The Emergence Of AI-Driven Processes In Finance

0
The Emergence Of AI-Driven Processes In Finance

Raghavkumar Parmar is a portfolio manager at MMA Pan Asia Fund Management.

It’s no exaggeration to say that AI has had a meteoric impact on nearly every industry today, with experts predicting an annual growth rate of 37.3% from now until 2030. However, even before the current generative AI boom, major tech companies were already incorporating artificial intelligence to improve product and service offerings. Today, the use cases for AI are finally here not only for businesses but also for individual consumers (many of whom may not be attuned to the technical intricacies of these tools). Importantly, we have also reached a stage where AI can also help create products on its own based on instructions and algorithms fed to LLMs (large language models).

One industry that can benefit significantly from generative AI tools is finance. Interestingly, finance has been in some ways an early adopter of AI tools, but the uptake has been steadier and less sharp than in other industry spaces. Corporate finance, in particular, relies on several different applications of generative AI models to perform such tasks as budget analysis, scenario analysis and data analysis. One of the niche sub-segments of the finance industry, the hedge fund space, also has some interesting use cases for AI. Here, these tools could allow users to save time and make more efficient decisions:

1. Financial model building: AI can be used to build financial documents such as income statements, balance sheets and cash flow statements for publicly listed companies based on their 8-K or 10-K annual reports and earnings releases.

2. Collecting data: AI models can be built to create algorithms that can scrape data through public websites; these data (which are hard to download manually) can be used to help make investment decisions. For example, AI tools can be used to scrape pricing data for airlines or cruise lines, car sales data, etc., to predict sell-through rates.

3. Data analysis: The data collected through AI can be re-fed into LLMs to train them, allowing these models to infer data in order to drive more informed investment strategies.

4. Event-based trading: Using AI tools, algorithms can be built to trade publicly-traded securities in microseconds following the release of new information, thus allowing real-time trading driven by specific events such as earnings releases, conference calls, investor days or launches of new products.

5. Parsing industry contracts and government restrictions: Government restrictions and industry contracts can be read in a matter of seconds using AI tools, in order to help portfolio managers make time-sensitive decisions that both account for and remain compliant with current regulatory environments. For example, AI tools have been used to parse the impact of the government ban on exporting semiconductor chips to China on global supply chains.

6. Producing AI-based financial commentary and presentations: AI tools can generate real-time transcripts of earnings calls and compile the data from presentations and earnings releases to update financial statements.

7. Portfolio risk management: Risk appetite metrics of portfolio managers can be fed into an AI-driven system, with AI keeping track of risk metrics in real-time and alerting portfolio managers when the metrics are breached.

Of course, concerns around AI remain an industry priority, particularly when the conversation turns to the use of sensitive financial data in these systems. How do we prevent AI from being fed with and then producing data that will lead to erroneous conclusions? Data verification and validation are important. On the training side, we have to make sure we are feeding the right kind of data into AI tools—that we aren’t feeding data with a lot of “one-off” numbers, which would then become normalized. Essentially, we have to teach the AI that certain data are incorrect and should be discarded.

On the inference side of AI, we have to understand whether the output that comes out is reliable or not. This may require a second set of human eyes, reviewing AI output and reviewing it against previous human-generated data. Having a two-step authentication process is one of the ways to make sure that the data is inferred correctly, and the model is trained right.

But what if there is a data security breach? In the current infrastructure setup, data breaches are common, and security parameters for non-AI-based data systems would still apply: When AI is programmed to execute certain sets of tasks, they still have to pass through automated compliance checks. Some of these automated checks are better than those performed by humans, simply because it is easier for a computer to determine if, say, it is being attacked by MS-DOS, or if there is other malfeasance going on. When an AI is programmed correctly, it can actually promote a higher level of data security than is possible purely with human oversight.

Is that to say humans are no longer needed? Certainly not. Particularly in the financial sector, human review, analysis and judgment are part and parcel to successful decision-making and long-term strategies. However, by infusing these processes with AI tools and the wide range of capabilities they offer, these decisions and strategies are greatly improved.


Forbes Finance Council is an invitation-only organization for executives in successful accounting, financial planning and wealth management firms. Do I qualify?


link

Leave a Reply

Your email address will not be published. Required fields are marked *