Moveworks Inc., a startup using artificial intelligence to help enterprises automate information technology support tasks, is worth $2.1 billion following a new $200 million funding round it announced this morning.

Mountain View, California-based Moveworks said that the round was jointly led by Tiger Global and Alkeon Capital. Bain Capital, ICONIQ Growth, Kleiner Perkins and a number of other institutional backers participated as well. Palo Alto Networks Chief Executive Nikesh Arora is joining Moveworks’ board as an independent director following the investment. 

Moveworks provides an AI-powered chatbot that can automatically answer technical support requests from a company’s employees. The chatbot is capable of providing explainers on how to set up a printer or install a piece of software. It can also perform much more complex tasks, such as resetting a password or creating a Salesforce account for a newly hired member of the sales team.

Such tasks are normally performed by companies’ information technology departments, which often take hours or days to respond to a request. The reason for the delay is that administrators can have dozens of outstanding technical support tickets at any given time, meaning employees must wait until the earlier requests are resolved. By partly automating the process, Moveworks says that its platform can not only speed up troubleshooting but also free up administrators to focus on other tasks.

Under the hood, Moveworks’ chatbot processes support requests using an array of machine learning algorithms the startup collectively refers to as its Intelligence Engine. One of the machine learning algorithms is an implementation of BERT, a neural network originally developed by Google. Moveworks’ implementation of the neural network turns natural-language user requests such as “how to fix the printer” into a mathematical representation that its chatbot can understand.

BERT, which powers parts of Google’s search engine, was seen as a fairly significant advance when the search giant shared it with the research community in 2018. That’s because the AI introduced a new feature, a so-called attention mechanism, that significantly improves the accuracy of processing results.

The attention mechanism allows BERT to infer not only the meaning of a word but also the context in which it’s used by analyzing the broader text. This context makes it easier for the AI to correctly interpret words that can have multiple meanings depending on how they’re used.

Deploying BERT in production requires fairly large amounts of training data. In the case of Moveworks, which is using BERT to analyze employee support tickets, the training data used by the startup consists of more than 100 million support requests that its chatbot has processed for enterprise customers.

Moveworks had to overcome a fairly significant technical challenge to build its chatbot’s AI layer.

Customer support tickets can vary greatly in their contents, which makes AI training difficult. For example, if an AI is trained to answer requests containing the question “how to open Chrome on Windows,” it may not be capable of answering the related question of “how to open Firefox on Windows,” though the process is practically identical. Moveworks’ engineers solved the challenge by turning sets of similar support requests into a generalized form that allows the chatbot to answer different versions of the same question.

“Everything we do at Moveworks is inspired by a simple idea: it shouldn’t take days to get help at work,” said Moveworks CEO Bhavin Shah. “But to make that idea a reality, we’ve had to build one of the most complex machine learning systems in the history of the enterprise.”

Moveworks says that its chatbot can fix 40% of the support requests received by a typical company’s IT department. The number of active users who rely on the chatbot increased by 700% in the past 15 months, the startup says, while the number of issues it has resolved increased fivefold. Moveworks’ customers include major enterprise tech firms such as Broadcom Inc., Nutanix Inc. and Palo Alto Networks, among others.

The round comes a few months after Moveworks expanded its platform’s focus from IT support requests to inquiries received by other departments such as finance and human resources. Now that its chatbot can perform more tasks, the startup can compete for more types of deals, which increases its revenue growth potential. This increase in the startup’s revenue growth prospects no doubt factored into its hefty $2.1 billion valuation.

Another factor behind the steep valuation may be the fact that there are certain parallels between Moveworks’ approach and robotic process automation. Similarly to Moveworks’ software, RPA platforms use AI to automate relatively simple but repetitive manual tasks.

The RPA market is experiencing rapid growth thanks to strong demand from enterprises looking to automate the repetitive aspects of their employees’ work. Moveworks’ investors may believe that its chatbot, which has the same core focus of automating routine chores, may allow the startup to catch the enterprise spending wave that is propelling the RPA market’s growth.

Moveworks has raised $315 million in funding to date. 

Image: Moveworks