Nucleus Software

FinnOne Neo is a lending platform first. Lenders build loan products from rules, policies, masters, forms and workflows, and an embedded rule engine handles scoring policies, deduplication and exceptions, so origination, servicing and collections run on configuration. AI sits at specific points. Underwriters get AI confidence scores and recommendations tied to internal credit policy, and origination checks applications for fraud with machine learning pattern recognition in real time.

Release 9.0 adds AI credit decisioning, comparative credit summaries and pre delinquency prediction, and collections ranks accounts by likelihood of repayment and suggests the best time and channel for each borrower. Nucleus also sells AI services separately, building models, retrieval applications and agents for banks. Take the models out and the platform still originates, books and collects loans.

Credit decisions run inside workflows the lender configures, which Nucleus pitches as automation with human oversight. Origination supports multilevel approval hierarchies, and release 9.0 adds logic based committee decisioning built on comparative credit summaries. The AI confidence score reaches the underwriter as a recommendation rather than a decision. Straight through processing is also on offer, with automated underwriting, fraud identification and rule based exception handling.

In collections, rules allocate cases to staff, agencies and field agents, approvals can be given by email, case visibility follows authority levels, and contact hour limits and consent rules govern outreach. Nucleus does not say where an automated decision stops and passes to a person, or what review a declined application receives.

Two mechanisms support validation. Underwriters see an AI confidence score beside each recommendation, and the collections module runs champion and challenger tests so a lender can compare strategies on its own portfolio. Scoring policies can be set at product and scheme level, and the rule engine supports verification and auditing of rules.

No accuracy figure accompanies the credit decisioning or pre delinquency models, and there is no validation report, model documentation or mapping to a supervisory standard. Gartner named Nucleus a representative vendor in a November 2025 report on applying predictive AI and synthetic data to lending risk assessment.

Bank of Sydney, one of the first FinnOne Neo implementations, cut loan application processing time by about 97 percent and broker loan agreement turnaround by about 95 percent, according to an AWS case study, and desktop property valuations there fell from days to about two hours.

EastWest Bank in the Philippines chose FinnOne origination in 2011, and a Nucleus case study covering 2011 to 2014 has it running across 645 branches with more than 700,000 loans processed across more than 50 products. Relationships run long. Cholamandalam has used Nucleus since 2006, and Axis Bank reports nearly two decades.

Nucleus counts more than 200 bank and financial institution customers, with more than 100 on the origination system, more than 95 on collections and more than 80 on loan management, and exports made up 37.55 percent of turnover in its 2024 to 2025 sustainability report.

Nucleus's privacy notice, updated 15 April 2026, says the company may use AI and machine learning tools, both third party and built in house, with appropriate safeguards where they process personal data. It does not say whether lenders' data trains models, whether learning crosses from one client to another, or what an outside AI tool keeps from a request.

Release 9.0 masks documents and personal data, which limits what a model or a user sees, and lenders that run FinnOne Neo on premises keep their lending data inside their own environment. There is no AI data policy specific to the lending platform.

The privacy notice of Nucleus Software Exports, updated 15 April 2026, covers customers and end users of its products as well as website visitors, but only where Nucleus decides the purposes and means of processing. That leaves out the borrower data lenders hold in FinnOne Neo, where Nucleus works on the lender's instructions.

The notice keeps data as long as its purposes and legal obligations require, allows transfers abroad under standard contractual clauses and intragroup agreements, and lists service provider categories such as cloud hosting without naming the providers or any data protection law. FinnOne Neo itself comes with encryption, role based access and a stated commitment to GDPR compliance.

Nucleus's Business Responsibility and Sustainability Report, filed with the Indian stock exchanges in July 2025, states ISO 27001:2022 certification and a SOC 2 Type II report. Neither document is on its website, and the scope of each is not given. The same filing reports no data breaches and no data privacy complaints in either of the two years it covers.

A responsible disclosure program covers the website, public applications and APIs, mobile apps and FinnOne Neo and FinnAxia deployments, with a safe harbor for researchers and no bounty, and the product encrypts data and restricts access by role.

