The $2.5 Trillion Problem — And Why AI + Human Expertise Is the Only Credible Answer

---

Asia's small businesses power nearly half the region's GDP. Yet most of them can't get a bank loan. I recently sat through a demo of what RupeeBoss is building to fix that — and I came out genuinely excited. Not the polite, LinkedIn-friendly kind. The kind where you can't stop thinking about what this means at scale.

I'm a board member investor in RupeeBoss. Here's what I saw — and why the context matters enormously.

---

The Backbone of Asia's Economy Is Starving for Credit

Across Asia and the Asia-Pacific region, Micro, Small, and Medium Enterprises (MSMEs) are not a niche — they are the economy. In ASEAN alone, MSMEs make up 97% of all businesses, contribute 40–60% of regional GDP, and employ nearly 70% of the workforce. In India, over 6.3 crore registered MSMEs contribute roughly a third of GDP, account for nearly 46% of the country's exports, and generate more than 22 crore jobs. Indonesia tells a similar story: 65 million MSMEs, representing 97% of total employment and 60% of GDP.

These are not small numbers. These are the structural pillars of the most dynamic region on earth.

And yet, these pillars are being systematically starved of capital.

India's MSME credit penetration sits at just 14% — compared to 37% in China and 50% in the United States. The Asia-Pacific region as a whole faces a staggering **$2.5 trillion credit access gap**, more than half of the entire global shortfall in small business financing. In India specifically, the addressable credit gap is estimated at ₹30 lakh crore by SIDBI-Crisil. RBI's own expert committee put the figure at ₹20–25 lakh crore — and that was before the post-pandemic acceleration in MSME registrations.

The consequences are severe. Without reliable access to formal credit, MSMEs turn to informal lenders charging anywhere from 2% to 20% per month. Working capital dries up. Growth stalls. Businesses that could thrive fail within their first five years — 60% of SMEs in Malaysia, 30% in Singapore. India's ambition to become a $5 trillion economy rests substantially on unlocking MSME potential. The credit gap is one of the most critical structural constraints standing in the way.

So why does this gap persist despite decades of policy initiatives, government schemes, and the explosion of fintech?

The answer is deceptively simple: **lending to MSMEs is expensive, slow, and hard.**

---

The "Thin File" Problem and the Cost of Human Underwriting

Traditional banks are not villains in this story. They are rational actors responding to real risk signals. Most MSMEs lack comprehensive financial records. They often have no formal credit history. Their fixed assets are modest — the average fixed assets of unincorporated non-agricultural enterprises in India were just ₹3.18 lakh in 2022-23. Collateral requirements designed for large enterprises simply don't map to MSME realities.

This is the "thin file" problem. The borrower may be creditworthy; the data to prove it simply doesn't exist in the formats that banks traditionally accept.

The second problem is operational cost. Underwriting a small business loan requires nearly as much human effort as underwriting a large corporate loan — financial analysis, site visits, relationship building, document verification, credit memos. But the loan size is a fraction, making the economics punishing. Serving an MSME borrower costs almost as much as serving a large corporate, but the revenue is a small fraction. The math simply doesn't work at scale without a fundamentally different operating model.

This is exactly the gap that Direct Selling Agents (DSAs) like RupeeBoss have stepped into — and where the next generation of AI-enabled platforms is now creating a genuine competitive edge.

---

What RupeeBoss Does — and Why It Works

RupeeBoss is a loan distributor — a DSA operating across India, facilitating MSME loans across both secured and unsecured segments. Their people are predominantly ex-bankers. They understand credit risk intuitively. They know how to read a balance sheet, assess a borrower's character, and navigate a bank's credit committee. They have deep relationships — with MSMEs on the demand side and with banks and NBFCs on the supply side.

This is an intensely relationship-driven, judgment-driven business. And for good reason: the credit assessment of an MSME is not a purely algorithmic problem. It involves soft information — business quality, promoter credibility, local market dynamics — that experienced bankers pick up in a conversation and a site visit.

But here's the tension: the business is also fundamentally limited by human bandwidth. Each credit file requires sourcing, analysis, product matching, packaging, and submission. Every step is sequential. Every step relies on an individual's time and knowledge. Scaling means adding people — a costly, slow, and operationally complex proposition.

PN Shetty has built an incredible operating model, team, SOPs that is scalable and that expert model is getting augmented by AI. This comes from years of experience the RupeeBoss team has in this domain. This experience is getting fed back into the AI models. 

This is why what I saw in RupeeBoss's AI platform demo matters so much.

---

AI as an Operating System for Expert-Led Lending

The platform the RupeeBoss AI team is building isn't trying to replace bankers with algorithms. It's doing something far more sophisticated: **building an intelligent operating layer that amplifies what expert bankers can do.**

The workflow is elegant. When a lead is sourced — an MSME expressing interest in a loan — it enters the system. On the supply side, the platform has aggregated the lending criteria, product parameters, and risk appetites of multiple banks and NBFCs. The AI engine then performs an intelligent **demand-supply matching**: given this borrower profile, what products fit, what lenders are likely to approve, and what documentation will be required?

This single step — which today requires an experienced banker to hold all of this in their head or sift through separate portals and product sheets — is collapsed into a near-instant, structured recommendation. The loan officer now knows exactly where to focus. Internal pre-underwriting time drops dramatically. The file that goes to the bank is better prepared, more targeted, and more likely to be approved on the first submission.

