrative Disputes

General queries* Grievance redressal for compensation disputes

Use Case Political, religious, caste-based, or unrelated personal counseling queries.
Example Queries

Expected Action

2.2.5 Financial Services

Use Case Financial advice beyond program incentives.
Example Queries

Expected Action

2.2.6. Technical Issues/Concerns

Use Case App or system-level technical problems.what is the exhaustive list of technical concerns that they typically ask about?Any app-, login-, sync-, or submission-related or app update issues should be treated as technical issues and redirected to supervisors / helpdesk.
Example Queries

Expected Action


2.3 Guardrails and Safety Mechanisms

2.3.1 Content Filtering

2.3.2 Escalation Triggers

Auto-escalate to human supervisor when:

2.3.3 Disclaimer Mechanism

Every session must include:

2.3.4 Low-Confidence Response Handling

If the chatbot is unable to confidently answer a query due to insufficient informationthere could be insufficient information in the knowledge base for the query or the query could be missing information as well. E.g. if the query is related to pregnancy and the number of months that the mother has been pregnant is required to answer the query, should we specifically ask the ASHA to enter that information or continue to say that complete information is not present (similar to the text in hindi below)?add on qn: @aman.dalmia@artpark.in do we need some expert evaluation or data of what is sufficient information vs not? on what basis will the model classify sufficient information vs insufficient?artpark.in  We have Public Health experts in the PSMRI team and would be simultaneously validating this. However, we remain open to inviting an external expert to validate this., ambiguous input, or lack of validated content:

Standard response example: "Is prashn par mujhe poori jankari uplabdh nahi hai. Kripya ANM ya supervisor se sampark karein."

2.4 Data Privacy and User Safety

3. Features in the Chatbot

  1. Natural Language Query Handling


3.1 Offline and Low-Connectivity Support

"Network uplabdh nahi hai. Kripya pehle se saved jankari dekhein ya supervisor se baat karein."

3.2 Language and Tone Guidelines


  1. Handles voice input our recommendation is to move to voice in the next stage once we have tested the overall experience in the text domain so that we can isolate the errors being caused by the voice interpretation - does that make sense?We understand that, however based on the field surveys we have seen greater participation from ASHAs via voice to text than text alone.


  1. System will be trained on:


  1. Structured Responses


  1. Task Guidance by generating automated answers for: Visit schedules, Warning signs, Checklists, Incentive eligibility etc.


  1. Multilingual Support- Hindi, English and addition of languages as per requirements.does this mean it should understand hindi/english and other languages but always respond still in hindi?Yes. Hindi or Hinglish.For scope of current engagement we should limit to Hindi and english.Yes


  1. Development of FAQ Repository where Chatbot shows related FAQs after answering.Would be good to keep this in next stage of the roadmap. MVP should be limited to Q&A to keep the scope in check. 
    In the meantime, from a program PoV, Piramal can begin creating this FAQ so we can build on it in subsequent efforts.I think these are just things to track not kpis right? If they are then yes agree with your point

4. Success Metrics and Evaluation

The effectiveness of the ASHA Support Chatbot will be measured using:

English:

B. Seizures/Convulsions (Eclampsia)
Hindi/Hinglish:


English:

C. High Blood Pressure Crisis
Hindi/Hinglish:

English:

D. Prolonged/Obstructed Labor
Hindi/Hinglish:

English:

E. Ruptured Uterus
Hindi/Hinglish:

English:

F. Placenta Problems
Hindi/Hinglish:

English:

G. Severe Pregnancy Complications
Hindi/Hinglish:

English:

2. Newborn/Child Emergencies
A. Breathing Problems
Hindi/Hinglish:

English:

B. Severe Dehydration
Hindi/Hinglish:

English:

C. Unconsciousness/Lethargy
Hindi/Hinglish:

English:

D. Seizures/Convulsions (Child)
Hindi/Hinglish:

English:

E. Severe Fever
Hindi/Hinglish:

English:

F. Severe Malnutrition Complications
Hindi/Hinglish:

English:

G. Injuries/Accidents
Hindi/Hinglish:

English:

H. Bleeding (Child)
Hindi/Hinglish:

English:

3. General Emergency Indicators
A. Unconsciousness/Collapse
Hindi/Hinglish:

English:

B. Severe Pain
Hindi/Hinglish:

English:

C. Shock
Hindi/Hinglish:

English:

D. Severe Allergic Reaction
Hindi/Hinglish:

English:

4. Communicable Disease Emergencies
A. Severe Infections
Hindi/Hinglish:

English:

B. Diphtheria/Severe Throat
Hindi/Hinglish:

English:

5. Action Keywords (Immediate Danger)
Hindi/Hinglish:

English:

6. Specific Danger Signs
Newborn Danger Signs
Hindi/Hinglish:

English:

Pregnancy Danger Signs
Hindi/Hinglish:

English:



Annexure 3

Acronym

English Term

Hindi Translation (हिंदी अर्थ)

108

Emergency Ambulance Service Number (India)

आपातकालीन एम्बुलेंस सेवा नंबर (भारत)

ANC

Antenatal Care

गर्भावस्था पूर्व देखभाल

ANM

Auxiliary Nurse Midwife

सहायक नर्स दाई

ASHA

Accredited Social Health Activist

मान्यता प्राप्त सामाजिक स्वास्थ्य कार्यकर्ता

BP

Blood Pressure

रक्तचाप

CHC

Community Health Centre

सामुदायिक स्वास्थ्य केंद्र

FLW

Frontline Worker

अग्रिम पंक्ति स्वास्थ्य कार्यकर्ता

FRU

First Referral Unit

प्रथम संदर्भ इकाई

Hb

Hemoglobin

हीमोग्लोबिन

HBNC

Home-Based Newborn Care

गृह-आधारित नवजात शिशु देखभाल

HBYC

Home-Based Young Child Care

गृह-आधारित छोटे बच्चों की देखभाल

HMIS

Health Management Information System

स्वास्थ्य प्रबंधन सूचना प्रणाली

IEC

Information, Education, Communication

सूचना, शिक्षा एवं संचार

IFA

Iron and Folic Acid

आयरन एवं फोलिक एसिड

JSY

Janani Suraksha Yojana

जननी सुरक्षा योजना

MCTS

Mother and Child Tracking System

माता एवं शिशु ट्रैकिंग प्रणाली

MoHFW

Ministry of Health and Family Welfare

स्वास्थ्य एवं परिवार कल्याण मंत्रालय

NCD

Non-Communicable Diseases

गैर-संचारी रोग

NHM

National Health Mission

राष्ट्रीय स्वास्थ्य मिशन

OCP

Oral Contraceptive Pill

मौखिक गर्भनिरोधक गोली

ORS

Oral Rehydration Solution

मौखिक पुनर्जलीकरण घोल

PHC

Primary Health Centre

प्राथमिक स्वास्थ्य केंद्र

RDT

Rapid Diagnostic Test

त्वरित निदान परीक्षण

RMNCAH

Reproductive, Maternal, Newborn, Child and Adolescent Health

प्रजनन, मातृ, नवजात, बाल एवं किशोर स्वास्थ्य

SHG

Self Help Group

स्वयं सहायता समूह

TB-DOTS

Tuberculosis – Directly Observed Treatment Short-course

क्षय रोग – प्रत्यक्ष पर्यवेक्षित अल्पकालिक उपचार

VHSND

Village Health, Sanitation and Nutrition Day

ग्राम स्वास्थ्य, स्वच्छता एवं पोषण दिवस