Leap of Faith Labs · Program 03

AI for ALL — Pricing Navigator

Four institutional programs, one gated cohort. What each door costs, what it includes, and what we will refuse to do.
AI-FOR-ALL.PRICING
v0.13 · 2026-08-16 · supersedes v0.12
पहले विवेक, फिर उपकरण
Refusal-as-skill · the first deliverable is a refusal inventory Institution evaluates LOFL · never the reverse In-language · Marathi · Hindi · English Public documentation · by default
In motion: First pilot: underway Government & education — in sāṃvāda
One grammar, four doors
Every program runs the same arc: Saṃvāda → Discernment Lab → Integration Scaffolding → Sevā. Counts below are set-shaped tallies — never scores.
4
Programs
1
Live pilot
1
Gated cohort
2
Open decisions
A thinking tool, not a quote. The numbers are starting positions; every engagement is shaped to its context, and refusal is part of the offer. Pick your organisation’s size inside each program to narrow the range. All prices in INR, exclusive of GST (18%, professional services). Ranges reflect scope and institution size; market-checked August 2026. Sequencing is an order, not a set of dates: Nāgarika → Sahakāra → Vivekī → Hastakāra. First engagements proceed on engagement letters or documentation-in-lieu; GST invoicing begins at incorporation.

Nāgarika

नागरिक First pilot: underway
"Citizen data is a public trust."
If you are a civic-tech platform, citizen-led project, or transparency/RTI initiative — building software that holds public data or mediates between citizens and institutions. Refusal translation: citizen data — what the platform will never do with it, inventoried before any tool is adopted.
Your size:
Saṃvāda→ Discernment Lab→ Integration Scaffolding→ Sevā
TierEngagementRange
Tier 1 · SaṃvādaTier 1 is not a discovery call — it’s this instrument: the Succinct Sāṃvāda (5 questions). The full 25-question sheet and the founder’s own answers are public.
In-kind / documentation-in-lieu — open to the first 2–3 pilots. One underway; the door is not closed.
in-kind
Payment posture: sliding scale by org size and funding stage. The first 2–3 pilots can opt for documentation-in-lieu-of-fee — the co-authored public case study itself counts as payment.
Tier 2 · Discernment Lab
4 weeks · 1 session/week + async
Discernment cohort
Trust-surface map of one feature · refusal inventory (private working doc) · sovereign-by-design architecture sketch · public trust statement · internal governance note · co-authored case study to the commons · Padtalani add-on available (see module below)
₹1.5–4.5L
Tier 3 · Integration Scaffolding
8–12 weeks · embedded
Embedded build
Everything in Tier 2 · LOFL embedded with your engineering team · implementation support across the full feature · audit cadence and governance handoff · evidence instrumentationNEW — dashboards and logs so the platform verifies its own AI claims
₹6–16L
Begin with the sheet →
Open in your mail app →
Open Claude →
Paste it there — or into any model you trust, including one running on your own device. You carry the words; nothing is injected, so no caution screens. The conversation is yours; LOFL never sees it. You send the final email yourself.
Copy pastes the inquiry anywhere (WhatsApp works too). Mail opens your own app, pre-filled. Draft-with-AI shows you a prompt to carry into a model of your choice.

Sahakāra

सहकार Federation-paid
"Member data is collective property."
If you are a cooperative, FPO, dairy union, weaver coop, fisher collective, or producer federation — member-owned, often touching ONEST/Beckn rails for market access. Refusal translation: member data — held in trust, never surrendered to a vendor.
Your size:
Saṃvāda→ Discernment Lab→ Integration Scaffolding→ Sevā

Tier 1 is not a discovery call — it’s this instrument: the Succinct Sāṃvāda (5 questions). The full 25-question sheet and the founder’s own answers are public.

