CredPulse
Real-time Agentic credit intelligence for MSMEs
Created on 28th August 2025
•
CredPulse
Real-time Agentic credit intelligence for MSMEs
Description of your solution
CredPulse — Real-time Agentic Credit Intelligence for MSMEs
CredPulse is an agentic AI platform that turns live, consented financial and supply-chain signals into instant, explainable working-capital decisions for MSME suppliers and their corporate buyers. It continuously detects risk and opportunity across invoices, payments, and deliveries; plans the optimal financing route; acts with explicit user consent (e.g., auto-verify invoice, price risk, trigger disbursement simulation); and maintains a tamper-proof audit trail — all in under 60 seconds.
Problem We’re Solving
MSMEs face persistent cash-flow gaps because buyer payments arrive 30–90 days after delivery, while underwriting remains collateral-heavy and slow. Data is fragmented across bank statements, UPI flows, GST/e-invoices, e-way bills, delivery proofs, and buyer ERPs — making real-time, low-cost working capital hard to unlock. Existing factoring requires manual verification and static scoring, keeping costs high and approval times long.
Our Agentic AI Solution
CredPulse is a consent-first, autonomous credit OS built for supply chains:
- Ingests multi-party, verifiable signals (bank/UPI via regulated rails, GST/e-invoice IDs, e-POD events, buyer payment histories).
- Builds a live buyer–supplier “risk graph” that updates whenever a delivery, invoice, or payment lands.
- Generates explainable, real-time credit scores and prices risk per buyer and tenor.
- Executes financing workflows (with pre-approved consent rules): invoice verification, instant scoring, rate selection, simulated disbursement, reconciliation on buyer payment.
- Preserves a full reasoning and data-lineage audit for every decision (who consented, what data, why priced this way).
This moves beyond dashboards to true agency: detect → plan → propose → act (with consent) → audit.
Where the Agent Acts (Detect → Plan → Act → Audit)
-
Supply Chain Intelligence Agent
- Detects: invoice creation, delivery confirmation, buyer delay patterns.
- Plans: updates supplier risk graph, flags financeable invoices, selects tenor/rate bands.
- Acts: promotes eligible invoices to “ready for financing,” queues consent prompt.
- Audit: logs evidence links (invoice ID, e-POD reference) and reasoning steps.
-
Invoice Intelligence & Factoring Agent
- Detects: authenticity and match (GST/e-invoice ↔ delivery ↔ PO).
- Plans: optimizes factoring rate by buyer risk and historical SLA.
- Acts: executes simulated factoring with disbursement and collections schedule.
- Audit: appends verification proof and dynamic pricing rationale.
-
Alternative Credit Scoring Agent
- Detects: cash-in/out velocity, GST cadence, UPI working-capital cycles, utility signals.
- Plans: updates an on-chain/off-chain score (0–1000) with reason codes.
- Acts: adjusts limits; pauses exposure if early-warning signals spike.
- Audit: versioned models, input features, and human-appeal workflow.
-
Cash-Flow Orchestration Agent
- Detects: upcoming cash gaps (14–60 days), seasonality, buyer behavior shifts.
- Plans: chooses best route (early-payment, factoring, limit top-up).
- Acts: triggers simulated disbursement on consent; auto-reconciles on payment.
- Audit: end-to-end ledger of proposals, approvals, actions, and outcomes.
Why It Fits the Fintech Track (Bring Your Own Problem)
- Fintech Core: Tackles MSME credit access and working-capital friction with consented data, real-time scoring, and automated financing.
- Agentic by Design: Autonomous agents don’t just analyze; they execute compliant, user-approved financial actions with full transparency.
- Real Impact: Faster, cheaper, explainable working capital stabilizes supplier cash flows and strengthens entire supply chains.
Key Features (MVP Scope for Hackathon)
- Consented data ingestion (mock/sandbox): bank/UPI CSV import, GST/e-invoice lookup simulation, delivery/e-POD webhook simulation.
- 60-second decisioning: instant score with explainable “Why this decision?” and data lineage.
- Autonomous workflows (simulated): invoice verify → price → consent modal → “Simulated Disbursement” → auto-reconcile on payment event.
- Risk graph & watchlists: buyer-wise aging, disputes, SLA drift; alerts that throttle/boost limits.
- Full trust controls: aggressiveness slider (Suggest-only / Auto-with-rules), one-tap pause/revoke, immutable audit.
Compliance & Trust (Principles)
- Consent-first data access; no credential storage; field-level encryption.
- Explainable decisions with human appeal; one-tap pause/revoke automation.
- Lending via regulated partner; all finance in demo is simulated.
Tracks Applied (1)
