Ashmita Gupta, global head of analytics at Linedata, and David Boot, senior product manager, explain the promise of generative AI to increase operational efficiency and support better results.
Over the past decade, automation has taken giant leaps in the financial services sector. However, many manual tasks remain in day-to-day asset management and fund operations processes. Advances in data analytics and the rapid development of Generative AI promise efficiency gains and enhanced investor services. From complex document analysis and report generation to client information processing, risk management, and compliance, Generative AI can eliminate drudge work while creating time for higher-value activities.
Ahead of the Curve
Generative AI has taken the world by storm. In the asset management space, it builds on existing analytics and Artificial Intelligence / Machine Learning (AI/ML) capabilities, explains Ashmita Gupta. For some years, Linedata has used descriptive analytics to help asset managers understand and address process-outcome correlations, such as trade failures and compliance breaches. The next step has been helping firms proactively manage operational issues, using AI/ML-enabled predictive analytics to anticipate and prevent errors and potential losses.
Now, Linedata is starting to apply Generative AI – in the form of private large language models (LLMs) – to a range of use cases. One example is reporting. Asset managers devote much effort to crafting customised reports for internal and external use. Generative AI can streamline and transform the report-generation process. Each client’s data requirements, whether institutional, wealth-focused, or family office-based, can be tailored. This is especially pertinent for asset managers with numerous smaller clients that require detailed reports, explains Boot. Standard reports can be enriched by integrating diverse information with specific client details. Automating such processes enhances time allocation for ‘money generating’ activities and enables investment firms to offer better bespoke services to clients.
Futuristic Approach
Generative AI can ease access to cumbersome data sets across an organisation, eliminating the need for manual scanning and data extraction. Boot highlights that business intelligence is often based on generating insight from data, but it can still be static. Generative AI adds dynamism to this process. It helps fund managers understand and address ongoing changes and instantly analyses the implications of daily news events on their holdings.
While many people think of Gen AI in terms of content generation, Linedata pursues broader applications, not least because of its ability to mimic humanlike conversations and generate answers to questions. For example, Gupta explains, when asset managers onboard new funds, they review reams of documentation to extract relevant information. Gen AI can help automate this process – making the process more efficient and less onerous.
Rapid data querying with Gen AI can also accelerate decision-making. The use of a ‘digital assistant’ expedites information extraction for traders and portfolio managers. The cherry on the cake is improved employee morale with more time to focus on meaningful activities.
Risk Management
Generative AI’s language models do carry ‘hallucination risk’ – in which AI makes up false information or ‘facts’ that aren’t based on real data or events. However, training models can mitigate this by flagging the doubts – rather than guessing the default answer. For instance, Linedata helps clients manage this risk with LLMs that provide references for potentially suspicious data, enabling fact-checking and improved reliability.
Another concern is the use of confidential data in public Generative AI platforms. Putting proprietary or customer data into a public LLM raises a host of privacy, information security, and intellectual property concerns. “Data must be protected against unauthorised third-party access and handled in the most compliant way,” says Boot. Linedata’s highly secure private LLMs are sealed off from the internet and restrict data access to authorised individuals.
While some overall risk is inherent in Generative AI, particularly if poorly implemented, the technology does show great promise in terms of operational risk management. One pertinent area is that of NAV calculation. Delays or errors in striking the NAV on funds are a source of monetary, commercial, and reputational risk, says Gupta. If NAV calculation issues are escalated or changes are required, this typically take place after the market has closed, complicating issue resolution. For more illiquid securities, reaching out to other trading desks takes time. She adds: “Generative AI can help traders and third-party data providers by scanning the news articles throughout the day to look for any anomalies related to a specific security or issuer, speeding up investigation by the valuation team.”
Data Extraction
Data extraction is fundamental to a broad range of asset management processes. Creating compliance rules based on individual investor requirements is time-consuming – from reviewing client and regulatory documentation to configuring and assigning compliance rules and updating them as requirements change. Asset managers can use LLMs to summarise rules and regulations so they can be easily translated into the system – a faster, less risky process for onboarding clients and maintaining up-to-date control processes.
Compliance is one of the most requested use cases at Linedata, but Boot and Gupta also note client interest in investment research and due diligence. For example, Linedata’s clients in the unstructured credit and private equity space must scan vast, unstructured data sets to make investment decisions. Extracting information and building company profiles can take weeks.
Enter Gen AI. A chatbot can be asked questions such as “What are the positive and negative covenants?” and “What are the underlying interest rates or debt secured?” AI-generated answers help research analysts complete due diligence much more quickly, enabling portfolio managers to make better investment decisions faster and with greater confidence.
Tangible Gains
While Linedata measures success for its Generative AI use cases using predefined metrics, the most significant success factor is time savings, which, in turn, means cost savings. Added accuracy is also an advantage. “Going from 99% accuracy to 99.5% using AI might not seem like a big change,” says Boot. “In fact, it means identifying 50% of problem cases, which is where most time is spent.” In short, Gen AI can help data scientists and empower non-technical teams – including senior management – to make better decisions faster.
Before jumping headlong into Generative AI, however, firms must understand where it would add value – and whether they need external support on their Gen AI journey. In case of the latter, the hallmark of a good potential partner is a deep understanding of the prospective client’s business and operations, and which (generative) AI use case is most appropriate.
Asset management firms often have highly competent technology teams with AI expertise, but that doesn’t guarantee results. As Gupta explains, building a proof of concept is one thing. Integrating LLMs within the company’s ecosystem and running and retraining them to meet ongoing business objectives is another.
And it’s not just about making the technology work; industry expertise is also critical. “If you don’t understand the problem, you can’t apply Generative AI to it successfully,” she says.
