Leveraging AI4All: Pathways to Inclusion

Contents

Executive Summary

This panel discussion at the AI Impact Summit addresses the critical gap between AI innovation and equitable access, emphasizing that technology alone cannot drive inclusion without deliberate design, institutional capacity, and investment strategies. The speakers, representing global perspectives from India, Rwanda, and Meta, present concrete case studies demonstrating that inclusive-by-design approaches yield better products, stronger business outcomes, and sustainable societal impact—contrary to the perception that inclusion is a CSR afterthought or late-stage add-on.

Key Takeaways

Inclusion is a Business Imperative, Not a Moral Add-On
- Corporations are shifting from "prove AI works" to "make AI sustainable and inclusive." Inclusive-by-design is becoming boardroom priority because it's economically sound, not just ethical.
Three Interconnected Pillars Drive Sustainable AI Inclusion: Design, Access, and Investment
- Design (participatory, universal): embed from start.
- Access (real-world conditions): low bandwidth, offline-first, multilingual, low-literacy.
- Investment (aligned incentives): government anchor buying, open standards, procurement policy innovation.
Language and Cultural Localization Cannot Wait for Phase Two
- Building English-only MVPs and planning Hindi/local language support for later is a recipe for failure. Operating in local language from day-one is both a user need and a market-unlock strategy (70%+ customer alienation if ignored).
Institutional Capacity and Ecosystem Matter More Than Technology Alone
- Technology is force multiplier, but without training (Adalati Academy curriculum), procurement standards, and institutional expertise, adoption stalls. Government-backed capacity building is a high-leverage investment.
Real-World Constraints Are Design Constraints, Not Excuses
- Humanitarian response systems must work offline. Agricultural advisory must work on feature phones and poor connectivity. Judicial transcription must handle Indian accents and legal jargon. These are non-negotiable design parameters, not afterthoughts.

Key Topics Covered

Design Principles for Inclusion – Participatory design, universal design, and accessibility-first innovation
Access & Digital Equity – Connectivity gaps, last-mile delivery, language localization, and low-resource environments
Institutional Barriers & Solutions – Procurement policy innovation, nonprofit models, and capacity building
Business Case for Inclusion – ROI, cost-benefit tradeoffs, and market incentives
AI Applications in Justice – Judicial transcription, case tracking, and information access via WhatsApp chatbots
Language & Localization – Multilingual AI, low-resource language datasets (Kinyarwanda), and cultural adaptation
Hardware & Assistive Technology – Ray-Ban AI glasses with Be My Eyes, smart wearables for accessibility
Government & Policy – DPI (Digital Public Infrastructure), public procurement for innovation, regulatory alignment
Sectoral Applications – Agriculture, healthcare, education, finance, justice, refugee aid
Investment Models – Government anchor buying, ecosystem development, data accessibility initiatives

Key Points & Insights

Access is Multi-Layered, Not Technical Alone
- Good technology without addressing connectivity, skills, user interfaces, and community needs will not drive inclusion. The "last mile gap" requires focus on infrastructure and cultural fit, not just AI capability.
The Purple Economy is Real Business, Not Charity
- The assistive technology market for people with disabilities represents $150 billion in India alone. These are purchasing customers, not marginal populations. Building for them is sound business strategy, not charitable obligation.
AI Products Are Stuck in Pilot Purgatory
- Many solutions fail to scale due to insufficient focus on surrounding systems: last-mile diffusion, funding, and institutional support. Scaling requires ecosystem thinking, not just technical prowess.
Language is Foundational for Inclusion
- All AI systems must operate in local languages and cultural contexts. Whether banking systems, educational tutors, or advisory tools, localization from day-one is non-negotiable, not a phase-two addition.
Institutional Capacity Multiplies Impact
- Government technical expertise directly translates to better procurement standards and procurement decisions. As institutions build capability, adoption and standards improve exponentially (seen in Kerala's mandatory judicial transcription requirement).
Inclusive Design Yields Better Products for Everyone (The Curb-Cut Effect)
- Accessibility features initially built for people with disabilities (e.g., screen readers, captions, voice interfaces) become universal benefits (wheelchairs → prams/luggage, flatbed scanners → document digitization). Inclusive design is good design.
Participation is Non-Negotiable
- "Nothing about us without us" — involving end users (Asha workers, judges, farmers, people with disabilities) in design prevents costly late-stage failures. Direct field observation and co-design are essential, not optional.
Nonprofit Models Lower Friction & Align Incentives
- Nonprofits (e.g., Adalat AI in courts) reduce institutional barriers (data security concerns, budget constraints, accountability issues) and enable faster adoption. They also improve procurement design for future commercial solutions.
Cost & Tradeoffs Are Real but Solvable
- Diverse datasets and localization require 3:1 data-to-AI spending. However, as governments provide accessible datasets (e.g., India AI Kosh), economies of scale kick in, making inclusion financially viable for enterprises.
Accessibility-First Innovation Drives Competitive Advantage
- Companies that embed accessibility from inception (Ray-Ban glasses with Be My Eyes, judicial transcription) gain market leadership and avoid costly redesigns. Short-term ROI pressure and minimum viable product thinking replicates past errors (1990s GUI inaccessibility).

