Saudi Banking’s AI Inflection Point
Saudi Arabia’s banking sector manages more than $800 billion in total assets, serves 35 million account holders, and operates within a regulatory environment that combines internationally recognised prudential standards with Islamic finance principles specific to the Kingdom. This combination — scale, sophistication, and unique structural requirements — makes Saudi banking both a natural and a demanding application environment for artificial intelligence.
The sector’s AI transformation is being driven simultaneously from three directions: competitive pressure from fintechs forcing incumbents to modernise customer experience, regulatory guidance from the Saudi Central Bank (SAMA) creating a framework for responsible AI deployment, and the Vision 2030 mandate for a diversified, technology-enabled financial sector that can compete regionally. The result is a banking AI landscape that is more developed than most outsiders expect, grounded in concrete deployments rather than pilot projects.
Al Rajhi Bank: AI at the World’s Largest Islamic Bank
Al Rajhi Bank occupies a unique position in global banking. With total assets exceeding $200 billion and a customer base representing a substantial fraction of Saudi Arabia’s banked population, it is both the world’s largest Islamic bank by assets and one of the highest-throughput retail banking operations in the emerging markets. Its scale makes its AI deployments consequential not just for the institution but as a signal for the entire Islamic finance industry globally.
Al Rajhi’s AI deployment spans the full customer lifecycle. At the customer acquisition stage, AI-driven credit scoring has replaced — or substantially augmented — traditional scorecard approaches for consumer lending decisions. The challenge of Shariah-compliant credit scoring is distinctive: Islamic finance prohibits riba (interest), which means that the loan products being underwritten are structured as murabaha (cost-plus financing), ijara (leasing), or diminishing musharakah (declining partnership) arrangements rather than conventional interest-bearing loans. The credit risk of these structures is fundamentally similar to conventional loans, but the documentation conventions, repayment profiles, and legal remedies in case of default differ in ways that require credit scoring models specifically trained on Islamic finance performance data.
Al Rajhi’s credit AI is trained on its own portfolio performance data, incorporating bureau data from the Saudi Credit Bureau (SIMAH), which maintains credit records on approximately 20 million Saudi individuals and businesses. The AI system assesses creditworthiness across dimensions that SIMAH data captures — payment history, utilisation ratios, inquiry patterns — while integrating behavioural signals from Al Rajhi’s own transaction data, including patterns of halal-compliant spending that may indicate financial stability.
Fraud detection at Al Rajhi operates at the scale of one of the highest-transaction-volume banks in the region. Processing millions of daily transactions across ATM networks, point-of-sale terminals, and digital banking channels, the bank’s fraud detection AI uses graph neural network models to identify suspicious transaction patterns, real-time velocity checking to flag anomalous spending behaviour, and device fingerprinting to detect account takeover attempts. The system operates with millisecond latency requirements: a fraudulent transaction authorisation decision cannot wait for batch processing.
Al Rajhi’s Arabic-language customer service chatbot, deployed across its mobile banking app and web platform, represents one of the most widely used AI customer interfaces in Saudi banking. The chatbot handles routine enquiries — account balance checks, transaction history, branch locations, product information — as well as more complex interactions including complaint lodgement, account management requests, and product cross-selling. Unlike generic Arabic NLP deployments, Al Rajhi’s chatbot is fine-tuned on banking domain vocabulary, including the specific Arabic terminology for Islamic finance products, and handles Saudi dialect variations with accuracy that generic models struggle to match.
Saudi National Bank: Digital Banking AI and Data Infrastructure
Saudi National Bank (SNB), formed in 2021 through the merger of National Commercial Bank and Samba Financial Group, is the largest bank in Saudi Arabia by assets with a balance sheet exceeding $260 billion. The merger created a technology integration challenge of corresponding scale — combining two banks with different core banking systems, customer databases, and AI platforms — while simultaneously accelerating a digital banking transformation.
SNB’s AI strategy centres on personalisation at scale: using transaction data and behavioural signals to deliver relevant product recommendations, proactive financial guidance, and contextual service interactions to millions of customers. The bank’s digital platform — Al Ahli Mobile — processes tens of millions of sessions monthly, generating behavioural data that feeds recommendation engines trained on Saudi consumer financial behaviour patterns.
The investment banking and corporate lending divisions use AI-assisted credit analysis for SME and corporate borrowers, where the combination of financial statement analysis, industry benchmarking, and cash flow modelling can be partially automated to accelerate credit decisions while maintaining analytical quality. SNB’s corporate banking AI draws on Aramco’s supply chain data, SABIC’s supplier payment records, and Ministry of Finance procurement data (with appropriate consent frameworks) to build enriched credit profiles for corporate borrowers that go beyond what balance sheet analysis alone reveals.
