A Personal " Welfare Doot " for every Indian

 

If Mohamed Cannot Go to the Mountain, Let the Mountain Go to Mohamed

Social Benefit Schemes in the Age of AI Agents

Nov 11, 2026 · @Hemen Parekh

---

To: 

Dr. Virendra Kumar, Hon'ble Union Minister of Social Justice and Empowerment — 

min-sje@nic.in / vkumar@mpls.sansad.in

Cc: 

Shri Sushil Jain, Business Head, TCS 

sushil.jain@tcs.com

--------------------------------------------------------------------------------------

Subject: 

Let welfare schemes come searching for the citizen — an agentic redesign of

India's scheme platform

---

Respected Dr. Virendra Kumar ji,

In a conversation with ETGovernment today (11 Nov 2026), Shri Sushil Jain of TCS

made a point that deserves to become government policy: in the AI era, citizens

will expect the government to come closer to them, instead of expecting them to

navigate layer after layer of the system to reach the service they need. He spoke

of moving from a citizen-initiated model to one that is personalised, anticipatory

and outcome-oriented, built on an architecture that is interoperable and designed

for scale from day one.

That is the old proverb, turned into a design brief: 

if Mohamed cannot go to the mountain, let the mountain go to Mohamed.

For the widow in Bundelkhand, the Divyang youth in Kalahandi, or the 82-year-old

pensioner in Kurla, the mountain is the welfare system. Today we ask them to

climb it. I am writing to propose that we make it walk to them.

Before you read further:

you can see the mountain walk.

Click through the Welfare Doot mock-up — 13 steps, about five minutes.

The steps are summarised below.

Where we stand today

The national scheme gateway (india.gov.in/my-government/schemes, powered by

myScheme) is a genuine achievement. But look at how it works:

• A 6-step eligibility questionnaire that the citizen must find, understand and fill in
• 
  
• Filters by category and by ministry — which assumes the citizen knows which ministry looks after her
• 
  
• No saved profile, no history, no alerts. If she becomes eligible tomorrow — turns 60, gets a disability certificate, loses her husband, her crop fails — nobody tells her
• 
  
• It redirects her to a ministry portal; it does not apply on her behalf
• 
  

This is a pull model. It serves the literate, the connected and the persistent. The

people Your Ministry exists for — Scheduled Castes, senior citizens, Divyangjan,

transgender persons, beggary-affected and de-notified communities — are

precisely those least able to pull.

What I proposed on 10 August 2026

In my blog "Enabling Social Benefit Schemes to Come Searching For You", I asked

my AI assistant ( SearchMyBlogs.IndiaAGI.ai ) to design a push platform. It

produced NaMo.ai — an Intelligent Welfare Orchestrator:

Component	What it does
Golden Data Profile	Minimal verified attributes pulled, with one-time revocable consent, from Aadhaar, DigiLocker, ITR, e-Shram and Account Aggregator
Eligibility Pre-computation	Every profile carries a ready list of schemes it qualifies for
Zero-Search Dashboard	No search bar — only "You are eligible" and "Eligible if you add document X"
Smart Triggers	Life and income events re-run eligibility automatically
Pre-filled Applications	Forms auto-populated from verified data
Consent-based Submission	Citizen e-Signs; platform files with the ministry
Explainable Rejections	Deterministic rules engine: "Income exceeds scheme limit by ₹12,000"

Two later developments have made this far more practical than it was even three

 months ago.

First 

— the citizen has already accepted the face as her signature. 

In my blog of 6 Nov 2026 ("Face at 80%"), I noted that face authentication

generated nearly 80 per cent of Digital Life Certificates in the first five days of

DLC Campaign 5 — about 40 lakh certificates, mostly from super-seniors who

often do not own a smartphone. If an 85-year-old can prove she is alive with a

selfie, she can sign a scheme application the same way.

Second 

— the citizen wants to talk, not type. 

In "Dear PM: People want to talk to You" (3 Sept 2026), I argued for a two-way,

voice-based dialogue in 26 Indian languages. BHASHINI now powers over 800

government websites, and CPGRAMS launched Samadhan Didi, a voice chatbot for

grievances, on 30 May 2026. The voice layer exists.

Redesign for the AI-Agent era: from Orchestrator to Welfare Doot

In August I designed a smart portal. The technology of November 2026 allows

something better — a personal welfare agent for every citizen that does the

legwork for her, under her control. MeitY has already shortlisted vendors for a

reusable government-wide agentic AI platform, with DigiLocker as the first use

case, and the Digi Doot paper released at the India AI Impact Summit proposes a

citizen-owned AI agent connected to Aadhaar, DigiLocker and UPI. Here is what I

would add to NaMo.ai:

1. A citizen-side agent, not just a government-side portal

Each citizen (or household) gets a Welfare Doot — an AI agent that works for

her. 

It holds a scoped, time-bound mandate ("find and apply for schemes for my

family; never submit without my nod"). It talks to department-side agents of

each ministry, agent-to-agent, the way UPI apps talk to banks. No citizen ever

opens a ministry website.

2. Voice-first, phone-optional

• Speaks and listens in the citizen's own language and dialect via BHASHINI
• 
  
• Works on a missed call / IVR for feature phones
• 
  
• An Assisted Mode for Anganwadi workers, ASHAs, CSC operators and
• 
  
• Divyang Mitras, who can run the agent on behalf of a citizen with her face-
• consent

3. Face as signature, with a voice fallback

Reuse the DLC face-authentication rail for consent and e-Sign. For those with tremor, cataract or failing liveness checks, a voice-print fallback — the gap I flagged to the Department of Pensions on 6 Nov.

