Groot OS layer
The AI operating layer behind Brolly's private credit platform.
Groot is the AI operating layer behind Brolly, running risk, servicing, capital and compliance workflows across the community lending platform.
Wholesale, sophisticated and professional investors only. Confidential information is released after eligibility confirmation.
Almost one year of live marketplace data. Operating at credit-platform maturity Level 2 today.
What Groot is
Live credit operations, becoming an AI-native operating layer.
Live private credit fintech
~2,500 funded loans, 10,000+ waitlist, 1.43% contract default rate, 0% lender capital losses and almost one year of live marketplace data. Operating at credit-platform maturity Level 2.
AI-native credit operating layer
16 agents in production or shadow mode today, ~28 by Year 3, 45+ at peak state. Targeting Level 5 maturity over 5+ years.
Six structural unlocks
Predictive credit, dynamic pricing, capital velocity, product velocity, infrastructure licensing, multi-market expansion. Converts Brolly from a single-product lender into financial infrastructure.
Maturity model
Five levels of credit platform maturity.
Brolly is architected toward L5. Active focus today is L2, moving toward L3 over the next 12 months.
Interactive flywheel
Every loan outcome improves the next.
Tap each node to see the measurable outcome that improves over time.
Default / bad rate by band
Tighter risk segmentation as repayment outcomes feed model calibration.
Five engines
Bounded scope, deterministic authority.
Groot is organised into five engines spanning the full borrower and lender lifecycle. Each engine is bounded. Each agent has a single responsibility.
Engine
4 liveRisk
Understand borrower quality
Reads bank-data patterns, flags missing information, checks stale consent and turns raw signals into explainable risk bands and product caps. Reproducible and audit-tagged.
- Data Quality
- Affordability
- Reason Code
- Manual Review Router
Engine
6 liveServicing
Support repayment at the right time
Detects salary cadence and liquidity windows. Coordinates nudges, partial payments, splits and human handoff. Reduces failed debits without harming borrowers.
- Payday Detection
- Liquidity Window
- Collection Strategy
- Hardship Guardrail
- Comms / Nudge
- Payment Orchestrator
Engine
Year 2-3Behaviour
Learn trust and step-up readiness
Builds a longitudinal profile from repayment history, bank health and engagement. Determines when a borrower is safe to step up to a larger product.
- Behaviour Profile
- Step-up Recommendation
- Fraud Pattern
Engine
2 liveCapital
Match lender funds to eligible loans
Matches eligible borrower demand to lender capital using risk band, term and availability. Enforces concentration and liquidity guardrails in real time.
- Capital Match
- Concentration Guard
Engine
4 liveCompliance
Create evidence, route exceptions
Packages AML/KYC evidence, monitors DDO and target-market rules, builds immutable audit decision packets, routes borderline cases to humans with context.
- KYC / AML Packager
- DDO / TMD Monitor
- Audit Evidence Builder
- Exception Queue Router
Agent growth
Scaling agents, not headcount.
16 agents in production or shadow mode today. Scaling agents, not headcount.
Infrastructure stack
Institutional rails. Mapped to each Groot engine.
Groot does not run on prototype infrastructure. Each partner has a defined function inside a specific Groot zone or engine.
Basiq
Live · CDR-accreditedOpen Banking data feed. Live income, expense and liquidity signals into the Affordability and Data Quality agents.
Monoova
Live · AFSL-authorisedNPP and PayTo orchestration powering the Payment Orchestrator agent. 24/7 instant transfers and programmable repayments.
FrankieOne
LiveUnified KYC, KYB, biometric and AML/CTF screening orchestration for the KYC/AML Packager agent.
InDebted
LiveAI-driven arrears management and digital-first borrower engagement for the Collection Strategy agent.
AWS
LiveEnterprise-grade compute, storage, encryption and monitoring for policy tools, model versioning and immutable audit events.
ConnectID
In discussionsBank-backed digital identity exchange that may extend KYC verification through the major-bank network.
Mastercard
In discussionsCard-rail and distribution infrastructure on the medium-term roadmap for adjacent borrower payment surfaces.
Institutional rails. Mapped to each Groot engine.
Guardrails
Where humans stay in control, by design.
Investor and regulator diligence focuses on what an AI system is prevented from doing.
Canonical decisions stay deterministic
Score and cap decisions are produced by versioned policy and model tools, not LLM agents. Same inputs always produce same outputs.
No black-box credit decisions
Every borrower-impacting decision has a reason code, a trace and an override path.
Hardship is a hard guard
Vulnerability and hardship signals trigger hard policy guards that override agent recommendations.
Every action writes an audit event
Onboarding, scoring, capping, servicing, collection, capital allocation and reporting write immutable audit events.