TierEngagementRange
Tier 2 · Discernment LabDiscernment cohort across member societies
Refusal inventory in the federation's own vocabulary
₹3–6.5L
Tier 3 · Integration ScaffoldingEmbedded build
Cooperative-owned deployment · evidence instrumentationNEW — the federation verifies its own claims, not LOFL's
₹10–20L
Tier 4 · SevāTrain-the-trainer
Federation staff carry the grammar forward without LOFL
₹8–15L
Begin with the sheet →
Open in your mail app →
Open Claude →
Paste it there — or into any model you trust, including one running on your own device. You carry the words; nothing is injected, so no caution screens. The conversation is yours; LOFL never sees it. You send the final email yourself.
Copy pastes the inquiry anywhere (WhatsApp works too). Mail opens your own app, pre-filled. Draft-with-AI shows you a prompt to carry into a model of your choice.

Vivekī

विवेकी Full-fee
"AI will not hollow out our practice."
If you are a mid-sized NGO (50–500 staff) running livelihood, education, or health programs — with an active L&D function and donor pressure on responsible AI use. Refusal translation: practice protection — the parts of the mission that stay human. Tier 4 is where LOFL sells capability, not delivery.
Your size:
Saṃvāda→ Discernment Lab→ Integration Scaffolding→ Sevā
TierEngagementRange
Tier 1 · Saṃvāda
Half-day or full-day
Tier 1 is not a discovery call — it’s this instrument: Succinct Sāṃvāda (5 questions). The full 25-question sheet and the founder’s own answers are public.
Facilitated awareness conversation · real-question worksheet for the institution · shared vocabulary across leadership · recommendation memo · Padtalani add-on available at Discernment Lab stage (see module below)
₹75K–2L
Tier 4 · Sevā cohort
8 weeks · train-the-trainer
Train-the-trainer
3–5 certified internal facilitators · custom-designed Saṃvāda for the org · practice-protection rubric · beneficiary-data sovereignty checklist · board-ready refusal policy framework · facilitator community access — the NGO teaches itself thereafter
₹7–15L
Bundle · T1 + T4
Saṃvāda first, then Sevā · ~9 months
Full arc, one engagement
Everything in both tiers · continuity of facilitators from Saṃvāda → Sevā · the common path: buy capability once, cascade internally
₹8.5–18L
Payment posture: full fee, from L&D or program budget — donor-tagged budget is often easier than core.
Begin with the sheet →
Open in your mail app →
Open Claude →
Paste it there — or into any model you trust, including one running on your own device. You carry the words; nothing is injected, so no caution screens. The conversation is yours; LOFL never sees it. You send the final email yourself.
Copy pastes the inquiry anywhere (WhatsApp works too). Mail opens your own app, pre-filled. Draft-with-AI shows you a prompt to carry into a model of your choice.

Hastakāra

हस्तकार Funder-paid
"Our craft is not training data."
If you are an artisan cluster, women-led MSME network, or anchor NGO supporting craft livelihoods — community-structured units, never a sector — where collective IP and craft sovereignty are at stake. Refusal translation: craft IP — what the cluster will not feed to any model.
Cluster size:
Saṃvāda→ Discernment Lab→ Integration Scaffolding→ Sevā

Tier 1 is not a discovery call — it’s this instrument: the Succinct Sāṃvāda (5 questions). The full 25-question sheet and the founder’s own answers are public.

TierEngagementRange
Full program
6 months · reshaped Tier 1 + defensive sprint + Tier 4
All four tiers, funder-invoiced
Story-and-circle workshop series in the cluster’s language · defensive design sprint — what to protect, watermark, license · cluster IP & AI sovereignty charter · trained cluster facilitators for cascade · documentation other clusters in the tradition can reuse · evidence instrumentation at scaffolding stageNEW
₹15–40L
Payment posture: funder-paid (DC Handicrafts, state cluster programs, Ministry of Textiles, philanthropic capital). The cluster contributes nothing direct, or a token in-kind — space, food, facilitation support. The amount shown is what the funder underwrites. USD invoicing available for international funders (engagement letter until incorporation).
Begin with the sheet →
Open in your mail app →
Open Claude →
Paste it there — or into any model you trust, including one running on your own device. You carry the words; nothing is injected, so no caution screens. The conversation is yours; LOFL never sees it. You send the final email yourself.
Copy pastes the inquiry anywhere (WhatsApp works too). Mail opens your own app, pre-filled. Draft-with-AI shows you a prompt to carry into a model of your choice.
◆

