Generative AI (GenAI) has the potential to transform the asset management industry at just the right time. After nearly two decades of strong performance, there is pressure on firms to cut costs, diversify revenues, and introduce more personalized services.

GenAI is a synthesis machine. The underlying models absorb large amounts of information, are trained to understand the context and meaning, can answer seemingly abstract questions, and are able to create net new information, including copy and images. These models also learn incredibly fast. Deployed at scale, GenAI is well-poised to enhance and disrupt asset management, which at its core is a knowledge-based industry where information is consumed, processed, and created, and trillions of dollars in client assets are managed.

In this article, we consider the possible medium-term impact of GenAI on asset management. We explore GenAI’s impacts on performance, address security and risks, and offer some practical thoughts on how asset managers can begin their transformation journeys.

Five Potential Impacts

We identify GenAI’s five major applications that are at the heart of the asset management business and have the potential for significant impact.

Improved Operating Efficiency. There’s potential for 10%–15% improvement in operating efficiency if AI were to be implanted at scale to augment current job functions. For some functions, the range may be as high as 40%–50%. (These estimates are based on our analysis of each step in the value chain and each function and subfunction.)

GenAI can be deployed to assist and significantly expedite work in departments such as marketing, finance, and human resources. In marketing, for example, GenAI is already assisting with design, copy writing, image selection, and compliance checks.

Personalization at Scale. The content creation and synthesis capabilities of GenAI make customization achievable across large numbers of clients/customers. BCG estimates that gains of 30% or more on engagement and 5–10% on sales are achievable.

In the years ahead, owning the customer experience will be as important as offering excellent products. However, until recently, weak data foundations, insufficient data science capabilities, and difficulties with change management (convincing salespeople to use new capabilities), have constrained asset managers from delivering to the maximum potential.

GenAI enables a step change in the ability to offer individual customers what they need. Historically, sales enablement teams spent large amounts of time precisely organizing data sets. Today, GenAI can leverage data sets and infer the meaning and intention of questions being asked. (For example, “Which of my clients are showing signs of potential redemptions?”)

Adoption can also be significantly improved. In the past, a major stumbling block was sales professionals not believing model outputs and encountering challenges in explaining them. The answer was usually along the lines of “trust me.” Now, GenAI-enabled models create two-way natural language interfaces that make it possible to have a conversation and ask questions. If the answer to a question does not satisfy the salesperson, they can dig deeper and get confirmation or flag a potential issue to improve the model.

In the future, GenAI could also help salespeople do quick meeting prep on the road, send notes which can be processed and submitted by GenAI in summary form to customer relationship management teams, and routinely access powerful recommendation engines. Much of this is already in development and promises to be a major unlock for personalization at scale.

Knowledge Compounding. True cross-enterprise knowledge sharing—the ability of an organization to never forget, to always have the best thinking available, and to share seamlessly across silos—has been the great unfulfilled promise of asset managers for decades. GenAI can finally help unlock this potential by easing the burden of knowledge retrieval through a natural language user interface and human quality synthesis.

GenAI is already helping organizations overcome siloed structure of investment teams, frictions around uploading and tagging documents, and challenges related to quality control and compliance by streamlining processes and making the user experience more intuitive.

Knowledge systems powered by GenAI can “understand” the intent of a question, rather than just the words. This enables retrieval of groups of related documents, including specific words and images. For example, a GenAI query could be “I am interested in our firm’s thinking on the merits of buy versus build decisions for AI capabilities, with emphasis on cost considerations.” This level of depth would not have been possible just a few months ago.

The benefits of unlocking knowledge include reducing key person risk, more effective information sharing across asset classes, better/easier access to key documents and/or policies for enabling functions, and reduced time and cost spent recreating old documents. Organizations are already re-engaging on knowledge with GenAI, and the trend is likely to continue.