SAMA’s Regulatory Framework: The 2023 Responsible AI Circular
The Saudi Central Bank (SAMA) published its Guidance on Responsible AI for Financial Institutions in 2023, establishing the first formal regulatory framework for AI governance in Saudi banking. This document, which applies to all SAMA-licensed banks, insurance companies, and finance companies, represents the most detailed AI regulatory guidance issued by any Gulf central bank.
The guidance identifies six principles for responsible AI in financial services: transparency and explainability, fairness and non-discrimination, robustness and security, privacy and data governance, accountability, and human oversight. Each principle is accompanied by specific implementation requirements that have practical implications for AI system design.
The explainability requirement is particularly consequential for credit AI applications. SAMA requires that adverse credit decisions — loan rejections, credit limit reductions, unfavourable pricing — be accompanied by explanations that are comprehensible to the affected customer and auditable by SAMA supervisors. This effectively prohibits black-box neural network models for decisions that directly affect customers, requiring banks to either use inherently interpretable models (logistic regression, gradient boosting with SHAP explanations) or implement post-hoc explanation frameworks like LIME or SHAP that generate human-readable explanations for complex model outputs. Most Saudi banks have adopted SHAP-based explanation infrastructure to satisfy this requirement.
The fairness requirement creates additional complexity. Saudi labour market demographics — with significant differences in income distribution, employment sector, and financial behaviour between Saudi nationals, long-term expatriates, and short-term contract workers — mean that credit models trained on historical portfolio data may embed systematic biases against certain demographic groups. SAMA requires banks to conduct regular fairness audits of their AI systems, testing for disparate impact across protected demographic characteristics.
Fintech AI: Tamara, Tabby, and BNPL Underwriting
Saudi Arabia’s buy-now-pay-later sector has emerged as one of the most dynamic fintech segments in the Gulf, driven by the Kingdom’s young, digitally native consumer base and the regulatory vacuum (now being filled by SAMA’s BNPL framework) that allowed rapid product innovation. Tamara and Tabby, the two dominant Saudi BNPL platforms, have collectively raised more than $600 million and built AI underwriting systems that are among the most sophisticated consumer credit decisioning platforms in the region.
The underwriting challenge for BNPL is distinctive from traditional bank credit scoring. BNPL transactions are small (typically $50–$500), fast (approval must occur in under two seconds during checkout), frequent (a single user may make dozens of transactions monthly), and served to a population that often lacks extensive formal credit history. Traditional credit bureau data, which banks rely on heavily, is less informative for this population segment because many young consumers are making their first credit commitments through BNPL.
Tamara and Tabby compensate by incorporating alternative data signals that credit bureaus do not capture: e-commerce purchase history patterns, delivery address stability (proxy for residential stability), device ownership characteristics, app usage patterns, and merchant-specific repayment performance. Their AI models are trained on proprietary repayment performance data that now spans hundreds of thousands of Saudi consumers across multiple economic cycles, making the models progressively more accurate as the portfolio matures.
The risk management implications are significant. Both platforms have demonstrated that AI-driven alternative data underwriting can achieve loss rates comparable to traditional consumer lending portfolios — typically 2–4% net credit loss — while serving a population that conventional banks underwrite conservatively. This challenges the assumption embedded in traditional banking that alternative data signals are too noisy to underwrite at scale.
PIF and the Saudi Fintech Investment Ecosystem
The Public Investment Fund’s role in Saudi fintech AI extends beyond direct investment. PIF’s portfolio companies — including stc Pay (now STC Bank), Riyad Bank (PIF holds a significant stake), and various payments infrastructure investments — serve as deployment platforms for fintech AI applications across the Saudi economy.
STC Bank, evolved from stc Pay with a full digital banking licence, has deployed an AI-native banking architecture that uses ML for every significant customer decision: credit scoring, fraud detection, product recommendation, and customer lifetime value modelling. Its advantage over incumbent banks is architectural: having been built as a digital-first institution, STC Bank does not need to integrate AI into legacy core banking systems designed in the 1990s. Its data pipelines, model serving infrastructure, and decision engines were designed with ML in mind from the start.
PIF’s investment mandate includes specific targets for Saudi fintech development under the Financial Sector Development Programme (FSDP), one of Vision 2030’s realisation programmes. The FSDP aims to increase the financial sector’s contribution to GDP from approximately 6% to 7.5% by 2030, with fintech explicitly identified as a growth engine. PIF-backed fintech investment includes Sanid (SME financing AI), Lean Technologies (open banking infrastructure that enables AI applications), and several stealth-stage AI credit companies operating in the SME lending space.