Padtalani

पडताळणी · verification
New module · DPDP-aligned AI impact assessment

A structured assessment of what an AI deployment does to the personal data your institution holds — mapped to India's Digital Personal Data Protection Act before the tool is adopted, not after. Deliverables: data-flow inventory, DPDP obligation map, refusal boundaries, and a filed assessment the institution owns. DPDP has no certification body — compliance is demonstrated through documented operations, which is what this assessment produces and your institution keeps.

Credibility root: Mrunmai ran the DPIA on a biometric facial-recognition access-control product at NVIDIA, jointly with legal — the same assessment discipline, carried into civic India.

Standalone ₹2L Bundled inside the Discernment Lab — Nāgarika & Vivekī · bundle price stated at Saṃvāda
Who delivers this. Delivery is led by Mrunmai, founder of Leap of Faith Labs — a former NVIDIA product manager across enterprise AI, observability, and edge AI (including a DPIA on a biometric access-control product), with a Georgia Tech HCI certificate, Stanford GSB design-thinking training, and delivered design-thinking workshops. Solo delivery through first pilots is evidenced, not aspirational.
⚑ Register · open decisions & pressed record — internal; gates are M’s to press, this page never closes them
Unresolved. Padtalani bundle price inside the Discernment Lab — the standalone point is pressed at ₹2L; the stated discount is not yet stated. It wears a proposal ring until M names the number.
Position filed, parked. A free program for teachers — library-led, trace-filing, no fifth program named on this surface. parked · wakes on the sāṃvāda presented set returning
Held at its own gate. Sākṣī (the witness primitive, spec v0.1) is referenced by the record but deploys nowhere on this surface until M advances it. Its first live act is planned for the presented-set compilation, not here.
Decision — pressed, 2026-08-16, by M. Padtalani standalone: ₹2L, single point, band closed — bundled into the Discernment Lab at a stated discount. The no-certification-body line joins the module verbatim.
Decision — pressed, 2026-08-16, by M (via the sāṃvāda thread). Tier 1 is the instrument, not a call. All four programs now open with the Succinct Sāṃvāda; the founder answers first, publicly. Education status renders in honest present tense: in sāṃvāda. Invoicing note added: engagement letters until incorporation.
Decision — pressed, 2026-08-15, by M. Pilot names dropped from the offer surface — pilots are described by state, never named until the case study publishes. the names stay in the private record
Decision — pressed, 2026-08-15, by M. “Viveka Lab” dropped; Vivekī kept as the program name. The tier renders as “Discernment Lab.” Sanskrit tier name (e.g., Vimarśa) — proposal, M to press
Decision — pressed, 2026-08-15, by M. Founder named on the offering. The credibility line carries Mrunmai by name; role-anonymity is closed. a decision, not a marriage — reversible by supersession
Decision — pressed. v0.3 adopts the LOFL Design Language (Cormorant Garamond · Inter · paper · suit tokens), superseding v0.2's Fraunces/#FBF6EC surface. Structure, pricing ranges, sequencing, and brand pattern carry over unchanged. a supersession, never an overwrite — v0.2 stays in the record
Unresolved. Two size-band floors clamped on import from the demo build: Vivekī Sevā "small" floor raised 6L→7L and bundle "small" floor 8L→8.5L to stay inside the locked bands. M presses either the clamp (bands hold) or a band revision (floors drop).
Position filed, half-awake. Polyglot-translator × Shabdayatra. On-device, zero-egress translation encouraged as the visitor’s own practice; no toggle shipped. Discovered on publish-audit: /tools/shabdayatra.html already lives in the commons — half the wake condition predated the parking. The navigator now links it. The polyglot-translator itself stays on the horizon. remaining wake: a contributor, or M pressing the translator spec
