India AI Impact Summit 2026

ai-at-scale-driving-adoption-productivity-and-market-access-for-indian-smes-startups

7 min read

TRACK:Democratizing AI Resources
SPEAKERS
  • Anand Kamannavar, Applied Materials Inc
  • Ashok Chandak, IESA & SEMI India
  • Jyothis Indirabhai, NetraSemi
  • Navin Bishnoi, Marvell India and IESA
  • Preet Yadav, NXP Semiconductors
  • Sundeep Gupta, Alphawave SEMI (Qualcomm)
SESSION OVERVIEW

This session will examine adoption, productivity, and market access challenges for SMEs and startups, aligned with the People, Planet, and Progress framework through a focus on democratising access to AI resources. The discussion will explore approaches to enabling broader AI adoption through safe and trusted systems, while supporting economic growth and human capital development. It will also consider how scaling AI deployment can contribute to inclusive outcomes and wider participation across sectors.

VIDEO RECORDING

AI at Scale: Driving Adoption, Productivity and Market Access for Indian SMEs & Startups

Abstract

The panel examined why AI adoption in India has largely remained cloud‑centric, why that model may not serve the country’s unique privacy, cost and latency constraints, and how a shift toward edge‑optimized, domain‑specific AI can unlock massive opportunities for SMEs and startups. Drawing on global deep‑tech experiences, the speakers highlighted the need for new business models (shared data platforms, government‑backed grants, and “AI‑coach” data aggregation), stressed the importance of trustworthy and ethically‑aligned AI, and illustrated concrete ROI‑driving use‑cases in manufacturing, logistics, and services. The discussion concluded with audience questions on standards, liability, and data‑center architecture, underscoring the urgency of coordinated policy, industry, and academic action.

Detailed Summary

  • The session opened with a quick thank‑you from the moderator and a reminder that AI “is everywhere” – both in cloud services and increasingly at the edge.
  • Key Statistic (cited from Times of India): AI tools are used by ~70 % of software developers and content creators, whereas adoption among financial advisors (~20 %) and accountants/auditors (~36 %) lags far behind.
  • The panel framed the discussion around the three‑pillared “People, Planet, Progress” framework that underpins the conference agenda.

2. Phase 1 AI: Cloud‑Centric, High‑Cost Foundations

  • Anand Kamannavar described the current global AI landscape as Phase 1 – dominated by large LLMs and generative‑AI models hosted in centralized data‑center farms.
  • Benefits: rapid innovation, economies of scale, easy access for developers.
  • Drawbacks for India: high energy consumption, expensive bandwidth, privacy concerns (data leaves the premise), and limited relevance to local use‑cases.

3. Indian Opportunity: Edge AI & Domain‑Specific Optimisation

  • Jyothis Indirabhai (NetraSemi) argued that Edge AI is the true opportunity for Indian SMEs because:

    • Privacy: Data can be processed locally, avoiding transmission of sensitive information.
    • Latency: Real‑time response for surveillance, smart‑city, and manufacturing applications.
    • Cost: Reduces reliance on costly cloud‑compute and broadband fees.

  • He illustrated a surveillance market example: sending raw video streams to the cloud compromises privacy and inflates cost; an on‑device AI pipeline can filter, flag anomalies, and only transmit metadata.

  • Ashok Chandak reinforced that domain‑specific optimisation (custom chips, lightweight models) can create a “deployment‑ready platform” that is largely absent in India today. He estimated that only ~5 % of the surveillance‑AI market has been captured, implying a huge upside.

4. Global Deep‑Tech Lessons

  • Anand Kamannavar highlighted a stark truth: no Indian deep‑tech company is currently selling AI‑centric hardware or models at global scale.

  • He identified two systemic issues:

    1. Misaligned SaaS expectations – deep‑tech product cycles (chips, sensors) span 9‑18 months, not the rapid “pivot‑in‑days” mindset of software SaaS.
    2. Metrics mismatch – deep‑tech success should be measured by problem‑solving at scale, not speed to IPO.

  • Successful global examples:

    • Korea‑Taiwan joint funds focused on memory‑display (Korea) and semiconductor fab (Taiwan) ecosystems.
    • Companies that anticipated future bottlenecks (e.g., photonics firms forecasting data‑transfer limits 5‑10 years ahead).
    • The $20 billion acquisition of Grok (illustrating that deep‑tech pay‑offs often materialise after years of foundational work).

  • Practical takeaway – Indian innovators should identify high‑value problems that will emerge in the next 5‑10 years (e.g., data‑center energy, high‑precision metrology) and invest in strong IP stacks early.

5. Funding Models & Shared Infrastructure

  • Navin Bishnoi shifted the focus to the enabling ecosystem: compute, capital, and demand aggregation.

    • Government schemes exist (e.g., AI‑Coach platform) that aggregate anonymised industry data, allowing startups to train on a larger, representative dataset without exposing proprietary information.
    • Shared GPU access at subsidised rates is already offered, but awareness remains low.

  • Preet Yadav (NXP) underlined the manufacturing sector’s growth potential: a 40 % CAGR in AI‑driven manufacturing through 2030.

  • Sundeep Gupta emphasized human‑in‑the‑loop validation: even when AI suggests a chip layout, a human designer must verify before tape‑out, ensuring safety and liability remain under human control.

6. Ethics, Trustworthiness & Responsible AI

  • Preet Yadav opened the ethics segment with an automotive analogy: a defective brake system is intolerable; AI used in vehicles must meet the same “day‑one‑to‑end‑of‑life” reliability.

