Blog

1. The Strategic Inflection Point: Assessing the Current Landscape

India’s technology services sector is the undisputed cornerstone of our national economy, contributing $265 billion in annual revenue and accounting for 7% of the Gross Domestic Product (GDP). Historically, this industry has been an outperformer, consistently outpacing global growth by 200 to 300 basis points. However, the paradigm that sustained this—centered on cost-arbitrage, labor scale, and incremental productivity—has reached its structural limit. We have entered a “post-pandemic reset” that demands a total reimagination of our industrial strategy.

The sector’s evolution from 2015 to 2024 reveals a cooling engine that requires urgent intervention:

• Rapid Adoption (2015–2020): A period of steady 6–7% CAGR driven by cloud migration and enterprise modernization.
• COVID Surge (2020–2022): A hyper-acceleration phase (11–13% CAGR) fueled by the global rush to remote work and digital transformation.
• COVID Reset & AI Discontinuity (2022–2024): A critical slowdown where growth has moderated to 7–8% industry-wide.

The “So What?” Layer The most alarming indicator is found among our Tier-1 India-headquartered providers, whose growth has decelerated to a mere 3–5%. This is not a cyclical dip; it is a structural warning. If our industry champions continue at this pace, the “Viksit Bharat 2047” vision—which requires an $11 trillion GDP by 2035—is under immediate threat. We cannot achieve national technological sovereignty on the back of a slowing “billable hours” model. The transition to an AI-native architecture is no longer optional; it is a national security mandate.

——————————————————————————–

2. The $300 Billion Challenge: Quantifying the Aspiration Gap

To secure our economic future, the technology sector must reach $750–850 billion in revenue by 2035. This target is a fiscal necessity for the sector to sustain its 7–8% share of a projected $11 trillion GDP. Current trajectories, however, reveal a velocity deficit that risks a permanent loss in global market share, which must expand from 20% to over 25% to maintain leadership.

Path to 2035: Revenue Projections vs. Aspirational Targets

Growth Component	Projected Revenue Contribution (2035)	Required CAGR
Current Base (2024)	~$265 Billion	–
“Protect the Core” (Dual-Track)	$150–180 Billion	4–5%
“Pivot to New Vectors” (New TAM)	$100–120 Billion	6–7%
Projected Size on Current Path	$500–580 Billion	Sub-Optimal
The Aspiration Gap (Shortfall)	$250–300 Billion	The Velocity Gap
Total Aspirational Target (2035)	$750–850 Billion	10–11%

The “So What?” Layer Closing the $250–300 billion gap requires a “Dual-Track Strategy.” We must protect our $265 billion base while simultaneously capturing adjacent Total Addressable Markets (TAM) in AI and software. This is not merely about revenue; it is about Geopolitical Leverage. In an era where technology is the central instrument of economic diplomacy, failing to close this gap means ceding the “architect” role to global competitors, effectively turning India back into a high-tech “back office” for foreign IP.

——————————————————————————–

3. Growth Vectors: The Five Frontier Plays for Global Leadership

Bridging the aspiration gap requires a pivot to a full-stack digital enablement model. Indian firms must move beyond routine delivery and institutionalize programmatic M&A—benchmarking against firms like Accenture (50+ deals annually)—to rapidly acquire niche AI and digital engineering capabilities.

• Agentic AI Play Moving from effort to outcomes by building hybrid “human + agent” service models.
  • Services as Software: Productized services for the software lifecycle ($15–20B potential).
  • Hybrid Workforce Models: Domain-specific AI agents substituting traditional labor spend ($25–30B potential).
  • Total Projected Potential (2035): $40–50 Billion
• Software/SaaS Play Rearchitecting CRM, ERP, and DataOps value pools while capturing cybersecurity hotspots to establish India as the global SaaS capital.
  • Projected Potential (2035): $20–25 Billion
• Infrastructure Play Establishing India as the world’s data services hub. We must scale capacity from 1.4 GW to 10–12 GW and increase GPU compute share from 4% to 20%.
  • Projected Potential (2035): $10–15 Billion
• Innovation (ER&D) Play Targeting the $1.1 trillion global R&D market by positioning India as the pre-certification and design hub for MedTech, Semiconductors, and Defense.
  • Projected Potential (2035): $25–35 Billion
• India for India Play Capturing domestic demand through multi-lingual AI agents and customized solutions for MSME credit and agri-advisory.
  • Projected Potential (2035): $40–60 Billion