This is the power of AI when applied to operational complexity: it doesn't eliminate human judgment — it gives human judgment a dramatically better starting point.

---

Why "Expert in the Loop" Is Not a Compromise — It's the Competitive Advantage

There is a tempting but flawed vision of MSME lending that goes fully automated: a borrower uploads documents, an algorithm spits out an approval, money moves. This may work for certain standardized consumer loan products. It does not work for the nuanced, relationship-intensive reality of MSME credit.

MSME lending involves businesses that are often informal, have irregular cash flows, operate in local contexts that algorithms don't understand, and are run by promoters whose intent and capability is as important as their balance sheet. A fully automated system misses this. And when algorithms miss, the consequences — bad loans, exclusion of creditworthy borrowers, regulatory risk — are significant.

The most sophisticated thinking in AI-enabled lending now recognizes this. The emerging consensus is a "Human-in-the-Loop" (HITL) model — where AI handles high-volume, structured tasks (document parsing, data aggregation, product matching, compliance checks, risk flagging) and human experts handle the judgment-intensive, contextual, and relationship-driven dimensions of the decision.

Research confirms the value: AI implementation in lending drives productivity gains of 20% to 60%. Lenders using AI-based systems have reduced per-loan origination costs by up to 14% and cut processing defect rates by 40%, with loan production cycles shortened meaningfully. But the critical insight is that the best outcomes don't come from full automation — they come from the thoughtful combination of AI efficiency and human expertise.

RupeeBoss has an asset that most AI-first fintech startups lack: **a large, experienced team of ex-banker professionals who already understand the credit ecosystem deeply.** They have the human expertise baked in. The AI platform amplifies it. This is not retrofitted — it is purpose-built.

This is a pattern I know from firsthand experience at Credibl. In sustainability reporting and supply chain compliance — the space we operate in — the data is complex, the regulatory frameworks are evolving rapidly, and the judgment calls are genuinely hard. Which emission factors apply to a particular activity? How should a company interpret a supplier's partial disclosure under CSRD? Is a scope 3 boundary defensible to an auditor? These are not questions you can answer with an algorithm alone. What AI does brilliantly at Credibl is compress the operational overhead: ingesting and normalizing data from dozens of sources, flagging inconsistencies, mapping disclosures to the right regulatory frameworks. What our sustainability experts do is bring the contextual intelligence — the knowledge of sector-specific nuances, the regulatory interpretation, the auditor's lens — that makes the output credible to boards, regulators, and assurance providers. The expert in the loop is not a concession to the limitations of AI. It is what makes the output trustworthy at the point where it matters most. I believe MSME credit underwriting is structurally identical.

The AI handles what AI is good at: aggregating lender criteria, matching borrower profiles to products, structuring documentation, flagging inconsistencies, accelerating internal pre-screening. The banker handles what bankers are good at: the relationship, the soft assessment, the local knowledge, the final judgment call.

The result is a loan officer who can handle significantly more files — with higher quality, better targeting, and faster turnaround — than was possible in a purely manual model. The economics of serving MSMEs, long structurally challenged, begin to shift.

---

The Bigger Picture: What This Means for Asia

Step back from the product for a moment and consider what's at stake.

India is home to over 6 crore MSMEs. Only 20% of micro and small enterprises accessed credit through scheduled banks as of 2024, up from 14% in 2020 — progress, but profoundly insufficient. NITI Aayog has estimated that in FY21, only 19% of MSME credit demand was formally met.

Closing this gap is not just a business opportunity — it is an economic imperative. Faster credit with better risk management means more MSMEs survive their first five years. It means more entrepreneurs in Tier 2 and Tier 3 cities get the working capital they need to grow. It means employment. It means the broad-based, inclusive growth that India's demographic dividend demands.

The same story plays out across Indonesia, Vietnam, the Philippines, and the rest of the Asia-Pacific region — where MSMEs are equally central to economic structure and equally underserved by formal finance. The ADB estimates the ASEAN MSME financing gap at over $300 billion annually.

AI-enabled platforms built on expert human networks are arguably the most credible path to closing this gap at scale — more credible than pure algorithmic lending (which struggles with thin files and novel situations), and more scalable than pure relationship-based models constrained by human bandwidth.

---

Why I'm Excited

I've spent my career building platforms that change the ratio of what one person can do. At Qwiklabs, we built cloud learning infrastructure that scaled expertise across millions of learners globally. At Credibl, we're building data infrastructure that scales sustainability reporting. What I see RupeeBoss building follows the same logic: **take a team of deeply skilled professionals and give them a system that multiplies their capacity without diluting their judgment.**

The MSME credit problem in India and across Asia is real, large, and stubborn. But it is not unsolvable. The combination of experienced bankers who understand local credit realities, an AI platform that dramatically compresses operational overhead, and the institutional relationships built over years of operating at ground level — this is a powerful combination.

I went into that demo as a board member doing diligence. I came out genuinely excited about what a platform like this could mean for millions of small businesses across India that deserve faster, fairer access to capital.

This is early. But the direction is right. And in my experience, when the direction is right and the team is capable, that's usually enough.

---

Jitesh Shetty is the Founder & CEO of Credibl ESG, a climate risk and ESG data management SaaS platform. He previously co-founded Qwiklabs, acquired by Google. He serves as a board member of several early-stage technology companies across India and the US.