Technology Alpha

How the tech stack is transforming the investment management process

Two things are certain in the Age of AI: the volume and variety of data will continue to increase; and the tools available to extract signal from that data will become more powerful and easier to use. These trends are transforming every industry, and asset management is no exception.

Much ink has been spilled about the proliferation of data. Less attention is typically paid to the rapid improvements in the tools available to manage and analyze all this data. As the cost of compute continues to decline, it becomes easier and more affordable to develop models that have the power to extract investable signal from this rich array of data.

But the story is about much more than models. Better data management and analytics tools are accelerating every step in the investing value chain:

• Accelerating the discovery and testing of new data sets that contain investment signals
• Simplifying data governance and cleansing to expedite onboarding
• Improving indexing and storage to enhance semantic search
• Expediting the creation of apps and workflows that translate data into better decisions
• Creating a richer feedback loop to enable deeper understanding of how decisions impact investing outcomes.

Data might be the new oil, but oil is useless without infrastructure to extract, refine and distribute it. We’re facing a future where there will be so much data – and such a variety of data – that best-in-class tooling to separate the signal from the noise is becoming the most important capability for investors to have.

I’ll illustrate with a project Microsoft recently did with a large sovereign wealth fund:

• October 2023: The global head of PE reached out to discuss how AI could help his business. We presented the concept of a Deal Assistant, largely focused on helping analysts find and summarize information in deal documents. After some job shadowing sessions to understand key analyst tasks, we scoped an MVP and estimated the benefits in a business case.
• November 6th: GPT-4 Turbo was released. Its larger context window and vision capabilities unlocked new opportunities so we shifted our scope and development got underway in late November.
• February: The prototype was complete, featuring: Data extraction from tables, charts and infographics; a tagging technique that improved vector search and generated more accurate, relevant results; automated pre-processing of large volumes of documents.
• March: Research Copilot MVP was complete, empowering the customer with a tool that could: Perform multi-modal Q&A on large volumes of complex pitch decks; translate extracted data into a simple financial model in Excel; draft deal memos for the Investment Committee.

In less than six months, this asset manager got a tool that enables a material increase in the pace of investment decision-making. This was possible because the development process leveraged application building blocks like document preparation, vector search and prompt orchestration. And because the project has been open sourced, others will be able to replicate and extend it to drive new innovations even faster.

Development tools are improving at an accelerating rate because they’re harnessing previous waves of tech innovation: connectivity, cloud computing, APIs, IoT devices, developer platforms and more. Each new innovation is adopted more quickly and with greater impact because it leverages the foundation of everything that has gone before it.

AI development tools have progressed rapidly in the 18 months since ChatGPT was released:

• Individual models confined to generating text, images or code have been merged into single multi-modal models;
• Platforms host models-as-a-service to enable developers to select the optimal model for their task, cost, and performance needs;
• Platforms integrate all the components developers need to build AI applications – enterprise data, data preparation tools, prompt orchestration tools, and a plethora of AI safety guard rails;
• Inferencing has become 12 times cheaper and six times more powerful;
• Most strikingly, many of these experiences are accessible to citizen, developers, without the need to write any code

Given the increasing power of these tools to unlock investment insights, technology is rapidly becoming a key source of alpha generation. We’re entering an arms race in which asset managers that learn to effectively incorporate AI into their investment process will leave their peers behind. Success requires three things:

• understanding how the latest tools are evolving across the tech stack
• having processes to rapidly make those tools available to investment professionals
• ensuring that those professionals have the skills to use them.

The integration of AI into asset management isn’t just a trend, it's driving a paradigm shift in the investment process. It requires new skills, incentives and organizational designs that have the potential to substantially change the competitive landscape in the years ahead.