  • Jyothis Indirabhai highlighted bias and data‑poisoning risks:

    • Training data reflecting gender stereotypes (e.g., “men work, women don’t”) can embed bias into AI‑driven decision‑making, especially in hiring or content moderation.
    • Continuous monitoring and open feedback loops are necessary to identify and remediate such bias.

  • Anand Kamannavar added that trust hinges on two pillars:

    1. Edge‑based privacy – keeping data on‑premise.
    2. Cost of intelligence approaching zero – as LLMs become commoditised, the differentiator will be how well AI integrates with existing workflows (e.g., on‑shop‑floor manufacturing).

  • Consensus: Ethics must be baked‑in from conception, not retro‑fitted; standards and certification (akin to automotive functional safety) are required for each sector and region.

7. Concrete ROI‑Driving Use‑Cases

SectorUse‑CaseMeasured Benefit
Manufacturing (SMEs)Inventory optimisation – AI predicts demand, reduces over‑stock/under‑stockLower holding costs, higher service level
Predictive maintenance – vibration / energy‑spike analysis alerts before failureUp to 30 % reduction in unplanned downtime
Logistics & workflow orchestration – AI routes raw‑material flow across multiple production stagesFaster throughput, reduced bottlenecks
Legal ServicesDocument review & contract analysis – AI flags errors in minutes vs weeks of manual review1000× speed‑up, higher accuracy
Software DevelopmentCode generation & testing – LLMs write boilerplate, run simulationsHours → minutes for prototype cycles
HealthcarePerinatal monitoring – AI assists delivery‑room decision supportZero tolerance for missed alerts (life‑critical)
  • Across these examples, initial investment ranges from ₹ few lakhs to ₹ tens of lakhs, but ROI materialises quickly once the AI model is embedded in the operational workflow.

8. Impact on Jobs & Skills

  • Preet Yadav expressed concern that entry‑level roles in coding, legal drafting, and marketing may be dramatically reshaped; graduates need AI‑augmented skillsets (prompt engineering, model evaluation) rather than pure manual execution.

  • Panelists agreed that human expertise will shift towards validation, ethical oversight, and systems integration – roles that AI cannot fully automate.

9. Liability, Standards & Policy

  • Audience Q&A – Liability (Lawyer Karthik):

    • Analogy to autonomous vehicle crashes – liability resides with the service provider (e.g., Waymo/Google) while the human operator holds residual responsibility.
    • Conclusion: Legal frameworks will evolve to assign product‑liability to AI providers (or the firms that deploy them), not the end‑user.

  • Audience Q&A – Standardisation (Arun Sharma):

    • Current gap: Unlike telecom or 5G, AI lacks a mature standard‑setting process.
    • Panel noted that standardisation will emerge as the ecosystem matures (similar to the transition from GPU‑only compute to hybrid edge‑cloud models).
    • Government role: early‑stage policy (privacy, security) can guide industry toward India‑centric standards, creating a competitive advantage for local chip and software vendors.

  • Audience Q&A – Data‑Center & Global Access (Yash, student):
    – Hybrid compute model (on‑premise for sensitive data + cloud for metadata) can meet latency and sovereignty requirements.

10. Closing Synthesis

  • Key consensus – AI’s future in India hinges on edge‑first, domain‑specific, ethically‑grounded deployments that directly solve pain points for SMEs (inventory, downtime, logistics).
  • Policy recommendation – Accelerate AI‑Coach data aggregation, broaden subsidised GPU access, and formalise privacy‑by‑design standards.
  • Call to action – SMEs should partner with startups that provide plug‑and‑play edge AI modules; startups should leverage government grants to build shared data platforms and industry‑specific IP.

Key Takeaways

  • Phase 1 AI (cloud‑only) is insufficient for India; the next wave must focus on edge AI that respects privacy, latency, and cost constraints.
  • Domain‑specific optimisation (custom chips, lightweight models) offers a ~5 % market capture now, indicating a large untapped opportunity for Indian startups.
  • Global deep‑tech success stems from long‑term problem identification and robust IP creation—not from rapid SaaS‑style pivots.
  • Shared infrastructure (government‑backed data pools, subsidised GPU farms) is essential to give SMEs access to high‑quality training data without breaching confidentiality.
  • Ethics and trust must be built in from the start; sector‑specific safety standards (akin to automotive functional safety) are required to prevent bias, data poisoning, and catastrophic failures.
  • Concrete ROI use‑cases—inventory optimisation, predictive maintenance, logistics orchestration, and legal document analysis—show AI can deliver significant cost savings and speed gains for SMEs.
  • Job market will shift: routine, entry‑level tasks are at risk, while human‑in‑the‑loop validation, ethical governance, and system integration become premium skills.
  • Liability frameworks will evolve to hold AI service providers accountable, mirroring trends in autonomous‑vehicle regulation.
  • Standardisation and policy are nascent but critical; early Indian‑centric standards around privacy, security, and edge‑AI performance can give the nation a competitive edge.
  • Hybrid edge‑cloud architectures solve the double‑edged problem of data sovereignty and global accessibility, ensuring Indian data‑centres can serve both domestic and international users efficiently.

The panel underscored that AI is not merely a tool for code generation or content creation—it is a catalyst for transforming India’s manufacturing floor, legal services, and SME ecosystems, provided that deployment is ethical, trustworthy, and optimised for the local context.

See Also:

Graph View

Backlinks