Notable Quotes or Statements

Nirmal (Report Presenter): "It's not a charitable cause. It's a simple business proposition" — referring to the $150B assistive tech market for people with disabilities.
Nirmal: "The question or doubt really isn't that AI is going to expand access and opportunity. The question is whether ecosystems will now choose to build systems that are required to make this expansion durable, equitable and sustainable."
Arjuna (Enterprise AI Lead): "If you don't do inclusive by design, you are leaving money on the table and it's just plain smart good business."
Arjuna: "If you do it this way [English-only phase-one], most of the folks calling you speak Hindi... you're alienating 70% of your customers... why are you even attempting it?"
Augusta (Meta): "Curb cuts, which are divots on sidewalks for wheelchairs, benefited everyone using prams, strollers, or luggage. They just make cities better. That's inclusive design."
Augusta: "Nobody has a crystal ball... the key thing is learning to be nimble... avoid the temptation of eating the candy before [showing painkiller or vitamin value]."
Arjua (Adalat AI): "Justice in these settings is really not a question of law. It's become a question of logistics."
Arjua: "Judges don't know how to update their Chrome browsers" — highlighting that technical capability of end-users, not just technology, determines success.
Olivia (Rwanda AI Hub): "Building a plane as we fly it" — describing Rwanda's approach to simultaneous DPI and AI development with emerging local language datasets.

Speakers & Organizations Mentioned

Speaker	Organization	Role/Focus
Nirmal	(Report Lead)	Report on AI4All pathways; findings on design, access, investment
Rutu Ja Paul	Guidepoint / Panel Moderator	Partner moderating enterprise and design discussions
Arjua	Adalat AI	Founder; judicial transcription, case tracking, WhatsApp chatbot for justice access
Olivia	Rwanda AI Scaling Hub	Hub lead; local language AI, DPI integration, sectoral implementation
Arjuna	(Enterprise AI Consulting)	Advises corporates on inclusive AI design; humanitarian, banking, insurance cases
Augusta	Meta	Engineering lead; Ray-Ban AI glasses, Be My Eyes, multimodal AI accessibility
(Unnamed)	Government (India/Rwanda context)	Policy on public procurement for innovation, DPI alignment

Technical Concepts & Resources

Digital Public Infrastructure (DPI) – Mature, trusted foundational systems (e.g., India's payment/ID systems) enabling AI deployment; Rwanda building DPI simultaneously with AI.
Multilingual AI & Low-Resource Languages
- Kinyarwanda language models and datasets for Rwanda AI Hub.
- Legal transcription understanding Indian accents, legal jargon (e.g., "res judicata").
- Voice, text, and vision LLMs adapted for local context.
Offline-First & Low-Bandwidth Design
- Humanitarian crisis AI working without cloud connectivity.
- WhatsApp-based justice chatbot (accessible without high-speed internet).
- Newborn measurement tool (Shishuman) using offline photo/video analysis.
Vision & Multimodal LLMs
- Ray-Ban glasses with Be My Eyes (visual impairment assistance).
- Document OCR, transcription, and summarization (with caution on safety).
- Vision LLMs for accessibility wrapping (e.g., video captioning for sign language).
Datasets & Data Governance
- India AI Kosh – government-provided diverse datasets for inclusive AI training.
- Dataset cleaning and curation cost (3:1 data-to-AI spending ratio in typical projects).
- Participatory data collection involving end-user communities.
Procurement Models
- Public Procurement for Innovation (PPFI) – agile, adaptive RFP process for fast-moving tech.
- RFP design informed by nonprofit pilots (e.g., Adalat AI's early court work improving Kerala's official procurement).
Use Cases & Tools
- Shishuman (Vadwani AI) – Photo/video-based newborn measurement for frontline health workers.
- Ray-Ban Meta Glasses with Be My Eyes – Multimodal AI for visual navigation and accessibility.
- Yes To Access App – Community database of disability accessibility in buildings/public spaces.
- Adalati Academy – Judicial training curriculum integrating technology literacy.
Universal Design Principles
- Screen readers, gesture interfaces, and voice input built in from inception.
- Curb-cut effect – accessibility features benefit all users.
- "Nothing about us without us" – participatory design methodology.

Document Type: Conference Panel Discussion
Event: AI Impact Summit
Key Organizers: Meta (noted as partner)
Tone: Practical, evidence-based, emphasizing business case alongside social impact
Audience: Corporate leaders, policymakers, AI practitioners, social entrepreneurs