Islamic Finance AI: Shariah-Compliant Recommendations and Zakat Calculation
Islamic finance AI presents genuinely novel challenges that have no direct parallel in conventional banking. Shariah compliance is not simply a constraint imposed on conventional financial products — it represents a fundamentally different conceptual framework for financial relationships, one that prohibits certain types of uncertainty (gharar), speculation (maysir), and interest (riba) while encouraging equity-based financing and asset-backed structures.
AI-driven Shariah-compliant product recommendation must therefore incorporate an understanding of Islamic finance jurisprudence that goes beyond product labelling. A recommendation engine that suggests murabaha home financing to a customer whose query indicates they are seeking a product that generates a guaranteed return — a product description that would fit a conventional certificate of deposit — must understand that the murabaha structure does not provide the guaranteed return the customer seeks, and redirect appropriately. This requires training on Shariah advisory opinion (fatwa) literature and a nuanced understanding of Islamic finance product structures that general-purpose ML cannot provide without domain-specific fine-tuning.
Zakat calculation AI addresses a specific and widespread need. Zakat — the Islamic obligation to give 2.5% of qualifying wealth above the nisab threshold to designated categories of recipients — requires precise calculation of qualifying assets. For individuals with diverse investment portfolios, business ownership stakes, and cross-border financial relationships, calculating the correct zakat base is genuinely complex. AI tools trained on Shariah advisory guidance from AAOIFI (Accounting and Auditing Organization for Islamic Financial Institutions) and individual bank Shariah boards are now available through several Saudi banks, providing customers with guided zakat calculation that incorporates their full asset picture.
ZATCA (the General Authority for Zakat and Tax) has deployed AI internally to assist in zakat audit workflows, identifying businesses whose reported zakat base is inconsistent with transaction and financial data accessible through the National Data Lake. This represents an interesting intersection of sovereign data infrastructure, Islamic finance, and tax compliance AI.
PDPL Compliance in Financial Data AI
Financial institutions in Saudi Arabia occupy a particularly sensitive position under the PDPL because they hold the most complete financial profiles of Saudi citizens and residents available anywhere — more complete, in most cases, than what the government itself holds. Credit data, spending patterns, investment positions, and insurance claims collectively constitute a comprehensive financial biography that must be protected under PDPL’s personal data framework.
SAMA’s 2023 AI guidance explicitly requires that AI systems in financial institutions comply with PDPL requirements, including data minimisation (using only the personal data necessary for the specific AI application), purpose limitation (not repurposing training data beyond its original collection purpose), and data subject rights (providing customers with the ability to access, correct, and in some cases delete data used in AI decisions about them).
The practical compliance challenge is significant for banks that have built credit AI systems on transaction data accumulated over years before PDPL was enacted. These institutions must retrofit consent frameworks, establish data lineage documentation for training datasets, and implement model auditing processes that satisfy PDPL’s accountability requirements. SAMA has provided a grace period for compliance but has made clear that AI systems processing personal data must be fully PDPL-compliant.
The convergence of PDPL requirements, SAMA’s AI governance circular, and the need for sovereign-hosted AI infrastructure for the most sensitive data applications is driving Saudi banks toward domestically hosted AI infrastructure for their core risk and compliance AI. This creates opportunity for Saudi cloud providers and AI infrastructure companies — including SDAIA’s commercial infrastructure arm — to provide compliant AI platforms to the banking sector.
Sector Outlook: Numbers That Define the Opportunity
Saudi banking’s AI transformation can be anchored in a few key numbers. Total banking sector assets of over $800 billion, growing at 8–10% annually as credit penetration increases. Thirty-five million bank accounts in a population of 37 million, with account penetration approaching saturation but product penetration — investment accounts, insurance, SME finance — still well below regional peers. Non-performing loan ratios below 2% at major banks, partially reflecting conservative underwriting but also the benefit of AI-assisted early warning systems. Digital banking adoption above 85% among retail customers at major banks, generating behavioural data that feeds AI models with the frequency and volume required for high-quality personalisation.
The fintech sector’s AI investment, supported by PIF’s mandate and a regulatory sandbox that allows rapid iteration, is producing credit and payments AI that will progressively challenge incumbent banks on customer experience while the incumbents press their advantage in data scale and regulatory capital. The next five years of Saudi banking AI will be shaped by this competitive dynamic as much as by any specific technology development.