Dissent — open, filed 2026-08-16. No-cookies note vs. host telemetry. This page sets no cookies and stores nothing — true and stated. But the host (Netlify) injects a performance-telemetry script (RUM) at serve time, sitewide. RUM is cookieless, so the note holds; the stronger claim — “this page phones no one” — does not, yet. Options: disable RUM in the Netlify site settings and earn the stronger sentence, or keep RUM and keep the claim scoped to cookies. M presses. the note is worded to stay true either way
Decision — pressed, 2026-08-16, by M (this thread). v0.10 carry-the-prompt: the URL-prefilled Claude link triggered Anthropic’s injection caution — correctly, since prefilling is injection. Rather than route around a platform safeguard, the AI action now hands the visitor the exact prompt, readable in full before anything moves: copy it, open a blank Claude, paste — or use any model they trust, including one on their own device. No injection, no caution screen, model-agnostic, and the human carries the words at every step. The safeguard taught the design.
Decision — pressed, 2026-08-16, by M (this thread). v0.9 inquiry panel: every CTA opens a three-way door — copy the composed inquiry (WhatsApp-ready), open it in your own mail app, or draft-with-AI (a Claude conversation prefilled to co-author; the visitor sends it themselves — nothing is ever sent for them, stated on the panel). The inquiry text is visible and editable before any action; the AI prompt instructs the model to fill nothing the visitor hasn’t said. Prices untouched.
Decision — pressed, 2026-08-16, by M (this thread). v0.8 buyer-path + integrity pass: jump-nav now switches tabs before scrolling; tables scroll on mobile; print forces the record open; every program card ends in a door (mailto CTA carrying program + size); Svarāj bands are selectable and feed the waitlist CTA; Nāgarika in-kind reopened honestly (first 2–3 pilots, one underway); bundle wording made public-honest ("stated at Saṃvāda" — the number stays M’s to press); Hastakāra gains USD invoicing line; the recordHash canonical string now ships inside the file, auditable by anyone. Prices unchanged; hash re-labeled for the new canonical text.
Decision — pressed, 2026-08-16, by M (this thread). Svarāj v0.7 language pass: philosophy translated to plain verbs against the AI_vibe_Coder market analysis. The trust gap (84% adoption / 46% distrust / 45% failed security tests) replaces abstraction as the frame; Week 4 becomes explicit review discipline; the artifact becomes a demo-you-can-show. Gate, bands, cohort shape, and strict trigger all unchanged — the analysis validates the sovereignty framing; only the clothes changed. Bands cross-checked against global cohort-course rates ($199–999/seat): ₹15/25/40K holds.
Decision — pressed, 2026-08-16, by M (this thread). v0.6 clarity layer: audience self-checks, sovereignty epigraphs, size-graded ranges, payment postures (incl. Sahakāra’s never-free token fee ₹25–50K), tier durations and full included-lists, Svarāj spine + bands named, jump navigation, internal notes folded into accordions. Headline bands unchanged.
the Sita Tai Test
“Would this price, this promise, and this refusal list read as fair to the person with the least power in the room?”
Every engagement on this page passes through this question before it is signed. If it fails, the price changes or the engagement doesn't happen.
Build your own offering. Everything here — the four-tier arc, the Sāṃvāda sheet, the refusal-inventory method — is published under CC-BY-SA 4.0 so it can be run again without us. Take the sheet from the commons, keep the license, and name what you change.
Reading in Marathi, Hindi, Tamil, or any language? Translate this page with an on-device model — zero-egress translation is itself LOFL practice, and your word-journeys — शब्दयात्रा is already live in the commons — feed the polyglot-translator on the horizon.
This page sets no cookies and asks for no login. Everything you type stays in your browser and leaves only when you press copy, mail, or send — to your own apps, never to ours. Close the tab and it’s gone.
recordHash 5031dfef… recompute: sha256 of the canonical string in this file’s source comment no cookies · no login CC-BY-SA 4.0 LOFL-AIAS 3 · AI Alongside scribed 2026-08-15 · supersedes lofl-pricing-navigator-v0.12 · explore@lofl.world