Manish Garg on the right amount of paranoid about AI

Editor-in-Chief Sarah Wheeler caught up with Manish Garg, senior vice president of product and technology EarnOpan autonomous financial wellness platform, to talk about how his company is using gen AI to deliver a personalized experience to customers at scale. Garg has a deep background in building enterprise software and has worked with fintechs in the mortgage lending space for the past decade.

This conversation has been edited for length and clarity.

Sarah Wheeler: What sets your technology apart?

Manish Garg: In building our tech stack, we focused on borrower, financial health, compliance and risk data protection as guiding principles. We always work backwards from the desired outcome, with an emphasis on the financial health of the consumer. We work in aspects where we offer our technology to servicers, credit unions or banks to help them reduce the risk of default by doing a lot of data analysis behind the scenes to help them identify how they can reduce the risk of non-payments and reduce the risk of default. , and keeping their books healthy, but also keeping consumers in a healthy place. So we have a lot of technology for predictive analytics.

SW: How do you use AI?

For a long time, we have mainly worked with traditional AI, where we build predictive models and predictive analytics and can categorize risks and give it all back to the companies. Over the past eighteen months, things have changed dramatically.

We were in the fortunate position that we had some insight into this very early on. And that’s what we invested in from the beginning and we’ve now built core capabilities into our platform to do things that we’ve only been talking about in the industry for years. It’s like pipe dreams are finally coming true: being able to generate engaging, hyper-personalized content for consumers to help the loan officers, back office, underwriter or processor do their jobs in even more efficient ways. These are capabilities that we all hoped would one day become a reality, but they seemed like science fiction, and suddenly they don’t. Suddenly it’s here.

SW: How have AI capabilities changed over the past six months?

MG: We’ve had AI around for a while. Most people on the technical side understood and appreciated it, as did the data scientists, but for many business users the value was not clear. But for the first time, it is something that everyone can touch and feel. So that’s something that has fundamentally changed and why there’s so much adoption and why there’s so much optimism about that. The second part is doing things that seem almost magical or very, very difficult to do – which has become very easy because of large language models (LLMs) and AI.

For example, creating hyper-personalized content for a consumer. We do that a lot with our clients, where we can take in a lot of personal financial information about consumers, from their banks, from credit bureaus, from many other sources, and then the consumers can interact with their personal information to understand more. about it. That was not possible before; Previously you had to build a complete application for it, but now I can talk to my own data.

For companies and lenders, it is about competing on rates. As the refi market hopefully starts to come back with interest rates falling, everyone will be competing for the same group of borrowers. They are inundated with very similar offers, such as a lower rate. But now someone can actually use gen AI, and if they work with us, they can create a very personalized offering for a consumer. “Hey, it looks like you have this kind of debt. It looks like you have enough equity in your home that if you were to take out $62,000 in cash, you could pay off some of this debt, and financially you’ll be so much better off.” I am much more likely to go to such a lender.

SW: What are your thoughts on safety?

MG: I think safety is a very big and serious topic. There have always been security risks, and there will always be new security risks: it’s an arms race. AI has allowed us to address security in ways that were not possible before, by helping us identify patterns of security threats that we may not have modeled. If you have to build a predictive model, it has to be able to predict certain things, which means you assume certain things. But it is very difficult to accept new security risks that will arise in a year’s time. Like no one knows, but with gen AI you don’t have to know everything. It can identify new patterns on its own without you having to tell it.

So that has made it a really powerful tool and an ally to be able to identify and address new threats, but it has also brought new security threats. For example, there is a new type of security threat called fast injectionwhere you can input malicious cues and make AI do things it shouldn’t do, and give you answers it shouldn’t respond to.

Other things we see with generative AI is that the output of the AI ​​is not always something that you can accurately predict because that is the nature of it. It generates brand new content that has never existed before, so you can’t really predict what it’s going to generate. So how do you test if it’s secure and compliant? We’ve looked at a lot of new technologies around this.

For example, something called generative networks and discriminative networks, where one AI model tests the work of another AI model based on probabilities – as if this kind of thing is becoming a reality. So even the way you build and test new applications will completely change.

And that’s the whole point of it generative adversarial networkor GAN, which is essentially a network where AI models test each other’s work. And there’s a whole framework for that, because we have to do that in a methodical way and not just randomly. So we really need to be at the forefront to make sure we’re ahead of what’s happening in the industry today. This is what it means to make AI applications enterprise-ready. It’s not just about building sexy new interfaces and awesome demos, but it really digs deep into what goes into building compliance and security and making it safe.

SW: What keeps you awake at night?

It’s part excitement, part fear that keeps me awake at night. And as exciting as it is, like you really have to be paranoid about certain things. I’m really excited that AI is finally starting to take off. That’s really exciting, but the pace of innovation is also very, very fast and accelerating like we’ve never seen before. We measure something known as AI years, where a week or two of AI is like a human year compressed into a few weeks.

But while all that’s happening, companies are going to have to run very, very fast to stay in the same place, and those who innovate will far outpace those who won’t be able to innovate. We’ve seen that with common technology, but it will become even more apparent now.

SW: How do you build a technology team that can handle the scale and pace of AI innovation?

MG: I think our team is one of our key differentiators. Our core team is made up of highly specialized engineers who can build mission-critical fintech applications that allow us to move hundreds of millions of dollars and reconcile and account for all of that, and that’s a huge undertaking that we do all day, every day. Highly specialized types of engineers are needed to work on such business-critical applications. It’s mainly our developers, security and compliance: people who are very skilled in the cloud and build things that are cloud-native data platform APIs.

And then we have a dedicated AI division where we continue to continually evaluate our core strengths. As the world of AI changes, we need to reshape our team and bring in expertise where necessary. We’ve moved very quickly from what we now call traditional AI to what we do with LLMs in generative AI, and the kind of expertise I need from the team is very different.

We have to think a lot about the end-user experience, because what does an end-user experience actually mean in this case? It can’t just be a conversational interface, because a conversational interface is like a room with infinite doors, like you can go from one place to another – but you also have to limit it. So how do you combine a conversational interface with a traditional point-and-click application so that you provide enough flexibility, but also provide structure for consumers to be able to use your application and be productive. We have highly specialized design and development teams who are constantly thinking about these issues and testing them in the market, in addition to our core engineers who are highly skilled in LLMs.