Layoffs continue to plague the technology world, but with the need for tech talent only increasing in organizations, more attention is being paid to how internal talent is managed.
A startup in Belgium called TechWolf is taking a unique approach to meet this need. It has built an AI engine that collects data from internal workflows to learn about the people doing the work. This is then turned into data for managers and internal recruiters to more accurately gauge the interests and skills of different employees, help match them with different projects, and ultimately train them better. And much more can be provided.
The company is making some waves with its technology, boasting an impressive list of customers that includes GSK, HSBC, Booking.com and many others. And now he's raised about $43 million ($42.75 million, to be more precise) to expand his business.
London-based Felix Capital is leading the Series B, while SAP, ServiceNow and Workday — three titans in HR — are investing together for the first time. Other backers include Acadian Ventures, Fortino Capital Partners, Notion Capital, SemperVirens and 20VC along with DeepMind and unnamed “AI leaders” from Meta. From what we understand, the startup is now valued at around $150 million.
CEO Andreas De Neve, who co-founded TechWolf with Jeroen Van Hautte and Mikaël Wornoo, started the company in 2018 while the trio were still computer science students at the University of Ghent in Belgium and Cambridge in England. .
The original plan was to develop an HR platform — with which the startup builds its own language model “like ChatGPT,” he said — to help source and hire talent from outside.
“It failed,” he said simply. Recruiting, or at least the part of it they were trying to focus on, wasn't so broken. Employers “didn't need AI to filter the good applicants from the bad.”
But the founders discovered that their target customers had a different problem that needed to be fixed.
“They said: 'Hey, so this AI model, is there any chance we could use it on our 40,000 employees instead of our applicants? Because there might be people we could hire internally, De Nieuw said. “HR leaders pointed us to the right problem to solve: identifying employee skills.”
The question “What do you actually do?” The TV show “Friends” had a recurring joke about Chandler (an IT worker). But this becomes a big problem in real-world businesses, and it gets worse the bigger the organization gets. “You can have 100,000 employees who are all highly qualified, who all spend a lot of time in software systems that generate data,” De Nieuw said. “But structurally, these companies know very little about these people. So that's what we've started to do.”
This is just one type of problem that AI can solve, he said. “We started building language models that integrate with the systems people use for work: project trackers, documentation systems for developers, research repositories for researchers. And from all that data, We estimate what skills these workers have. You can almost think of it as a set of AI models that connect to the digital exhaust of an organization.
TechWolf is currently touching on a few key trends in the market that are worth noting:
- The real innovators dilemma? The seminal book, “The Innovator's Dilemma,” paints a persuasive picture of how even the most successful, large companies can be undone by smaller businesses that are too slow to respond to change. Grow up. But looking at it differently, the primary asset that helps one organization operate more flexibly than another is its people: how easily teams can be formed around different projects and goals. What makes or breaks these efforts? And it turns out that organizations are willing to pay good money for tech that can help them do this.
- LLM vs MLM vs SLM. “Big” language models and the companies that build them continue to generate a lot of interest. And “generate” is really the operative word here, as they're what underpin buzzy generative AI applications like ChatGPT, Stable Diffusion, Claude, Suno and more. But there is certainly a growing tide for “smaller” language models that can be applied to very specific use cases, which are potentially less complex to build and run, and ultimately more compelling and Thus there is less risk of fraud. TechWolf isn't the only company operating in this space, nor the only one attracting investor attention. (Another example is startup Poolside, which is also building AI for a specific use case: developers and their coding tasks.)
- Focus really matters a lot. I asked De Neve if TechWolf has any ambitions to expand its platform into other areas such as enterprise search or business intelligence. After all, it's already consuming a lot of enterprise information, wouldn't it be an easy step to build more products around it?
nonDe Nieuw had a clear answer: “We can process data like no one else in the market, but we're very focused on solving the expertise problem, because at the moment, in the market where we work We do, there's a lot of demand for us.”
At a time when it feels like there's a lot of noise in the AI world, the focus rings an obvious bell and may be one reason why investors are interested in companies like this.
Julien Codorneau, the partner at Felix who led the deal, believes TechWolf can outpace much larger companies coming from other corners, such as AI-based enterprise search. “A job well done can really pay off,” he said. “They don't want to be Workday or ServiceNow. They want to be the Switzerland of the HR department.