DAILOGUES: You’re both originally from Italy. What inspired you to found a startup in London?

Pietro Arina: I came to the city to continue my research job and to expand my CV. I completed my PhD in London. While still working in academia, I realized that I wanted to do something more in line with reality. At that time London was a perfect fit to create a startup because it is a great space to meet people, gather professional experience, and conduct business.

Davide Ferrari: I also came to London to pursue a PhD but, I always wanted to create my own business. I just had no idea what to do and when it would happen. As my PhD progressed, I realized that academia was not really the place where I felt home. I desired the freedom that you can have outside of academia. Meeting Pietro was the first step toward this freedom and I found London to be a thriving environment for building a startup. In Europe, it is probably the best place to start a business. The regulation is easy to navigate, and the bureaucracy is way more manageable compared to Italy. Then one thing led to another.

DAILOGUES: Cities that are successful in AI usually have a broad ecosystem comprising companies, research labs, academia, and governmental institutions. London appears to offer such an ecosystem. For example, your startup is backed by institutions such as University College London, King’s College and London Venture Capital Network. What has your experience been with London's AI ecosystem?

Pietro Arina: Let’s have another look at academia. I would point out that the British institutions always want you to study something useful for society. Potential products or new services are well received. This “utilitarian” culture is different from research in Italy which is more theoretical. It helps when you want to found a startup. Combine this mindset with all the talented students that arrive from all over the world in London, and you get a truly dynamic environment. With its proximity to the United States and Europe, the city is fertile ground for fresh ideas. Some people even argue that AI was born in London. On top of the excellent research, London offers a rich layer of startups and established businesses. However, the U.K. is not like the U.S. For example, in London it takes more time to acquire pre-seed funding, which can quickly reach $10 Million in the U.S. There are also fewer college dropouts and people prefer growing a business more organically instead of rushing to the next stage.

DAILOGUES: What has your experience been in attracting funding to grow your startup?

Davide Ferrari: Getting funding is the most challenging aspect of an early-stage startup. It is not so much about the quality of your technology or product, but about what investors consider valuable. This is what you need to find out when pitching your idea. Since the COVID years, the sentiment has also changed and there are fewer funds available now. We are lucky enough to be very close to closing our pre-seed round. It took us basically one year from start to finish: from the conception of our idea to the actual signing of term sheets.

DAILOGUES: Who would be an ideal investor?

Pietro Arina: An ideal investor is somebody who understands the vision of the people in front of them, and who also has a good grasp of the market. As we are first-time founders, we have encountered prejudice where people doubted our ability to carry out our vision. In that regard, I think that a good investor is also somebody who is open to new people. It should be a person who invests in you to thrive and not for the direct return. Things might be different for second-, or third-time founders.

Davide Ferrari: A good investor not only provides you with money, but with expertise, a network, and even help.

DAILOGUES: Let's talk a little more about your company. What does Lapis offer?

Pietro Arina: We define Lapis’ main product as an AI data concierge. The vision of Lapis is to use AI to empower professionals in the finance industry. For example, our platform can be tailored to the undisclosed and private data of an investment firm. It provides the firm’s employees with a secure platform to manipulate their information and data, such as financial reports, in an easy-to-edit and easy-to-work-with format. In this fashion, we help finance professionals avoid repetitive tasks and increase their productivity.

DAILOGUES: What do you mean by AI? What does your tech stack look like?

Davide Ferrari: Our assumption is that the user doesn’t know anything about AI and doesn’t want to know anything about AI. We want to provide a software platform that exactly embodies this assumption. It should work while we use the best AI tools under the hood. Of course, large language models are the latest toy in our arsenal. They are incredibly powerful for many things, and for the manipulation of text in particular, which is an important aspect for our customers. Another important part of our platform is a tool for table extraction from screenshots or PDFs.

DAILOGUES: Since you're also using large language models, how has your experience been with both proprietary and open-source models? Do you have any preferred ones?

Pietro Arina: We use many different models ranging from open-source to proprietary models. For some tasks, we work with a model from OpenAI to which we have access due to a sponsorship that we receive from Microsoft Startup Hub. In fact, the challenge is that there are so many new language models being published almost every other week that it is hard to keep up with the releases. On the other hand, we’ve experienced that some tasks can already be solved with existing models to our satisfaction. So we don’t need better models in all respects. Eventually, we want to be as independent as possible and hope to implement everything with open-source models in the future.

DAILOGUES: Some argue that AI-first companies should go beyond just automating repetitive tasks and explore entirely new opportunities enabled by AI. Where do you see the greatest potential for using AI to create new business opportunities?

Pietro Arina: Lapis offers a good place where a lot of information from public as well as private sources comes together. I believe that this place enables our customers to gain new insight into the data that is rarely considered together. One reason for that is that public data from the internet is often low quality and messy. So with our platform we help our customers to separate the important facts from the noise. Much of the financial data in the industry is of a different nature. It is often very condensed and accurate. However, a finance analyst usually needs only a few data points from this source for their work. In other words, there are two different data sources, which we combine efficiently with AI to get the best out of them for our customers.

DAILOGUES: Davide, you said that the users of Lapis’ data platform shouldn’t know what is under the hood. However, there are people in the field of Responsible AI who argue that people working with AI-driven tools should be familiar with what they are using because this will help them to make the best decisions, avoiding for example known pitfalls of current AI systems. Isn’t there a conflict between this understanding of AI usage and your approach making your platform so smooth that the users are unaware of the technology?

Davide Ferrari: It's not that we don’t want to involve the users per se or want to avoid telling them about the AI behind what we're doing – some of it is just protected intellectual property. But even beyond that, we have experienced that most of our potential users don’t want or need that level of detail. If you're technical and can contribute meaningfully, then this is a different story. But most of the people we serve are working 12-hour days. They don’t have time or interest in understanding what models we're using. They just want software that works well and fits their needs. In fact, giving them too much technical info can hurt the product's effectiveness. So, it's a tricky balance, but in the finance industry, more transparency on the tech side often isn’t helpful – it can be counterproductive.

We thank Pietro Arina and Davide Ferrari for the DAILOGUE.