London and San Francisco-based startup Klu has today raised $1.7m to scale its platform that helps other companies get value out of generative AI models. The round was led by Firstminute Capital and also included scout investors from a16z, Sequoia Capital and Atomico.
What does Klu do?
Klu essentially helps other companies build useful products with AI tools like GPT-4, and is currently working with seven paying customers including hardware design platform CoLab and HR platform Zavvy.
Cofounder Stephen Walker tells Sifted he’s currently seeing the most interest from two main types of customer: SaaS companies and financial institutions like private equity firms.
He says one of the use cases for a private equity firm using Klu's product is lead scoring — and the analysis that typically happens before a PE firm decides to reach out to a company to potentially purchase it. That analysis has been moved over to Klu, and the startup provides an initial “first pass” score after being fed data on the company’s metrics. “That way, they spend more of their time actually making phone calls,” Walker says.
Klu has also helped an edtech company build a GenAI tutoring companion that allows students to ask questions that are then answered based on hours of recorded seminars from past cohorts. To make it work, Walker says that Klu has got three different models working together in a single, coherent product: one to turn the audio from the recordings into text, another to make the content easily searchable and a third to allow students to query it with a chat interface.
He says that, so far, Klu is seeing the most usage and traction from SaaS businesses, where engineers are accustomed to building new products. But he expects more traditional companies to start realising the potential of building AI-powered tools before too long.
“I do think more and more non-technical businesses will start adopting this once they realise that they can bring this into their work,” he says.
What’s the market like?
Klu is competing with a number of other companies working on building GenAI-powered products for businesses, including London-based Humanloop and Paris-based Dust.
Walker says that we’re still in the very early days of applying GenAI to products, and that the possibilities for use cases will broaden once multimodal models (ones that can understand different types of media like text and images) start being rolled out.
He adds that the issue of “hallucination” — where models like GPT-4 make up false responses to questions — is becoming a “last year problem”, as most of the clients he’s working with are now able able to quickly iron out issues with prompt engineering (the process of honing inputs into GenAI models).
Klu will use its fresh funding to increase the size of its team from four to eight, all of whom will be focused on engineering.