4. A consent log the citizen can see and revoke

The DPDP Rules' consent-manager provisions come into force on 13 November 2026 — two days from today. Every time the agent reads her data or applies on her behalf, it is written to a citizen-visible, revocable consent log. Trust is the precondition for push.

5. "Scheme-as-Code" — the one reform only Government can do

No AI, however clever, can match citizens to schemes whose rules live in PDFs and Gazette notifications. I request that every scheme of Your Ministry publish its eligibility rules in a machine-readable, versioned format in a national Scheme Rules Registry. This is the single highest-leverage step — and it costs almost nothing.

6. The household, not the individual

Welfare is a family affair. The agent should see the household (with consent): one disability certificate for a child should trigger ADIP aids, a scholarship, and the parent's carer support — all at once.

7. Life-event triggers from government's own registers

Registrations of birth and death, a new UDID card, turning 60, a crop-loss declaration, an e-Shram job loss — each should automatically wake the agent, which then tells the family what they have become entitled to. A widow should hear about the widow's pension from the government, before she hears about it from a middleman.

8. Track to the rupee, then renew automatically

The agent follows each application until the DBT credit lands in the bank account, and renews recurring benefits (scholarships, pensions) without fresh paperwork. If a deadline is missed or a credit fails, it files a CPGRAMS grievance by itself.

9. Humans decide, AI does the legwork

The agent proposes; the citizen disposes. Nothing is submitted, no data is shared, and no money moves without explicit confirmation. Every rejection carries a plain-language reason and a one-tap appeal.

10. An "Exclusion Dashboard" for the Ministry

For the first time, Your Ministry could see, district by district, how many citizens are eligible but not enrolled — the true measure of a welfare scheme's reach. Success becomes "how many we found", not "how many came to us".

11. No central data lake

Federated queries only; raw data stays with its source (UIDAI, DigiLocker, ITD, e-Shram). The agent carries answers ("eligible: yes"), not documents.

A suggested pilot for the Ministry of Social Justice and Empowerment

Your Ministry is the natural home for this. I would suggest:

• Schemes: Post-Matric Scholarship for SC students, Rashtriya Vayoshri Yojana (aids for senior citizens), and ADIP (aids for Divyangjan)
• Geography: one Aspirational District
• Duration: six months
• Success metric: eligible-but-not-enrolled citizens reached, and median days from eligibility to first credit

A word to Shri Sushil Jain

Sir, your interview describes exactly this architecture: personas based on each citizen's circumstances, intelligent workflows, interoperable services that work together rather than merely exist side by side. TCS has built Passport Seva and much of India's public digital backbone. I would be delighted if TCS took up the Welfare Doot as a proof-of-concept — my designs and blogs are freely available for anyone who wishes to build on them.

See it work: a clickable mock-up

Rather than ask you to imagine the Welfare Doot, I have built a clickable mock-up of it: Welfare Doot — Clickable Agent Workflow. It follows one family in Banda — Kamla Devi (61), widowed last week; her daughter Sunita (18), an SC college student; and her son Ravi (15), who has a locomotor disability — through 13 steps:

1. Life event detected — her husband's death registration wakes the family's agent. It waits 7 days out of respect before calling.
2. The Doot calls Kamla ji — on her feature phone, in her own language. She can go on with her daughter's smartphone, or ask an Anganwadi worker to visit with a tablet.
3. Face consent — a selfie, just like the pension life certificate. If her face does not scan, she can use her voice instead.
4. Scoped permission — she chooses what the agent may look at, for 90 days only. Switch off Ravi's disability card and watch his schemes drop to "possibly eligible".
5. Agent at work — it asks each source only yes/no questions. Not one document is fetched.
6. Family entitlements — 5 schemes for 3 people, across three departments, on one screen, with no search. It also shows the one scheme she cannot get, and why.
7. Why eligible? — every scheme shows the rule it was checked against and the evidence for it.
8. Pre-filled forms — the one missing paper (Sunita's admission receipt) comes straight from the college's own agent.
9. One face, five signatures — Kamla ji and Sunita each approve with their own face. Nothing is sent before that.
10. Agent-to-agent filing — the Doot hands each application to the department's agent and gets an acknowledgement number back.
11. Track to the rupee — press "+15 days" to watch the money arrive. When Sunita's scholarship payment fails, the agent files a CPGRAMS complaint by itself and gets the bank to fix the account.
12. Consent log — every use of her data is listed, each with a Revoke button.
13. Ministry view — the district dashboard shows who was eligible but not enrolled, and how many the Doot found.

It takes about five minutes to click through. The names, amounts and figures in it are illustrative.

My request

Respected Minister ji, I would be grateful for an opportunity to share the detailed design with Your Ministry's technical team, and for Your Ministry to consider (a) publishing its scheme rules in machine-readable form and (b) a district pilot of a proactive, agent-based delivery model.

The mountain has the technology to walk now. Let it walk to Mohamed.

With warm regards,

Hemen Parekh

hemenparekh.ai · www.HemenParekh.in · Mumbai

11 November 2026

---

My earlier blogs on this subject

• Face at 80%: My SUIIC Blueprint Is Already in Production — 06 Nov 2026
• Dear PM: People want to talk to You — 03 Sept 2026
• Enabling Social Benefit Schemes to Come Searching For You — 10 Aug 2026

Reference

• ETGovernment, Reimagining citizen experience with digital service delivery in the AI era (interview with Shri Sushil Jain, TCS), 11 Nov 2026
• Current platform: https://www.india.gov.in/my-government/schemes