Nanonets lands $10 million to expedite document processing with AI

Organizations are devoting more resources to deploying intelligent document processing, particularly as they embark on digital transformations. As Deloitte explains, intelligent document processing automates the processing of data contained in documents — understanding what the document is about, what information it contains, extracting that information, and sending it to the right place. According to Everest Group, the global market for intelligent document processing was worth between $700 million and $750 million in 2020.

Leveraging a blend of AI, including computer vision and natural language processing, intelligent document processing can help to automate tasks like invoice processing, insurance claims, patient records, proof of delivery, and order forms. A range of startups offer these types of services, including Mindee, Zuva, Rossum, PandaDoc, and Anvil. Another player is Nanonets, which can extract data from documents and input them into databases automatically.

Nanonets today raised $10 million in series A funding led by Elevation Capital, bringing its total raised to over $11.5 million.

AI-powered document processing

Thirty-five-employee Nanonets is the brainchild of Prathamesh Juvatkar and CEO Sarthak Jain, who launched the company in 2017. Juvatkar previously founded Cubeit, a content aggregation startup acquired by ecommerce company Myntra in 2016. Jain was the CEO of Cubeit from 2012 to 2016 and served as a senior product manager at Myntra before joining Nanonets.

Jain estimates that 70% of organizations still have paper-dependent processes today, often involving legacy systems and processes that require employees to review each document and manually enter data into customer relationship management software, enterprise resource management software, or Excel files.

“[A]utomation induces a better way of connecting information across documents — quick implementation, execution, and arrangement of the large volume of data in less time than mechanical procedure with the reduced overhead expense and improved experience for enterprises,” Jain told VentureBeat via email. “Typically, creating a machine learning model that’s custom to [a] business needs requires many months, millions of data points, and the world’s best talent. We have taken that process and condensed it into a simple workflow of just uploading … data and automatically getting a best-in-class model spit out by the system in a matter of minutes.”

Nanonets can capture data from documents using AI, reading unseen, semi-structured data and improving as it captures additional documents. Jain claims that Nanonets significantly reduces the need to create rules and layouts per document format, a common problem with traditional optical character recognition providers.

“[W]e have created pretrained architectures … suited to enterprise applications [that] only require fine tuning of a few parameters for a new customer … Hundreds of thousands of machine learning models have been trained on Nanonets, and billions of images and documents have been processed,” Jain added. “A majority of AI models commercialized today are single modal, i.e., only text or image or structured data. We combine both image and text and are in the specialty area of multimodal deep learning models, which is especially useful in extracting data from complex data-dense tables.”

Nanonets supports platforms customers can directly import from or export to existing workflows. The platform also gives users the ability to write their own business rules, connect to different business data sources, and update internal systems to automate document workflows and financial controls.

A growing market

According to an estimate by Bain, automation platforms with broader capabilities and technologies have an addressable market of over $65 billion. Jain sees Nanonets in this category, going up against process automation vendor heavyweights in the enterprise. Demonstrating the interest in the field, a recent Deloitte report found that 73% of organizations worldwide are now using automation technologies — such as machine learning and natural language processing — up from 58% in 2019.

“When we met Sarthak and Prathamesh, we were impressed to see that despite being at the cutting edge of AI and machine learning, Nanonets’ product is no-code and super intuitive — which makes it extremely easy for their clients to adopt,” Elevation Capital’s Mukul Arora said in a statement. “We were also really inspired by Sarthak’s vision of using document extraction as a wedge to build a much deeper process automation platform.”

Nanonets claims, of course, that its platform is accurate. But issues can crop up in any intelligent document processing technology. For example, documents might be too poor in quality for the software to read, or there might be problems integrating the software with existing systems.

Be that as it may, Nanonets claims to have “hundreds” of customers in teams at P&G, Toyota, Sherwin Williams, Roche, and Divvy.

“Our annual recurring revenue is growing at 15% month-on-month and we are cashflow-positive. We have increased our customer base five times in the last 18 months and we have thousands of users coming to our platform daily,” Jain said. “This demand [has] resulted in an astonishing 12 times revenue growth in business for the company from April 2020 … Since 2020, [Nanonets] has helped 30,000 new users build workflows, developed over 100,000 AI workflows, and processed billions of files.”

The company plans to use the new capital to expand its engineering and AI and machine learning teams as well as grow its operations and go-to-market divisions in new geographies.