Finding a buyer for their technology

Given our in-house expertise, in particular Shaun’s track record of building these types of systems with BAE, we’ve been confident in our ability to make this technology work. 

As with many deep tech companies, however, it’s not just a case of making the technology work. We had to ask ourselves—is there really a product there? Will people actually pay for it? 

Given my background in financial crime, we set out to serve financial services: a huge trillion dollar market that, even if you just make a slight improvement, offers massive upside. But we quickly realised how difficult these companies were to sell to. They have long, rigid procurement processes and fairly mature KYC technology, meaning they wouldn’t provide a great path to revenue. 

As a start up we had to earn revenue quickly and so we decided to start out “uncomfortably narrow” where we knew we could find product-market fit and close deals within a few months. 

So our attention shifted to universities. Speaking to my alma mater, the University of Cambridge, I realised these institutions expended vast resources trying to understand who they were receiving donations from. If they were writing someone’s name on the side of a library, it’s worth spending the time to vet them, right? But the process was manual, sitting across web browsers, corporate records, sanction lists, screening tools, and so on. 

Universities proved to be our ideal first customer—strong demand, faster sales cycles, clear target markets. And most importantly, this gave us a roster of globally renowned organisations to stick on our sales decks and websites, adding much needed validation to our product. 

Once we scratched the surface of this market, we glimpsed huge potential. It wasn’t just universities receiving these types of philanthropic donations—it was museums, orchestras, charities, and countless others. We spoke to everyone from the Museum of London to the Royal Shakespeare Company to St Paul’s Cathedral.

For each customer, a specific context that our AI was adept at accommodating. One cancer charity we worked with revealed one specific blindspot—around receiving donations from tobacco companies. We realised building in these nuances was key to differentiating our product from the crowd. 

"Dynamic due diligence" through AI

We’ve certainly reaped some of the benefits of being early in building with LLMs: it’s enabled us to build ahead of the competition and capture attention from the industry. And to be honest, I often think to myself—if I wasn’t already doing this, then I would start a company today around the application of AI to a specific industry problem. 

My advice for other founders building in or entering the AI market? 

Compartmentalise your work. We’ve gleaned huge advantages from keeping our core product development separate from exploratory R&D. To illustrate—we have two teams, an engineering team whose job is to build a specific piece of the product, and a small research team whose job is to go out and speak to experts, experiment with the new models, and truly find out what is possible. The interaction of these two teams ensures that we are constantly delivering while also evolving to push the boundaries for our industry. 

Our next phase is promising exactly this—exploring the concept of dynamic due diligence. As it is, due diligence is static, in that it captures a single point of time. Our vision is to create a category defining product for continuous monitoring, leveraging the power of LLMs to ensure that due diligence is ongoing, constant, and up-to-date. The focus is on ‘knowledge changes’ rather than just news mentions, helping us target regulated industries that require periodic KYC reviews (that typically happen annually or biannually).

The pace of change in AI can often be overwhelming. New models are launched every week, new unicorns minted every month. The focus, as with any company, comes down to something simple. Are you doing something that is really worthwhile, or is it just something that someone else can come along and build, whether that’s another startup or the companies that you’re selling to?

As Xapien scales, and we aspire to build a category defining product, this is something that remains front of mind.