Can AA systems and AI unlock innovation in mutual funds?

The panel on ‘AA and AI—What’s the next wave of innovation in Mutual Funds’ comprised Varun Sridhar, CEO, Paytm Money; Chetan Mehra, head - quantitative investment, PMS & alternative investments, Bandhan AMC Ltd; and Vetri Subramaniam, chief investment officer, UTI AMC.

Edited excerpts from the panel discussion:

What, according to you, will be the next big innovation in the MF industry?

Chetan: I see passive investing and a lot more quantitative investing coming through big time.

Varun: There are two ends of the spectrum to look at. One, how do you get UPI (unified payments interface) users to invest in MFs. There are about 300-350 million UPI users but the MF base is small. It is easy for people to buy something for ₹500 with UPI but to make your first investment in MF for ₹500 involves hopping through KRAs and other things. So, an asset management company (AMC) or a tech company like us will flip this around to make investing easy with UPI. The day that happens, this whole sector will see an important change.

The second breakthrough will be the account aggregator system. On the other end of the spectrum are people with more wealth and are advanced investors. When asked what their wealth is, people generally calculate their household wealth. But, because each family member has an independent PAN and thus, different accounts, it is difficult to determine and refer to the family wealth at one place. I think the account aggregator system can flip this problem statement and make it easy to see one’s wealth in an app.

Vetri: We are a fiduciary business and I wouldn’t prioritize innovation for the sake of it. Some of the innovations have happened in the external ecosystem and we need to figure out how we can incorporate that into our business model to interact and educate new customers so as to onboard them. These are primarily the areas where we would like to use innovation.

How can data sciences, AI and machine learning (ML) help in aiding these innovations?

Vetri: When it comes to using AI in business, from an asset management point of view, I think, we are in the business of alpha. For the purpose of creating additional value for investors, AI can be used for pattern recognition and in the allied domain for better risk management. The second way in which AI can be used by an AMC is to simplify transaction- related tasks. We are in a high volume transaction processing business–right from the trades which flow through us to customer level transactions. There are different outsourced bodies and stakeholders that work with us on this front but, I think, a lot of that can be simplified by using AI. Finally, when it comes to conversations with the customers, both transactional or educational, there is tremendous scope to use AI.

To my mind, all good investing is a combination of science and art. We can debate which of these two has a higher share, but it is definitely a combination. Therefore, AI can play a significant role on the science part but you will eventually need a little bit of art to kick in, particularly in the active side of the business. As a good pilot once told me, when you hit a crisis, it really matters that you have an experienced pilot at the controls, you don’t rely on auto-pilot at that point of time.

Chetan: When you have good quality data, you can use ML to make better models. In fact, most ML will improve an existing model. In terms of using ML for that, you will always try to use an existing standard quant model.

When you have a lot of customers, as is the case with the MF industry, you have a lot of data. That means you can create profiles of customers with information on their income, where and how they like to invest and their holding periods. Using this information, you can create the right products to sell to the customers. So, I think, in sales, one can do a tremendous amount of things and the more data you have the better it is. But most of the time, the challenge is how do you treat the data.

The biggest challenge with most ML problems is that generally people without domain knowledge are hired and told to do something with the data. What you really need is domain knowledge to figure out what the boundaries are within which you will use ML, and how you will treat and interpret the data. On the face of it, this looks simple but this is a huge amount of work. We spend 75% of our time figuring out whether the data quality is good, which is a big challenge in India. If you don’t have good data, you can’t really do ML or AI because ML is essentially an optimisation function–with bad data it will just optimise bad data.

Varun: AI can solve the problem of education. MF’s default language is English–NAV, schemes, hybrid, multi-cap, mid-cap, equity, debt, etc. For people who are not in capital markets and have a different mother tongue, it’s impossible to understand all this. Here’s what AI can do. If we were to get basic domain knowledge in, we can have a Hindi only investor app, or an app with a mix of English and any vernacular language. This makes it easy to communicate during onboarding what these terms mean, or inform the investor of a market correction in the local language. Data sciences can convert the funnel from all the way to generating curiosity to onboarding to getting the first SIP started by simply solving the language problem.

Do you think the account aggregator (AA) system can make access to quality data easier?

Chetan: Yes, it can definitely enhance data usage dramatically. But, India still doesn’t have data protection laws. So while we can have all the data we need, soon the law will be out and we don’t know what kind of barriers it may create. That said, to have access to data and to be able to extract powerful information out of it is one thing but you have to know how to do it right.

In a recent consultation paper, Sebi has questioned whether AMCs and brokers should be excluded from functioning as financial information users (FIU) in the AA system. What are your comments on this?

Vetri: As an AMC, we see ourselves as manufacturers of products that investors use. They either use them themselves or use advisors. I’m not sure we, as asset managers, want to cross the Rubicon and be on the other side and will really need all that data.

Varun: The direction of the consultation paper is right. The most important thing in the AA system is consumer data privacy protection because they must understand what they are signing up for. In my opinion, regulated entities should be allowed to function as FIUs. I remember when I was in the lending business in 2006-07 and Cibil was launched, there were questions raised on the inaccuracy of data and how it won’t work. Fast forward to now, the lending space has done a fantastic job of getting it together. Similarly, AA makes a lot of sense in the long run as long as it is run conservatively and with consumer protection at the highest level.