The “So What?” Layer The transition from “Human-only” to “Human + Agent” models is expected to deliver non-linear efficiency gains of 70% or more. This rids the industry of its dependence on headcount-linked growth, transforming the margin profile and allowing Indian firms to dominate high-margin IP pools.

——————————————————————————–

4. Legislative Reform: The National Tech-Services Single Window

Regulatory friction is the greatest tax on innovation. We must mandate a National Tech-Services Single Window to facilitate radical Ease of Doing Business (EODB) and reduce the friction of global trade.

Requirements for the Single Window:

• Infrastructure Expediency: Fast-track approvals for land, power, and hardware specifically for GPU-enabled data centers to attract large-scale investment.
• IP & Export Streamlining: Radical simplification of intellectual property registration and reduction of export compliance complexity for SaaS.
• Fiscal Friction Removal: Immediate clarification on ESOP taxation and resolution of cross-border payment hurdles that currently force Indian SaaS firms to flip their headquarters abroad.
• Specialized Facilitation Units: Dedicated units to manage supplier partnerships for sovereign GPU infrastructure and high-quality power access (including renewable/nuclear).

The “So What?” Layer This reform shifts India from being a “high-friction service provider” to a “seamless global technology partner.” Without this structural ease, we cannot capture the $1.1 trillion global R&D and SaaS spend pools that prioritize speed-to-market over labor cost.

——————————————————————————–

5. Human Capital Transformation: The Coordinated AI Talent Mission

Talent is no longer just a resource; it is a strategic asset for trade negotiations. To counter the widening shortage of LLMOps and research specialists, a nationally coordinated AI talent mission is non-negotiable.

Critical Talent Shifts Checklist:

• [ ] Judgment Over Code: Shift training from pure coding to high-level problem-solving and business judgment.
• [ ] Emerging Specializations: Scale reskilling for Prompt Engineers, GenAI researchers, and LLMOps specialists.
• [ ] ANRF Missions: Embed “AI Literacy” through academic-industry partnerships via the Anusandhan National Research Foundation.
• [ ] Learnability: Institutionalize adaptive capability across all management tiers to manage AI-native workflows.

The “So What?” Layer This mission is the bedrock of Digital Sovereignty. By cultivating indigenous expertise, India can develop domain-specific Small Language Models (SLMs) for sensitive sectors like banking, telecom, and defense. This ensures our digital economy is powered by local intelligence rather than foreign-governed AI stacks.

——————————————————————————–

6. Funding the Future: R&D Incentives and Capital Allocation

Industry R&D must scale to 1–2% of revenues to move beyond routine delivery into defensible IP. The state must provide a fiscal framework that distinguishes between technical innovation and standard operations.

Fiscal and Investment Mandates:

• National Digital R&D Framework: Provide shared infrastructure, including national GPU clusters, to lower barriers for smaller firms.
• Targeted R&D Support: Capital and tax incentives specifically for frontier technologies like quantum computing and bioengineering.
• ANRF Partnership Models: Encourage industry consortia to join as limited partners in national R&D missions.

The “So What?” Layer Policy must explicitly differentiate “Technical R&D” (solving technical uncertainty to create reusable IP/algorithms) from “Routine Engineering” (client-specific maintenance). This distinction is vital for claiming tax incentives and ensures that national capital is building a globally competitive IP portfolio rather than subsidizing standard business processes.

——————————————————————————–

7. Conclusion: Securing India’s Position as the Global AI Architect

The journey to an $850 billion technology sector is a national imperative. We are at a junction where the traditional “effort-based” models of the past thirty years are being replaced by an “outcome-linked,” AI-native reality.

