Fascination with AI shows no sign of abating in Europe. While established companies may still be slow to pick the technology up, new startups in the sector continue to grow–in France by as much as 38% this year already. There is also no shortage of investment, with $4 billion pumped into the sector globally in Q3 compared to $2.8 billion in the same period last year, while the U.K. continues to set the research and development standard in Europe with 623 AI-related patents.

Despite the ongoing demand, AI startups aren’t safe from the statistics looming over every new business venture: there’s a 90% failure rate. It’s often said numbers like this are not intended to discourage entrepreneurs, but to motivate them to work smarter and harder. Unfortunately, it’s more than just hard work that sets a company up for success – and for those who are looking to scale up their AI venture, the best lessons to be learned can be drawn from those who’ve been in the same position – and thrived.

Looking at a selection of ten of the most successful AI companies of today–UiPath, Graphcore, DarkTrace, Benevolent AI, Behavox, Roborace, Babylon, Featurespace, Onfido, and Signal AI–there are common attributes and stand-out stories that led them to secure the funding, customers, and market share they needed to scale.

Both the sector and market a new start-up targets says a lot about its potential for success. Targeting high growth sectors can be a competitive undertaking, but also one of the best ways to succeed–as is diversification.

This is one of the reasons why companies such as Darktrace or Onfido, which offer AI solutions for security and compliance issues, are doing so well right now. Cybersecurity has become an increasing concern for companies in every sector, intelligent algorithms that either speed up compliances processes or make it easier for IT teams to spot external threats are in high demand. The introduction of GPDR last year means it’s financially risky to not invest in robust security systems – British Airways’ landmark £183 million fine from the Information Commissioners’ Office serves as a serious warning.

Other notable growth areas include process automation and processor acceleration – shown by the meteoric rise of UiPath and Graphcore. UiPath provides RPA and AI enhanced tools to businesses for building software to emulate processes. At the other end of the eco-system, Graphcore builds hardware to enable faster and more complex AI models to be run. Next year we’re likely to see more investment poured into both intelligent RPA and the growing need for AI hardware, ultimately enabling computer processors to catch up with the software being developed.

The increase in investment and roll-out of AI in the healthcare sector means there’s been a surge in valuations and gains for companies including Babylon, Benevolent AI and Behavox. As the sector grows, it will also likely incorporate more consumer-focused markets including “wellbeing and nutrition,” so there will be a growing number of opportunities for start-ups in this broader domain.

Start raising for expansion–but stop for development

VC investment is a proven way to scale up a company. From studying the selection of companies we can see a popular time to begin funding for expansion takes place around two years into product development and proof of concept testing, which is when most companies have generated enough momentum and refined the product market fit to begin scaling at pace.

In the group studied, the average time between Series A and Series B rounds is just over a year: 13 months. This steadily rises for funding rounds thereafter, to 16 months of time between Series B and C on average and 18 months between Series C and D.

These averages shouldn’t be the only guiding factor in when to seek further investment; the right time to wait in between rounds is dependent on several factors and differs from company to company. For example, when DarkTrace was scaling for global expansion in 2015, it completed its Series B funding just four short months after the Series A, raising a similar amount.

It’s essential for companies to allow space to deliver on the investment made. If a company is scaling and can demonstrate clear future growth potential, it is likely to attract further interest from new and existing investors, but it is important to analyse whether the extra capital is needed at that specific point in time, or if it can be raised at a higher valuation further down the road. Babylon AI waited 16 months between its Series A and B, but then took a break to develop its proposition in Europe before completing a momentous Series C of £550 million to expand to the U.S. and Asia–more than 28 months later.

The phenomenon of “mega-rounds” is growing, but as seen with WeWork, they are not necessarily a recipe for success. The companies studied who have managed to build sustainable long-term growth can generally be grouped into two camps: either raising smaller rounds but more often, or seeking larger raises less frequently and seemingly from a more focused pool of investors.

Onfido’s funding journey has now reached Series C, and notably from a wide range of investors at each stage. This is interesting to compare to Behavox, which has only raised from a small number of investors including Hoxton Ventures, Citi, Index, and Promus throughout its rounds to date. Both are valid funding strategies–but getting the balance right and recognising what is needed for their scale-up is the challenge for AI founders and investors alike.

Leadership matters

Attributes of successful AI companies’ CEOs are far from homogenous. While most start-ups thrive when their founder is also their CEO for a long time into the company’s lifecycle, this is not the only way forward. An industry leader in cyber security, DarkTrace has an innovative leadership model in Nicole Eagan and Poppy Gustafsson sharing the role of CEO. Meanwhile, Roborace’s Denis Sverdlov asked Lucas di Grassi to step up to CEO so his background as a race driver and businessman could be used to its full potential.

Most founders are understandably reluctant to let someone else look after the company they’ve built. However, in many cases, it is the founder’s business expertise and leadership style that has led to a company’s success. This is more important than the founder’s background in the sector. Serial founders, such as Graphcore’s Nigel Toon and Babylon’s Ali Parsa, use their managerial experience running other VC-backed companies to secure funding and scale their ventures.

Looking at the data, 60% of successful AI ventures are led by a founder-CEO, 20% are led by non-founder CEOs who were brought into the business because of their business development expertise, and another 20% are led by non-founders who were brought into the business because of their sector expertise. Both of these things are important–so one of the best recipes for success remains a strong CEO who has the expertise and stamina to do both.

Tying it all together

Ultimately, the most meaningful contributor to success is a combination of factors, and certainly not one set template. It is a mixture of product, market, management, and timing. An excellent illustration of this is Daniel Dines’ founding story for UiPath. Dines didn’t believe the automation was such a high-priority concern for businesses and was shocked to learn most do not operate as smoothly as a computer programme. He recognised that the problem with employees building business processes is that they always interact with the system through a user interface, which is readable only by a human.

Dines’ solution was to create robotic processes that replicate humans using the same tools, which allowed UiPath to scale RPA for any size of enterprise. It was this which set UiPath apart from its competitors, despite not being the first to launch in the space: a user-based, scalable solution that left its competitors behind. Evolving its solution to incorporate an intelligent edge and AI component has enabled continued market relevance and customer growth.

From a funding perspective, UiPath has recently raised a $568 million Series D at a valuation of $7 billion and is now the clear leader in the RPA space. Blue Prism, the company that arguably coined the term and put RPA on the map, has a market cap of only $1.07 billion in comparison. First mover advantage didn’t secure victory, it was the combination of factors that UiPath has played to its advantage which has enabled a meteoric rise. The next question is how long will it continue?