Nucleus supplies software to lenders, is not a lender itself, and does not set out its own regulatory position. Its compliance features serve the lender's obligations. Collections enforces contact hour limits and borrower consent rules and keeps communication audit trails, origination carries configurable policy controls, and release 9.0 adds data compliance and regulatory controls alongside document masking.

Islamic lending has its own product line, and the Vietnamese subsidiary lets banks there contract under local procurement and compliance rules. No regulator, statute or supervisory guideline is named, across a customer base spanning India, the Philippines, Vietnam, Indonesia, Australia and Nigeria.

FinnOne Neo sells AI credit decisioning, analytics based scoring models and AI recommendations to underwriters, and collections strategies can be segmented by risk level, repayment behavior or demographics. Nucleus offers no fairness testing, disparate impact analysis or guidance on which borrower attributes its models use, and says nothing about the reasons given to declined applicants. The AI services pages commit to ethical standards and privacy guidelines without a process behind them. Its sustainability filing records board training on new technologies including AI, and no responsible AI policy.

FinnOne Neo keeps audit trails for its rules and for collections communications, so a lender can trace how a case was handled and which policy applied. Where a lender switches on the self service portal, borrowers can raise disputes and negotiate settlements through it.

Nucleus's public terms cover only its website and disclaim all warranties for it, and the platform's license terms are private, so who bears the loss when an AI credit decision or a delinquency prediction is wrong is not known.

Amazon Web Services runs about 95 percent of Nucleus's cloud deployments, on Amazon EC2, Amazon S3 and Amazon RDS for Oracle and PostgreSQL, with customer environments moving to Amazon EKS. FinnOne Neo is listed on AWS Marketplace and Microsoft AppSource. The AI inside the lending platform comes with no named model provider, model family or subprocessor list, and the privacy notice says only that third party and in house AI tools are both used. The AI services practice builds agents for clients with LangChain, AutoGen and CrewAI, which describes consulting work rather than the product.

FinnOne Neo exposes more than 580 APIs across the platform, over 230 of them in origination and more than 85 in collections. It connects to core banking systems through APIs or middleware, to credit bureaus and outside scoring engines configured by region, and to payment gateways, electronic signature platforms, video KYC and eKYC. Collections works with outside speech analytics and chatbot tools.

At Bank of Sydney the platform links lending and core banking to broker platforms, desktop valuation services and solicitor systems. Nucleus names no specific core banking system or bureau, and the API reference is not public.

FinnOne Neo runs in the cloud or on premises, on any database and application server, and Nucleus leaves the choice to the lender's regulatory needs. Cloud deployments run mostly on AWS, where the platform is containerized and moving to Amazon EKS, with isolated test environments created on demand. At Bank of Sydney, the AWS case study describes support for local data residency requirements. The mobile collections app works offline, as do some other features depending on configuration. Implementation runs 6 to 12 weeks for origination and 12 to 16 weeks for collections.

FinnOne Neo and its modules are priced by quote. A buyer who registers on the AWS Marketplace page receives pricing with the module list, a deployment plan and support options, and the Microsoft AppSource listing for collections shows no plan. Implementation timelines are set out up front, 6 to 12 weeks for origination and 12 to 16 weeks for collections, and support after go live includes a helpdesk, service levels and optional managed services. Whether licensing runs per user, per loan or per module is not public.

FinnOne Neo is built for banks, non bank finance companies and housing finance companies, with dedicated products for retail, SME, corporate, automotive, gold and Islamic lending, plus finance against securities, co lending and Omni Loans in release 9.0.

Named customers run from Indian lenders such as Cholamandalam, ICICI Home Finance, HDFC Bank, Axis Bank, DCB Bank, Mahindra Home Finance and Poonawalla Fincorp to EastWest Bank in the Philippines, MB Bank and Vietcombank in Vietnam, Bank of Queensland and Bank of Sydney in Australia, and Stanbic IBTC in Nigeria. Nucleus says it has served Indonesian banks for nearly two decades. FinnAxia, its transaction banking suite, sells cash management and supply chain finance to the same banks.