Why Finance Industry Leaders are Investing in AI: An Interview with AI Consultant Francesco Corea

Article by Daniel Smith | October 26, 2018

Complexity scientist, tech investor and IPAM fellow Francesco Corea has spent his career so far consulting for financial institutions large and small. He currently uses his research experience in machine learning and PhD in economics to advise AI startups on the effective implementation of their solutions. You can find Francesco’s thoughts about AI in the financial sector on sites such as Forbes and Medium.

In this second interview in our series focusing on AI specialists, Francesco chatted with us about a diverse range of topics, from the ways AI is transforming economics to how quality data sets industry leaders apart.

Lionbridge AI: What sparked your interest in machine learning?

Francesco: As a complexity scientist, I am a strong supporter of an interdisciplinary research approach, and my life goal is to foster the interaction of different sciences in order to bring to light hidden connections. I am passionate about exploiting emerging technologies to solve high-value problems with impactful solutions.

As a result, I don’t think my interests were particularly focused on machine learning. I just found myself working with those tools! I wanted to make sense of phenomena I was observing for which my economic/financial models didn’t have an answer. Reading the data was the only way I could find patterns in messy observations, and this is how I ended up learning everything from logistic regression to neural nets. It was a necessity rather than a conscious choice.

L: Which issues in machine learning are you currently focused on?

F: There are three areas in my daily work where ML is involved: first, I see and screen a lot of machine learning companies, so of course I need to have an overview of recent developments to understand the solutions a specific team is building. Second, I help companies to implement AI solutions and make sense of the AI hype that we have – this is a bit more strategic and operational.

Finally, from a research point of view, I am focusing my efforts on understanding how I can exploit the tools I have to make the investment process more efficient.


“The venture industry is betting on AI as a disruptor for a variety of industries, but hasn’t really recognized how it will change the venture industry itself.”


Very few venture capital firms are using AI to improve their own operations. As a result, I’m interested in how analytics can be used to efficiently search for investment opportunities, improve your cherry-picking abilities as an investor, and support your portfolio companies later on.

L: You’ve written frequently about AI in the financial sector, from your Medium articles about AI and venture capital to your recent work for Forbes on Blockchain and the Internet of Things. What is it about AI that allows it to have such a massive impact in this particular industry?

F: We could talk about this for hours, but basically AI is what is called a General Purpose Technology. It is one of those technologies that not only enables new layers of tech to be conceived and built, but also completely reshapes the economic and social paradigm.

The financial sector is something that is on one hand extremely old-fashioned (technologically “antique”) and intrinsically related to our social constructs, and from the other a very competitive and fierce industry. It is a sector where you need to “innovate-to-survive” rather than “to-grow” and therefore you need to always look for solutions that give you an edge. After having been a human-intensive business and a capital-demanding one, the next step is of course leveraging technology to the extreme.

Then there are practical considerations that make the financial industry a good candidate for AI. In general, it’s a data-rich and information-poor industry, so companies have a lot of data to work with. Also, big banks and institutions are highly structured, which means they have a ton of processes that can be accelerated, automated and improved. Finally, it is the sector at the center of the cross-generational transfer of wealth. AI needs new data, feedback and to learn from its own use, and millennials and new generations are more prone to feed an AI engine than previous generations.

L: How do you see corporate investment in AI benefiting those companies willing to take the risk?

F: I think right now it’s unquestionable that the first-mover advantage is a real thing in applying AI. All companies should seriously consider this and start understanding where and how they can embed learning systems in their own workflows.


“There are many things we don’t know yet, like which precise AI approach will pay out in the future, but companies willing to take the risk and invest today will lead the market tomorrow.”


We should also mention that they will see immediate benefits: a study from a couple of MIT professors shows that a company leveraging analytics, in general, outperforms its peers by around 5%-6%.

L: It must be absolutely crucial for these companies to have high-quality datasets. What are the most important features of data for those looking to build an algorithm for use in finance?

F: The finance industry is huge and every problem is different. Although some features are the same whether you are dealing with fraud detection or high-frequency trading, everything has to be analyzed case by case. If you are building a trading algorithm, you want to be sure you have a continuous stream of real-time data on prices, volatilities, etc. to feed your systems. If you are building a personalized product for end-customers (such as credit scoring, robo-advisors, etc.) you want to be sure the data are accurate and capturing aspects of the person that are statistically significant.

In venture capital, the biggest problems are data sparsity (you don’t have all the data points for every company), variables definition (it’s not easy to understand what “success” means, for example), and the difficulty in forecasting/backtesting a model (especially given the long cycle of the investments).

L: In your experience, what role does data quality play in the development of great machine learning algorithms?

F: Data quality is hugely important: it can determine the success or failure of a model. This is especially true now that computation is (more or less) affordable and there are a variety of open-source models and frameworks. The type and quality of data are the real competitive advantages for a company today. This is destined to change in the next 5-10 years, but nowadays makes the difference between leading and following in the industry.

There are different interesting projects and companies I am constantly looking at such as Premise data, Streetbees, Dataloop.ai, Neuromation, and OpenMined, but also marketplaces such as BurstIQ, doc.ai, Datum, Streamr, Fetch.ai and others which I believe represent the next stage of AI development.

L: Looking forward, which hot topics do you expect to be writing a lot about in the coming year?

F: The list is long. I will definitely keep writing about AI in venture capital, and then I am looking at a couple of different verticals like healthcare and energy. I am also very interested in studying the convergence between AI and quantum computing as well as the hardware aspects of the AI wave.

L: Are there any industry applications of AI that you’re particularly excited about?

F: I am excited about deep tech in general and meaningful applications, particularly within healthcare, energy and material science. I am also very excited by enabling stack in general. An example is a company I am advising called Meeshkan, which is building a framework to simplify and optimize training. They are basically building a new way to do automated training and tuning for multiple models and to optimize the GPU resources, which means a huge cost-saving and potentially a new spectrum of accessible applications for a company.

L: Finally, do you have any advice for anyone thinking about integrating AI with their business model?

F: There isn’t a correct answer here because every reality is different. In general, I recommend to start small but with commitment. Test and experiment as much as possible, then increase your investment and effort level once you’ve achieved your initial milestones.

The Author
Daniel Smith

Daniel writes a variety of content for Lionbridge’s website as part of the marketing team. Born and raised in the UK, he first came to Japan by chance in 2013 and is continually surprised that no one has thrown him out yet. Outside of Lionbridge, he loves to travel, take photos and listen to music that his neighbors really, really hate.


    Sign up to our newsletter for fresh developments from the world of training data. Lionbridge brings you interviews with industry experts, dataset collections and more.