In September 2023, Goldman Sachs quietly disclosed that it had deployed an internal AI platform capable of generating code, summarising documents, and assisting with data analysis across its banking and markets divisions. More than 10,000 employees were using the system within months. The platform was not a single tool. It was an infrastructure layer that connected Goldman’s proprietary data, internal research, and workflow systems to large language model capabilities. An investment banker preparing a pitch book could query the system for relevant deal comparables. A trader could ask it to summarise overnight market developments across Asia. A compliance officer could use it to analyse regulatory filings. The platform transformed how thousands of professionals at one of the world’s largest banks did their daily work.
Goldman Sachs is not alone. Every major global bank is now building or deploying AI platforms that sit across multiple business functions. According to MarketsandMarkets, the global AI in finance market reached $38.36 billion in 2024 and is projected to grow to $190.33 billion by 2030. The growth reflects a structural shift from point solutions (AI for fraud, AI for lending) to integrated AI platforms that operate across the entire bank.
From Point Solutions to Platform Architecture
The first wave of AI in banking deployed standalone models for specific tasks. A fraud detection model evaluated transactions. A credit scoring model assessed loan applications. A chatbot answered customer questions. Each model operated independently with its own data pipeline, infrastructure, and team.
According to Mordor Intelligence, the AI in fintech market is projected to grow at a compound annual growth rate exceeding 20 percent through 2029, driven by demand for automated fraud detection, credit scoring, and customer service applications.
Research from McKinsey’s 2024 analysis indicates that organisations deploying AI at scale report efficiency improvements of 15 to 25 percent within the first 18 months of production implementation.
This approach worked for individual use cases but created fragmentation. A bank might have 50 separate AI models running across different departments, each built by a different team, using different frameworks, drawing from different data sources. The fraud model had no connection to the customer service model. The credit model could not access insights from the transaction monitoring model. The bank had AI, but it did not have an AI capability.
AI platforms solve this fragmentation by creating a shared infrastructure layer that all AI applications use. The platform provides unified data access (all models draw from the same data lake), shared model infrastructure (models are trained and deployed through a common pipeline), consistent governance (all models are subject to the same monitoring and compliance frameworks), and cross-functional intelligence (insights from one model are available to others).
JPMorgan Chase’s approach illustrates the platform model. The bank employs over 2,000 data scientists and machine learning engineers, but they do not each build AI from scratch. They work within a common platform that provides access to the bank’s data, pre-built model components, and deployment infrastructure. When a team in the consumer banking division builds a new credit model, that model can access data and insights from models operating in the commercial banking, trading, and asset management divisions. The platform creates compound intelligence: each new AI application makes the entire platform more capable.
What AI Platforms Do for Banks
AI platforms in banking typically provide capabilities across five layers.
Customer intelligence. The platform aggregates data from every customer touchpoint, including transactions, app interactions, customer service contacts, product usage, and external data sources, to create a comprehensive view of each customer. This unified view powers personalisation, churn prediction, and cross-sell recommendations. Personetics, an AI platform used by banks including US Bank, Royal Bank of Canada, and Santander, analyses customer transaction data to generate personalised financial insights and product recommendations. The platform serves over 150 million customers globally through its bank partners.
Risk and compliance. Grand View Research reports that risk management held 27.9% of the generative AI in financial services market in 2024. AI platforms consolidate fraud detection, credit risk assessment, anti-money laundering monitoring, and regulatory reporting into a unified risk layer. This integration matters because risk signals in one area often indicate risk in another. A customer whose transaction patterns trigger a fraud alert may also warrant a review of their lending exposure. An AI platform that connects these signals can detect risks that siloed systems miss.
Operational automation. The platform automates document processing, account reconciliation, report generation, and workflow routing across departments. JPMorgan’s COiN platform processes 12,000 commercial credit agreements per year, extracting key terms in seconds. Extending this capability across all document types, from mortgage applications to insurance claims, requires a platform approach rather than building separate automation for each document type.
Employee productivity. Morgan Stanley’s AI assistant, built on GPT-4, allows 16,000 financial advisors to query the firm’s research library using natural language. The system synthesises information from hundreds of documents in seconds, work that previously took advisors 30 to 45 minutes. Goldman Sachs’ internal AI platform provides similar capabilities across its banking, trading, and asset management divisions. These tools do not replace employees. They make each employee significantly more productive by automating research, analysis, and document preparation.
Product development. AI platforms accelerate the creation of new financial products by providing tools for rapid prototyping, A/B testing, and market analysis. A product team can use the platform to analyse customer behaviour data, identify unmet needs, prototype a new product, test it with a subset of customers, and iterate based on results. The data infrastructure and analytical tools are already in place. The team does not need to build them from scratch for each new product.
Three Models of AI Platform Adoption in Banking
Banks are adopting AI platforms through three distinct approaches, each with different cost, speed, and competitive implications.
Build internally. The largest global banks (JPMorgan, Goldman Sachs, Bank of America, HSBC) are building proprietary AI platforms. This approach provides maximum control and competitive differentiation but requires massive investment in talent, infrastructure, and time. JPMorgan’s AI and data science organisation includes over 2,000 specialists. Not every bank can afford to build at this scale. The advantage is that a proprietary platform creates capabilities that competitors cannot buy.
Buy from specialised vendors. Mid-size banks and credit unions are purchasing AI platform capabilities from vendors like nCino (banking operations), Temenos (core banking with AI), Personetics (customer intelligence), and Featurespace (fraud and financial crime). This approach provides faster deployment at lower upfront cost but does not create competitive differentiation because every bank using the same vendor has access to the same capabilities.
Partner with technology companies. Several banks have formed strategic partnerships with large technology companies to build AI platforms. Microsoft provides AI infrastructure to HSBC and Morgan Stanley. Google Cloud partners with Deutsche Bank and CME Group. AWS provides banking-specific AI services to multiple financial institutions. These partnerships give banks access to the most advanced AI technology without building it in-house, but they create dependency on the technology partner and raise concerns about data sovereignty and operational concentration risk.
Fintech Platforms as the Benchmark
While traditional banks build or buy AI platform capabilities, fintech companies have already been operating as AI platforms since their founding.
Revolut’s technology stack is essentially an AI platform that delivers banking products. Every function, from customer onboarding and identity verification to transaction categorisation, fraud detection, credit decisioning, and customer service, runs on machine learning models that share data and improve each other’s performance. Revolut did not retrofit AI onto a legacy banking system. It built a banking system where AI is the operating layer.
Stripe’s platform approach is even more explicit. Stripe provides payment processing, fraud detection (Radar), lending (Capital), treasury management (Treasury), and financial reporting (Revenue Recognition) through a unified platform. Each product generates data that improves the others. Fraud patterns detected through Radar inform risk models used by Capital. Payment flow data from processing informs treasury management recommendations. The platform’s intelligence compounds with every merchant and every transaction.
Nubank serves 100 million customers with a fraction of the staff that a traditional bank would require. The efficiency comes from an AI platform architecture where customer service, credit assessment, fraud detection, and product personalisation all operate on shared infrastructure. The company’s customer acquisition cost of approximately $8 per customer, compared to $40+ for traditional banks, is a direct result of platform-level AI automation.
These fintech companies set the benchmark that traditional banks are trying to reach. The gap between a purpose-built AI platform and a legacy banking system with AI bolted on is measured in years of development, billions of dollars of investment, and fundamental differences in organisational culture. The banks that close the gap fastest will survive the transformation. Those that do not will find themselves competing against institutions that can serve customers at a fraction of their cost and a multiple of their speed.
