Verbit is a very successful startup. The 4-year-old developer of an AI-powered transcription and captioning platform has reached unicorn status in June, raising $157 million at a valuation of over $1 billion, for a total of $319 million raised to date. It has 2,600 customers, 450 employees, and will reach $100 million in revenues by the end of the year. According to co-founder and CEO Tom Livne, Verbit enjoys Net Revenue Retention (the rate of revenue generation from existing customers) of 163%. “Our customers are growing with us,” says Livne.
This impressive performance is the result of executing on a well thought-out framework for what it will take to succeed in the future, no matter what business you are in and the market you are serving. Verbit’s technology foundation, its global community of freelancers, and its mass customization strategy are the three features of Verbit’s future of work model, the very model of a 21st century company.
Verbit entered a very fragmented market for transcription and captioning services, typically performed by listening to the audio and manually typing out the words. It also entered the market at the time when the field of Automatic Speech Recognition (ASR) has greatly benefited from the rapid progress since 2010 in AI capabilities, specifically deep learning techniques, which led to a significant reduction in error rates. The success of speech-related applications for mobile devices and the proliferation of smart speakers (e.g., Amazon Alexa, Google Assistant) also helped drive the development of open-source ASR tools.
The new world of work is also about a new kind of teamwork: humans and AI working together to achieve more than they can accomplish on their own. Regardless of its recent progress, AI is still not accurate enough to meet the enterprise-level requirements of speech-to-text in many industries. “If technology gives me 90% accuracy, humans can deal with the last mile. Human-in-the-loop is core to our product,” explains Livne. In addition to developing the required technology, Verbit has developed a global community of trained, free-lance transcribers. To guarantee the customer 100% accuracy, Verbit today employs 35,000 freelance transcribers around the world that work on the “last mile” editing of the text produced by its algorithms.
This flexible, work-from-everywhere new world of work, requires new management skills. “We have lots of IP around building and managing this community of freelancers—how you hire freelancers, training and onboarding, how you do the ranking system, compensation, quality control,” Livne lists all the qualifications required to manage and grow the new version of a global workforce.
The third feature of the three-dimensional (Meta?) new world of work is serving each customer as if it is the only customer. Verbit’s domain-focused go-to-market strategy (e.g., education, legal, media) helps in training both AI and humans in the specific language of specific domains. But it also helps in “understanding who you are selling to, how long is the sales cycle, what are the specific requirements and needs, customizing the solution,” says Livne. “We utilize all relevant data for the specific customer, we utilize all the data we accumulate for the customer, we create a model just for them,” he adds.
In short, Verbit demonstrates that the future of work gives a new meaning to Silicon Valley’s passion for doing everything “at scale”: develop your technology on the basis of other people’s (mostly free) talents; build a global, flexible, (relatively) easily expandable workforce; and rapidly grow your revenues by customizing your service for the specific needs of each and every customer.