Human override everywhere
No automated workflow is sealed off from human intervention. Overrides write audit events with operator identity, timestamp and reason.
The team building it
Infrastructure-grade execution.
Groot is not a research project. It is a production system being built by engineers with prior shipping history in enterprise software, infrastructure-scale ML and tier-1 financial integrations.
Nadun Perera
Architecture · AI · Platform
- 20+ years as a software engineer and platform architect.
- Built Brolly's platform from the ground up: Open Banking rails, decision engine and Groot.
- Founder of Agent Box, agentic AI infrastructure for enterprise.
Dineth Goonetilleke
Systems · ML · MLOps
- Senior systems and ML architect across security, availability and real-time credit-risk modelling.
- Designed systems architecture for Melbourne Metro rail lines.
- Deep MLOps stack: model ops, feature stores, drift monitoring and audit-grade evidence.
Malindu Sasanga
Engineering · Delivery
- Leads engineering, system architecture and secure delivery.
- Owns Open Banking, payments integrations, underwriting engine and cloud infrastructure.
- Day-to-day responsibility for Groot agent shipping cadence.
Ben Tavai (CEO) leads capital structuring and commercial strategy. Sonny Sethi (Advisor & Investor) brings 20+ years of consumer-credit and fintech research leadership. Currently Director of Global Research at Zip Co. Previously Head of Research ANZ at Uber. 1,500+ research projects across Foxtel, Tabcorp, Ipsos and Nielsen. Michael Hosking (COO) oversees servicing, customer experience and compliance operations from 15+ years in SaaS and fintech (LinkedIn, Certn, Absorb, Instructure).
Roadmap
The full sequence. Each stage gated by evidence.
No autonomous borrower-impacting decisions are made until shadow-mode performance is validated.
Stage 1 · NowPackage live data
Clean decision history, repayment outcomes and lender data. Stand up feature-snapshot infrastructure. Deploy audit-event backbone.
Stage 2 · 0-90 DaysShadow Groot
Merge v2 risk bridge. Deploy AI decision service with versioned tables. Run AI behind current workflows. Compare shadow to live daily.
Stage 3 · 3-6 MonthsAdvisory mode (L3)
Show Groot recommendations to ops, risk and lender teams with reason codes. Humans still make every decision; Groot makes them faster.
Stage 4 · 6-12 MonthsBounded automation (L3 to L4)
Automate low-risk gates, standard nudges and reporting. Activate behaviour profile and capital-allocation guardrails.
Stage 5 · 1-3 YearsMarketplace activation (L4)
Full two-sided marketplace live. Treasury, pricing, fraud and behaviour agents in production. Continuous re-pricing for top cohorts. Multi-product begins.
Stage 6 · 3-5 YearsInfrastructure layer (L4 to L5)
Self-calibration agents in advisory mode. Cohort discovery. B2B licensing to first external lenders. Multi-jurisdiction. Infrastructure revenue material.
Stage 7 · 5+ YearsPeak state (L5)
Outcome-tuned operating layer. Marginal cost materially lower. Capital velocity closer to real-time. Compliance latency materially reduced. Credit infrastructure.
Investor bridge
From current proof to peak state.
The investment case is credible when each existing asset is mapped to a specific Groot system and measurable investor outcome.
| Current Asset | Today (L2 to L3) | Peak State (L5) | Investor Outcome |
|---|---|---|---|
| Borrower waitlist + onboarding | AI origination triage | Conversational onboarding · cross-product routing | More funded loans without linear ops |
| Risk engine + Open Banking | Risk v2 with reason codes | Self-calibrating policy · continuous re-pricing | Bad rate down · false decline down · margin expansion |
| PayTo / Monoova / repayment | Smart servicing + collection | Predictive default · borrower wellbeing | Higher recovery · lower complaint risk |
| Lender matching code | Capital allocation + guardrails | Real-time market making · treasury optimisation | Capital velocity closer to real-time · idle capital reduced |
| AML/KYC integrations | KYC packager + audit evidence | Continuous audit · forensic replay · regulatory intelligence | Compliance latency materially reduced · audit cost materially reduced |
| Almost one year marketplace data | Data flywheel feeding five engines | Compounding moat across markets + products | Operating intelligence strengthens category position |
| Lending business | Principal book + marketplace | Multi-product · multi-market · infrastructure licensing | Lending funds the build · infrastructure defines the multiple |
Investor access
Stripe-like infrastructure for private credit.
Wholesale, sophisticated and professional investors under the Corporations Act 2001 (Cth). Confidential. Not for redistribution.
Contact: [email protected]
Figma-importable storyboard asset: /groot/groot-figma-storyboard.svg
Provided to wholesale, sophisticated or professional investors under the Corporations Act 2001 (Cth). Confidential. Not for redistribution.