What the AI?!... Schroders Capital

The Drawdown (TDD): Can you tell me about your data team?

Steven Yang (SY): Our data insight unit at Schroders has been 10 years in the making. It consists of more than 20 dedicated professionals, including data scientists, engineers and consultants.

AI is not new to us – we have been using it and machine learning across our business for quite some time. In the past three years alone, there have been more than 300 data projects completed to enhance the way we work.

TDD: Can you tell me more about the most interesting of those projects?

SY: We have an investment tool that has been live for almost 10 years called the ‘intelligent long shortlist’. Using machine learning, it takes data within our database and from different sources to rank a list of roughly 3,000 managers based on preset criteria. This has helped us move away from the traditional way of evaluating opportunities.

On the operational side of the business, we have tried to enhance portfolio construction and cashflow modelling, particularly relevant to our finance team. Each of our clients has its own set of mandates, objectives and constraints. We’ve started using AI to model investment portfolios around these needs. Being able to efficiently sort portfolios by geography or sector is relevant also from a risk management perspective.

Finally, in our venture programme alone, we have more than 5,000 companies in our database. So, when it comes to our quarterly reports, our data science team can perform benchmarking and call out any outliers, like if there is a significant increase or decrease in the fair market value. The AI Integrated Database can really assist the team in that process because humans would not be able to swiftly analyse 5,000 companies alone and Excel cannot support such a large dataset.

TDD: How does your cashflow modelling tool work?

SY: The purpose of the cashflow modelling tool is to have a better understanding of any real-world uncertainties associated with private assets cashflow. It simulates the cashflow to build more of an optimal asset allocation for our funds. It also helps build a more sophisticated approach towards risk and cash management.

Our clients have a mandate to make commitments and expect their assets to generate value on the commitments they've made. Their future investment activity is really tied into how much cash they expect to get back and what they can reinvest, as well as what underlying obligations they have.

In today’s market, investors are facing a denominator effect because their asset allocation for private equity tends to be significantly above target, since the value of their existing investment has increased. The tool can help to rebalance that portfolio as the valuation of assets changes over time.

TDD: What are your thoughts on incorporating generative AI in-house?

SY: It seems clear to me that global asset managers are pretty keen on implementing generative AI. The natural questions are: How do you incorporate generative AI into your existing setup and how does it adhere to compliance, security and data privacy regulation?

I think the industry is still in its infancy when it comes to generative AI adoption. The tool is dependent on how much data you make accessible to it, particularly between departments. Most companies would be reluctant to fully open that gate and let an application entirely loose into their whole ecosystem. That’s probably generative AI’s biggest use case but also its limiting factor.

Internally, we are using an automated transcript tool called Genie for our Teams calls. It can also summarise and translate meetings for our teams. These are probably lower-hanging fruits but it is definitely a time-saving activity.

One of the downsides of the types of tools that exist today is, unfortunately, the data is only up to date as of 2021. Digitally, a two-year gap is huge. So, that is a bridge we need to shorten over time.

TDD: Is generative AI on the radar for Schroders?

SY: We are currently working with a generative AI provider that operates within our firewall. This safeguards our data and ensures it does not go outside of the organisation. It is definitely an iterative process and we are in discussions with a number of other providers to discuss issues such as data privacy and compliance. We definitely tend to lean towards enterprise solutions and it is a live process.

To read more interviews in the What the AI?!... series, click here