BEST WISHES FROM
B Swaminathan
The integration of Artificial Intelligence (AI) in India’s credit intelligence landscape is revolutionizing lending practices and fostering financial inclusion. Traditionally, credit assessments in India relied on limited data, often excluding vast segments of the population with “thin” or no credit files. AI-powered systems overcome this by analyzing diverse, alternative data sources like digital footprints, mobile usage, utility payments, and social media activity.
Rubix Data Sciences provides B2B risk and credit intelligence that helps SMEs and MSMEs enable safer trade and lending decisions. With a deep understanding of compliance, credit ecosystems, and India-specific risk indicators, Rubix offers a grounded view on how software can help small businesses grow while staying financially secure. Tushar Bhaskar, Chief Business Officer at Rubix Data Sciences speaks on that.
With the increasing reliance on AI and machine learning for crucial business decisions, what are the most significant ethical considerations that organizations must prioritize when developing and deploying these advanced analytical models?
At Rubix Data Sciences, we believe that ethical AI is about principles as well as about operationalising trust at scale. A recent IDC–Credo AI survey (2023) found that “AI-first, ethics-forward” companies reported 22%–29% higher growth in revenue, customer satisfaction, and product innovation. Yet, only 39% of firms expressed high confidence in their ability to use AI ethically. This trust gap highlights the need for concrete governance. Leading frameworks like NIST’s AI Risk Management Framework and Singapore’s FEAT principles (Fairness, Ethics, Accountability, Transparency) offer actionable scaffolding. At Rubix, we embed ethical foresight by building auditability, explainability, and fairness checks into our credit and compliance analytics, from counterparty onboarding to trade credit risk assessments and monitoring.
Data privacy is a growing concern globally. How do you see the advancements in privacy-preserving AI techniques, such as federated learning or homomorphic encryption, impacting the future of data analytics, especially in highly regulated sectors like financial services?
Privacy-preserving AI technologies like Federated Learning (FL) and Homomorphic Encryption (HE) are reshaping the future of data analytics in regulated sectors. FL enables institutions to collaboratively train models without moving sensitive data, particularly valuable in scenarios like cross-institutional fraud detection or credit risk modelling across financial networks. FL allows firms to “leverage new data resources without requiring data sharing,” aligning with privacy mandates such as India’s DPDP Act and the EU’s GDPR. Meanwhile, HE allows analytics on encrypted data, ensuring sensitive information remains confidential even during computation. Though still maturing, both technologies are vital tools in Rubix’s roadmap toward privacy-by-design analytics, especially as we support clients navigating multi-jurisdictional compliance.
Algorithmic bias is a critical challenge in AI development. In the context of risk assessment and decision-making, what practical strategies can companies implement to identify, mitigate, and continuously monitor for bias in their AI systems?
The talk about bias in AI systems is not just a theoretical concern—it is a well-documented reality. According to a 2024 Urban Institute analysis of Home Mortgage Disclosure Act (HMDA) data in the US, Black and Brown borrowers were more than twice as likely to be denied a mortgage compared to white applicants. A 2022 UC Berkeley study on fintech lending further found that, on average, African American and Latinx borrowers were charged nearly 5 basis points more in interest than credit-equivalent white borrowers, amounting to an estimated USD 450 million in extra interest payments annually. These disparities often arise from seemingly neutral proxy variables such as zip codes, device types, or email domains—factors that inadvertently reflect socioeconomic divides. While such large-scale studies are more prominent in Western markets, concerns about algorithmic bias are equally valid in India, where deep-rooted inequalities across caste, gender, income, and regional access can subtly influence AI outcomes if not actively mitigated. At Rubix, we mitigate such risks by conducting pre-deployment bias audits, using disparate impact metrics, and implementing ongoing fairness monitoring. Embedding synthetic data to balance datasets and adopting fairness-aware ML techniques are part of our strategy to ensure that AI does not amplify systemic inequities.
Beyond regulatory compliance, what does ‘responsible AI’ truly mean for businesses operating with vast amounts of sensitive data? What frameworks or principles should guide organizations in building trustworthy and transparent AI solutions?
Responsible AI is not confined to ticking boxes in regulatory checklists. It means aligning AI systems with societal, ethical, and human values. As the World Economic Forum notes, the number of trained AI governance professionals must scale with deployment. The RBI’s FREE-AI committee (2024) and the DPDP Act are part of India’s growing regulatory architecture, but true responsibility requires internal rigor. Rubix’s commitment to responsible AI means maintaining human-in-the-loop oversight, documenting every decision node, and ensuring transparency in credit and compliance models.
The concept of ‘explainable AI’ (XAI) is gaining traction. Why is it becoming increasingly important for businesses to not only make accurate predictions but also to be able to explain the reasoning behind their AI-driven decisions, particularly when those decisions impact individuals or other businesses?
Explainability is fast becoming a cornerstone of trust, particularly in high-stakes financial decisions. The UK FCA’s 2024 survey found that 81% of financial institutions using AI also deploy explainability tools to validate decision-making. Global regulations like the EU AI Act (2024) and India’s Model Risk Management guidelines by the RBI now explicitly require interpretability for high-risk models.
A prime example is our Rubix Risk Score, a dynamic, algorithmically derived indicator that quantifies the financial and compliance risk of a counterparty across multiple dimensions. The score incorporates near-real-time data from statutory filings, litigation records, financial ratios, payment behaviour, compliance signals, and sectoral risk factors. Each component of the score can be traced and explained, making the model both auditable and transparent. The score dynamically adjusts as new data becomes available, reflecting a counterparty’s evolving risk profile. This level of granularity allows our clients to not only rely on the score but also understand the “why” behind it. When counterparties are denied credit or flagged for compliance, we believe they deserve to know why.
Looking ahead, what are the biggest challenges and opportunities that the broader data science community faces in fostering a culture of ethical AI development and deployment across industries?
The biggest challenge facing the AI community is the gap between innovation velocity and governance maturity. While model development cycles have accelerated, ethical risk reviews and fairness audits often lag behind. As highlighted by the World Economic Forum, the shortage of trained AI ethics professionals is a bottleneck. However, the opportunity is profound: to automate with accountability and democratise access to finance, credit, and insights. Rubix sees a future where ethical AI drives inclusive growth. That’s why we invest in explainability, bias mitigation, and governance frameworks to build data ecosystems that our clients and regulators can trust.