The Case for Action:

1. The Aspiration Gap: We face a $250–300 billion shortfall that threatens our 2047 economic goals; status quo growth is a failure of strategy.
2. Model Transformation: We must move from effort-based billable hours to productized, human-agent hybrids with 70% efficiency targets.
3. Unified National Effort: Success requires a synchronized push where industry leads in R&D and M&A, and the government enables through the Single Window and AI Talent Mission.

India possesses the talent and the scale to lead this transition. By acting decisively now, we will evolve from the world’s “back office” into the world’s AI-native architect, securing our place as the primary builder of the global digital order.

1. Introduction: The Uncomfortable Evolution of Trust

In our hyper-connected landscape, the foundational concept of “trust” has undergone a radical, uncomfortable evolution. We have officially entered an era where “seeing is believing” is a dangerous fallacy. As a senior specialist, I’ve watched the traditional security perimeter dissolve as sophisticated synthetic media and industrial-scale crime syndicates bridge the gap between digital fiction and financial reality.

The numbers from the 2025 scorecard are a sobering wake-up call: global fraud losses have eclipsed the $1 trillion mark. Even more distressing is the abysmal 4% recovery rate for victims. In a world where only four cents of every stolen dollar returns home, the message is clear: our traditional defenses aren’t just leaking; they are obsolete. The following analysis explores five critical lessons that define the new front lines of financial defense.

2. The End of the “Blame the Victim” Era

For decades, financial institutions treated fraud as a “user error” problem, leaving the burden of loss on the individual. That era is over. A wave of “Failure to Prevent” regulations is sweeping the globe, forcing banks and telecommunications giants to move from passive observers to proactive protectors.

This regulatory shift is transformative, fundamentally changing the economics of fraud for institutions:

• United Kingdom: Under new PSR rules, the liability for scam losses is now split 50/50 between the sending and receiving banks, incentivizing both ends of a transaction to intercept fraud.
• Australia: The Scam Prevention Framework (SPF) mandates that banks, telcos, and digital platforms take “reasonable steps” to prevent and report scams or face heavy civil penalties.
• Singapore: The Shared Responsibility Framework (SRF) has expanded the liability circle to include telecommunication companies, recognizing their role in the fraud delivery chain.
• European Union: PSD3 now expects institutions to assume liability for impersonation scams, particularly where bad actors pretend to be bank representatives.

As the Feedzai 2025 Scorecard correctly identifies:

“Financial institutions are moving from leaving victims to bear the burden of scam losses to having financial institutions and other players assume greater responsibility.”

3. The “AI Hype” Correction: From Autonomous Agents to Intelligent Co-Pilots

The market has faced a necessary reality check regarding “Agentic AI.” Early 2025 predictions suggested that fully autonomous AI agents would be running the show by now. However, as research from MIT highlights, many of these autonomous Proofs of Concept (POCs) fell short of the high bar required for financial security.

Instead of replacing the human investigator, we have seen the rise of the “Intelligent Co-pilot.” Banks are now focusing on integrating Generative AI directly into existing workflows to automate complex data summaries, recommend rules, and triage alerts. This co-pilot model mitigates the unpredictable risks of fully autonomous agents while supercharging human teams, allowing them to synthesize oceans of data in milliseconds. The focus has shifted from replacing the investigator to maximizing their efficacy through augmented intelligence.

4. Why Your “Voice” is the Newest Vulnerability

Artificial Intelligence has effectively turned personal biology into a high-risk liability. Fraudsters are now leveraging AI voice cloning and deepfakes to impersonate trusted figures with terrifying precision. By scraping just a few seconds of audio from social media or voicemails, bad actors can mimic tone, pitch, and emotional inflection to bypass human intuition.

However, the threat isn’t just social; it’s technical. Industry leaders like Microblink have identified the rise of “virtual camera injection.” This is a sophisticated vector where threat actors inject deepfake video streams directly into the KYC (Know Your Customer) process, tricking standard liveness checks by presenting a synthetic image as a “real-time” selfie. We saw the devastating potential of this in the $25 million Hong Kong CFO impersonation incident, where an entire video call was faked to authorize a massive transfer.

As experts at Banesco USA warn:

“These technologies are now easy to use, giving average threat actors the power to exploit human trust through audio manipulation.”