Research Accelerator. The speed and quality of investment research is poised to accelerate and challenge the competitiveness of larger asset managers that have historically benefited from big teams of researchers and data scientists. In the near future, GenAI-enabled research assistants will help smaller firms level the playing field by creating a “synthetic” army of research analysts at a low marginal cost.

Already investors are able to interact with data sets using intuitive natural language interfaces to generate investment hypotheses and even first draft investment memos. In the near future, GenAI will likely move up the value chain to suggest holistically reasoned investment views, trading strategies, and risk controls. However, quality oversight will be important, especially until the models can be fine-tuned and governed with appropriate guardrails.

Within these parameters, the potential for faster research, new idea generation, and reduced key person risk are significant. Simply having more people won’t necessarily be a competitive advantage. Instead, the investment edge in research will increasingly rely on the ability to connect GenAI models to compound knowledge faster than peers.

Democratization of Coding. English and other languages are the new coding languages. GenAI has already changed the game for application development, with functionality that can accept a natural language prompt and auto generate the required code. As a result, GenAI can enhance the software development process, allowing for greater democratization of code development, higher quality applications, and reduced reliance on offshore resources.

Today, technical employees such as data scientists and developers spend significant time on low-value query and code writing. As a result, they face backlogs in development and often deploy buggy applications. GenAI can reduce the number of people needed for the same work as more people will be able to create code. The technical experts will be freed up to take on more complex coding and testing tasks.

The operating model implications are potentially profound. We expect technology organizations to look very different in the near future—much smaller teams, more onshore capabilities, and more focus on high-value work powered by GenAI.

Where to Begin

Given how fast GenAI has taken off, the first step is to start with foundational education, so that everyone—including the Board, C-suite, and staff—understands its importance. This will be critical to creating a level playing field across the organization. Firms that invest time early will position themselves to navigate GenAI the most effectively.

All asset managers will need a proper GenAI strategy to maximize the benefits and minimize risks.

In particular, organizations need to tackle three questions:

Where is GenAI going to be a source of competitive advantage (and therefore should be managed internally)? If it is not a source of competitive advantage, will the market provide the service through partners? GenAI is likely to challenge historical sources of advantage (large data science teams). The role of these teams needs to be revisited. At the same time, most service providers are embedding GenAI into their products, which should unlock efficiency benefits across many functions. Asking those deeper questions early on will help shape the organization’s focus.

Which applications/use cases should we start immediately to maximize learning and impact across the enterprise? Thoughtful experimentation is key. GenAI is new for almost everyone, so finding the right fit in any organization will take time, as well as trial and error. The most advanced firms have launched multiple experiments. These are designed to balance impact and learning (meaning even if the experiment fails, the organization learns something valuable). Pilots are usually orchestrated from the center but led day to day by leaders who are willing to push boundaries.

How do we manage security and compliance to enable learning while avoiding the risks and pitfalls that could damage the business? GenAI is powerful and therefore also dangerous if not managed effectively. Firms are already taking action to mitigate the obvious security risk of data being disseminated outside the organization. The risks within the organization are more challenging. There are questions around the accuracy of GenAI created content, the provenance of data, and quality control. Firms need to think about regulatory requirements for certain types of data (especially financial reporting). A crystal clear set of responsible GenAI principles and controls will be critical.

GenAI can be a positive breakthrough for asset managers at a critical time for the industry. The natural approach will be to “wait and see,” as with most technology trends. However, we believe waiting is not an option because of the speed at which GenAI is growing and will continue to grow. As a matter of urgency, asset managers need a forward-looking GenAI strategy. This will position them to move forward with confidence, mitigate risks, and reap the benefits of a powerful new technology.

The authors thank BCG colleagues Julie Bedard, Jeanne Kwong Bickford, Ingmar Brömstrup, Juliet Grabowski, Helen Han, Sam Lambert, Kedra Newsom Reeves, Ian Pancham, Neil Pardasani, Ella Rabener, and Michelle Stohlmeyer Russell for their contributions to this article.