Incorrect password
Executive summary
5 pages · 8 minute read
An AI pipeline that eliminates the cost barrier preventing institutional capital from reaching nano-businesses. Not a fund. Not a lender. The operating system that makes last-mile capital deployment economically viable at institutional scale.
300+ million people run informal nano-businesses with zero access to equity capital. Not because investors don't want to invest — because it costs $5,000 to evaluate a $900 opportunity. The economics have never worked. Until now.
Seen Capital's seven-agent AI pipeline processes investment candidates through WhatsApp, in any language, autonomously. It conducts intake, researches the geography, scores the opportunity, interviews the candidate, makes the decision, structures the agreement, and sets up tracking — all without human intervention for standard applications.
The result: a $900 average investment with unit economics that work at institutional scale.
This is not microfinance. There is no debt. No interest. No fixed repayment schedule. Seen Capital deploys equity, provides formalisation support (business registration, tax ID, bank account), purchases a productive asset under title retention, and takes a 10% revenue share until the working capital is recovered. The entrepreneur keeps the asset, the business registration, and the growth.
The fund structure eliminates the J-curve. Revenue begins Week 1. Capital recycles without new fundraising. And a chain mechanism drives geometric portfolio growth from a single seed cohort.
Every woman who reaches break-even nominates three peers for investment. Those nominations are warm introductions — pre-qualified by someone with skin in the game. The marginal cost of sourcing new candidates approaches zero after the first generation.
The fund attracts blended capital by structuring risk in layers. Each tranche de-risks the one above it.
| Tranche | Investor | Target Return | Role |
|---|---|---|---|
| First-loss | DFIs, Foundations | 0–2% | Catalytic — absorbs first defaults |
| Mezzanine | Impact LPs | 4–7% | Protected by first-loss cushion |
| Senior | Institutional | 10–15% | Double-cushioned — venture returns, bond risk |
A competitor can copy the AI agents. They can build a WhatsApp-native pipeline. They can even replicate the revenue share structure. What they cannot replicate is the trained model, the NGO network, or the data.
Seven-agent system trained on real investment outcomes. The prompts encode institutional knowledge — rubrics, consistency checks, counterintuitive patterns — that took years to develop.
Default rates, recovery curves, revenue share behaviour, chain dynamics across 15+ geographies. This dataset does not exist anywhere else in the world.
Every weekly payment event trains the scoring model. 8.7 million events per cycle. Default rate falls from 8% to 3%. The model improves itself.
20+ partnerships across 15+ geographies, built over 14 years. Earned trust that cannot be purchased. These relationships are the distribution layer.
$500K/client/year for DFIs and development banks. Five clients by Year 5 = $2.5M recurring. Pure technology revenue.
Validated conversion rates, generation intervals, nomination quality metrics. The proof that geometric growth works — across cohorts and geographies.
The scoring model starts with uniform assumptions — all features weighted equally. As investments accumulate and payment data flows back, the model discovers what actually predicts success.
What the model has learned:
Business maturity is the strongest predictor. Mobile money history is critical. Community endorsement matters. Number of dependents and distance to market are noise — they add zero predictive power. And prior microfinance loan history is a negative signal: women conditioned by fixed-repayment debt struggle with income-linked payments.
That last insight is discoverable only through data, not theory. A human analyst would include prior loan history as a positive signal. The model discovers the truth.
25 jurisdictions assessed. 8 ready to deploy now. Deployment sequence follows regulatory simplicity, not market size — generate proof-of-concept data in the cleanest markets first.
| Phase | Timeline | Geographies | Purpose |
|---|---|---|---|
| Phase 1 | Months 1–3 | Rwanda | Proof of concept. Cleanest regulatory environment. First 50 investments. |
| Phase 2 | Months 3–6 | Kenya, Tanzania, Uganda | East Africa expansion. M-Pesa penetration. 150+ portfolio companies. |
| Phase 3 | Months 4–8 | Namibia, Kyrgyzstan, Ghana, Senegal | Six-geography proof for Series A. Geographic diversification. |
| Phase 4 | Months 6–12 | Nepal, Ethiopia, Morocco, India, Côte d'Ivoire | Local entity formation. Year 1 expansion markets. |
| Phase 5 | Year 2+ | Indonesia, Bangladesh, Uzbekistan | Complex structures. Named as Fund II story. |
Navigating FDI minimums, mobile money regulations, cooperative law, and capital repatriation rules across 25 jurisdictions creates a barrier that no competitor can shortcut. Seen Capital's 14-year NGO relationships provide the trust layer that regulatory approval alone cannot grant.
Nigeria is blocked. Despite being the largest African market, the naira's 70% collapse against USD (2023–24) makes capital deployment economically unviable. It moves to Phase 5 pending currency stabilisation. This kind of jurisdiction-specific analysis — prioritising data quality over market size — is what institutional investors need to see.
The management company is acquired as a technology platform at 10–20× revenue, not as a fund manager at 2–3% of AUM. The fund vehicles continue under new management. This is a clean, proven exit structure.
The difference is the six proprietary assets. A fund manager is valued on AUM. A technology company is valued on revenue, growth rate, and the irreplaceability of its data. Seen Capital is the latter.
Mastercard, Visa, M-Pesa. Seen Capital turns their payment rails into capital rails. Distribution play — access to 100,000+ portfolio companies as payment users.
BlackRock Impact, Nuveen, LeapFrog. Seen Capital enables deployment at ticket sizes they can't reach. Infrastructure play — plug into their AUM with proven pipeline.
Stripe, Block, Flutterwave. The compound risk model trained on 100,000+ nano-businesses is the largest dataset of its kind. Data play — unique, defensible, monetisable.
Stripe proved its infrastructure by processing its own payments. Seen Capital proves its infrastructure by managing its own funds. In both cases, the fund is the proof. The pipeline is the product. The data is the moat.
The full investor materials include detailed financial projections, jurisdiction-by-jurisdiction analysis, the complete self-optimising model simulation, and the live pipeline demo.