To counter these high-tech injections and clones, we must adopt “low-tech” authentication anchors:

• Establish “Family Safe Words”: A secret, offline phrase to verify identity during any high-stress call.
• The “Hang up and Call Back” Rule: Never trust an inbound call that demands urgency. Hang up and dial a verified, pre-existing contact number.
• Ask “Smart Questions”: Challenge the caller with questions requiring personal anecdotes or information about private conversations that aren’t available on social media.

5. The “Consortium” Advantage: Replacing Silos with Shared Intelligence

Criminal syndicates operate as highly coordinated, industrial-scale enterprises. To beat them, financial institutions must abandon their historical data silos in favor of “Consortium Analytics.”

According to Nasdaq/Verafin, the scale of the problem demands a network-level response. With ACH payment values growing by $1 trillion annually and check fraud losses hitting an estimated $21 billion, individual banks can no longer see enough of the board to win. By leveraging shared intelligence—exemplified by the Feedzai and Mastercard partnership—institutions can identify high-risk payees across thousands of banks before the money ever moves.

This is the cornerstone of “Cyber-Fraud Fusion.” For too long, a dangerous blind spot existed: Cybersecurity teams saw the how (network traffic and anomalies) but lacked payment context, while Fraud teams saw the who (behavioral and payment data) but were blind to the technical delivery. Fusion closes this gap, creating a “network effect” where a scam identified at one institution instantly hardens the defenses of the entire consortium.

6. Regulation with Teeth: The California ADMT Shift

The regulatory landscape will tighten significantly on January 1, 2026, with the enforcement of new Automated Decision-Making Technology (ADMT) requirements under the CCPA/CPRA. This shift grants consumers unprecedented transparency and control over the algorithms that govern their financial lives.

Consumer Right	Description
Right to Know	Consumers can demand information regarding the specific logic, parameters, and intended outcomes of the ADMT.
Right to Opt-Out	The ability to refuse decisions made without human involvement in “significant” areas like employment, credit, housing, or healthcare.
Pre-Use Notices	Mandatory plain-language explanations provided before the technology is deployed to process a consumer’s data.

7. Conclusion: Toward a “Know Your Actor” (KYA) Future

The industry is moving beyond the era of simple “Know Your Customer” (KYC) checklists. As Microblink has signaled in their 2026 roadmap, we are entering the “Know Your Actor” (KYA) era. In a world of synthetic identities and virtual camera injections, verifying that a document is “real” is no longer sufficient. KYA represents a shift toward “Identity Intelligence”—a continuous, multi-modal assessment of the intent and authenticity of the actor behind the screen.

As we look toward 2026, we must answer a fundamental question: In an era where AI can mimic any voice or face perfectly, are we prepared to move our primary source of trust from digital signals back to human-centric, verifiable relationships? The future of fraud prevention will not be found in better passwords, but in our ability to distinguish the human actor from the machine mimicry.

As generative AI adoption accelerates, most AI product teams default to proprietary large language models (LLMs) such as ChatGPT or Claude during early product testing. While these models offer exceptional performance, relying on them exclusively during the Proof of Concept (PoC) phase introduces hidden risks related to cost, architecture, compliance, and long-term scalability.

This whitepaper proposes an open-source-first hypothesis: AI products validated using open-source LLMs during the PoC phase result in more realistic cost models, stronger security posture, and more resilient system architectures than products validated exclusively using proprietary LLM APIs.

We argue that open-source LLMs are not a compromise for early testing—but a strategic advantage.

---

1. Introduction: The PoC Fallacy in AI Product Development

In traditional software development, PoCs exist to reduce uncertainty early. However, in AI product development, PoCs often do the opposite: they mask risk instead of revealing it.

This happens when teams:

• Use highly capable proprietary models from day one
• Ignore infrastructure realities
• Assume future costs and compliance will “work out later”

As a result, many AI initiatives succeed in demo environments but fail in production planning.

---

2. The Core Hypothesis

Instead of relying only on proprietary LLMs during early AI product testing, teams should primarily use open-source LLMs to validate feasibility, economics, security, and architecture—and introduce proprietary models only at later stages if required.

