AI and the Future of Intelligent Investing

Ensuring that massive volumes of disparate structured and unstructured data is fit for purpose requires a solid data management foundation capable of capturing, curating, enriching and delivering data sets to AI and ML modeling engines. Time to information is also critical. Many use cases require forecasts and predictions in near real time due to the short shelf life of actionable investment and risk data.

Legacy databases, spreadsheets and data silos present major obstacles to leveraging the benefits of AI. The considerable cost and effort of connecting these disparate data sources to a centralized, AI-ready repository rarely delivers the expected benefits. A new data management paradigm is required.

The ability to harness AI at scale is underpinned by significant advances in cloud computing. Data scientists can train massive AI models at whatever scale and speed they want, simply by requisitioning more computing resources from the cloud. Transparent usage metering enables firms to make informed decisions balancing computing cost with model training times. Additionally, investment firms can improve operational effectiveness by outsourcing non-core activities like server maintenance, performance monitoring and software upgrades to their cloud provider, thereby reducing the need for costly internal operations teams.

Deploying technology like the Snowflake Data Cloud also increases scale, by eliminating data movement – data is accessed in-place – enabling complex data sets for AI/ML modeling to be assembled in seconds. Financial time series and in-database feature engineering are natively supported capabilities that are key to building and training AI applications. This enables analysts and data scientists to focus on data exploration, hypothesis testing and other value-added work instead of manual data assembly.

A growing number of investment firms, liquidity venues and service providers are harnessing AI and cloud computing to uncover investment opportunities, improve staff productivity and eliminate low-value, time-consuming manual processes.

AI applications can be grouped into six broad categories. Generative AI chatbots provide customer-facing service teams and financial advisors with curated knowledge for more productive customer interactions, and help middle -office operations specialists automate routine trade reconciliations, freeing staff to focus on transactions requiring human intervention. Portfolio managers, traders and analysts can use AI-assisted chat to ask natural language questions about their portfolios, trades and counterparty exposures.

Data governance applications include detecting and escalating anomalies in mission-critical financial data, escalating suspicious records to human analysts to validate and remediate potential errors.

The ability to parse documents, spreadsheets and other unstructured sources helps inform portfolio construction, flag suspicious AML/KYC transactions and perform valuation and cash flow analysis of private equity and other alternative asset classes.

Software developers are harnessing applications that generate code, freeing them to focus on algorithm development. Image analysis is being used by hedge funds to generate buy and sell signals based on chart pattern recognition. Finally, AI algorithms are creating visualizations of investment performance, risk exposures and other portfolio analytics.

State Street Global Advisors has partnered with an AI-based analytics provider to uncover emerging companies focused on innovation, such as drones, biotechnology, electric vehicles and sustainable energy. The provider deploys an LLM-based application to scan through thousands of regulatory filings, aggregating the frequency that a particular term appears, and its accompanying context. The LLM then categorizes each company into one of 25 innovation buckets that are used by the ETF manager to inform portfolio construction of their popular “innovation ETF.”

FundGuard, a State Street Alpha strategic partner, is building a new generation of smart investment accounting systems that leverage AI to enhance processing capabilities, integrating both traditional statistics and ML models. This integration enables far greater efficiency and accuracy through a low- to no-touch process that reduces tedious, error-prone reconciliations. Automatic price checks and real-times analysis of factors impacting a specific security provide investment professionals with greater confidence in the veracity of their portfolio and risk data.

The Alpha Data Platform incorporates continuous learning neural networks trained across multiple data domains including reference, pricing and corporate actions to detect anomalies. While traditional rules-based validation methods flagged more than 31,000 data exceptions across a six-month period, of which there were only 250 true exceptions, the neural network AI identified only 4,000 exceptions while discovering 100 percent of the true exceptions1. The massive reduction in false positives means significantly less work for human experts tasked with investigating each exception as a potential error.

State Street Alpha for Private Markets uses natural language processing to parse large volumes of unstructured data contained in spreadsheets, call transcripts and prospectuses to help alternative asset managers expedite decision-making and evaluate new investment opportunities ahead of competitors. This has the potential to streamline manual workflows that previously required significant numbers of analysts.

LTX, a Charles River liquidity provider, empowers institutional traders with its BondGPT application, which provides LLM-based bond information on individual CUSIPS, sectors, companies and market trading activity. Powered by GPT technology together with bond and liquidity data and driven by LTX’s proprietary models, BondGPT saves traders time by providing accurate, timely responses to complex bond-related questions.

In middle- and back-office custody and accounting functions, our State Street team of more than 400 data scientists, engineers and analysts have deployed numerous other AI applications. These include an ML predictive NAV benchmark, inadvertent data disclosure prevention, client file quality controls, anomaly detection applied to vendor market data, ML-based predictive reconciliation matches and break resolution, as well as digitizing unstructured data from broker confirmation statements, trade instructions, bank loan agent notices, funding memos and many more. Our team is now actively developing Generative AI use cases that are poised to radically transform how our clients interact with all of our services, enabling conversational access to knowledge, documentation, trade status, portfolio data and news.

These use cases are just a few of many examples demonstrating how we’re helping forward-thinking organizations leverage AI, cloud computing and modern data management solutions to empower their employees, clients and partners with new insights for better-informed decisions, improved operations and automated workflows. To learn more, visit statestreet.com/alpha.