This hypothesis reframes PoCs not as intelligence demonstrations, but as risk discovery mechanisms.

---

3. Why Proprietary-First PoCs Are Misleading

3.1 Artificial Performance Inflation

Proprietary models:

• Mask poor data quality
• Compensate for weak retrieval pipelines
• Hide architectural inefficiencies through sheer model capability

This leads to false confidence.

---

3.2 Unrealistic Cost Assumptions

Early PoCs rarely reflect:

• Real token volumes
• Peak concurrency
• Long-term usage patterns

By the time costs are modeled accurately, architectural decisions are already locked in.

---

3.3 Vendor-Driven Architecture Lock-In

Designing around a single API leads to:

• Prompt-centric systems
• Weak abstraction layers
• Limited portability across models

This increases switching costs later.

---

3.4 Incomplete Security & Compliance Validation

SaaS LLMs make it difficult to validate:

• Data residency
• PII exposure paths
• Internal security audits
• Client-specific compliance constraints

These issues often surface after business commitments are made.

---

4. The Case for Open-Source LLMs in PoCs

4.1 PoCs Are About Feasibility, Not Perfection

At the PoC stage, teams must answer:

• Does the product work with real data?
• Is the experience useful?
• Can it scale economically?
• Is it deployable within constraints?

Open-source LLMs are more than sufficient to answer these questions.

---

4.2 Cost Realism from Day One

Open-source deployments force teams to confront:

• Infrastructure costs
• Latency tradeoffs
• Throughput limits
• Optimization requirements

This leads to better investment decisions earlier.

---

4.3 Security-First Validation

With open-source LLMs, teams can:

• Run models on-prem or in VPC
• Enforce zero data egress
• Validate encryption, logging, and access control
• Pass enterprise security reviews earlier

---

4.4 Architecture-Driven Product Design

Open-source testing encourages:

• Explicit RAG pipelines
• Model orchestration layers
• Observability and monitoring
• Fallback and degradation strategies

These systems are inherently more production-ready.

---

4.5 Model-Agnostic Thinking

Open-source-first PoCs promote:

• Model interchangeability
• Hybrid deployments
• Vendor flexibility
• Future-proof architectures

The product becomes independent of any single model provider.

---

5. Recommended PoC Validation Framework

Phase 1: Open-Source Validation

Purpose: Truth discovery Focus: Feasibility, cost, architecture, security

Validate:

• Data readiness
• Retrieval quality
• User value
• Latency and infra constraints

---

Phase 2: Selective Proprietary Benchmarking

Purpose: Capability benchmarking Focus: Quality uplift analysis

Test proprietary models only to measure:

• Reasoning improvements
• Edge-case handling
• Language nuance
• Multi-step task performance

---

Phase 3: Informed Production Decision

Choose deliberately between:

• Open-source only
• Hybrid deployment
• Proprietary with fallback strategies

---

6. Addressing Common Objections

“Open-source models are worse”

Yes—but PoCs don’t need perfection, they need realism.

---

“Clients expect GPT-level quality”

Clients ultimately expect:

• Predictable costs
• Secure systems
• Compliance readiness
• Reliable delivery

---

“Open-source increases engineering effort”

That effort reveals:

• Scaling bottlenecks
• Infra constraints
• Operational risks

Which is exactly what PoCs are meant to uncover.

---

7. Strategic Implications for Organizations

Organizations adopting an open-source-first PoC approach gain:

• Lower long-term risk
• Better capital efficiency
• Stronger negotiation leverage with vendors
• More defensible AI architectures

This approach shifts AI development from vendor-led experimentation to engineering-led product design.

---

8. Conclusion

Open-source LLMs are not a replacement for proprietary models—they are a filter for truth.

By using open-source LLMs during the PoC phase, organizations:

• Reduce uncertainty earlier
• Avoid costly architectural rewrites
• Make informed production decisions
• Build AI products that survive real-world constraints

We use open-source LLMs in PoCs not because they are better than GPT—but because they reveal reality earlier.

---

• 2