Are you ready to bring more awareness to your brand? Consider becoming a sponsor for The AI Impact Tour. Learn more about the opportunities here.

Businesses have countless documents to keep track of. And organizing, maintaining and updating them can be a challenge, both from a volume and a manpower standpoint. 

Intelligent document processing (IDP) tools are increasingly applied to process and analyze structured and semi-structured documents such as forms and invoices. This technology leverages such tools as computer vision, machine learning, natural language processing, and optical character recognition. 

But when it comes to delving deeper into more complex documents and assessing and parsing their contents for consistency, clarity and errors, these systems can often be simplistic in nature.

Contract intelligence (CI) picks up where IDP leaves off. Using a branch of ML known as natural language understanding, CI provides meaningful insights from extracted data in more complicated contracts, agreements, and other business documents. 

VB Event

The AI Impact Tour

Connect with the enterprise AI community at VentureBeat’s AI Impact Tour coming to a city near you!


Learn More

“The goal is to get computers to use language like humans use language,” said Francisco Webber, cofounder and CEO of NLU company

According to KBV Research, the IDP market will reach $4.1 billion by 2027, representing a compound annual growth rate of more than 29%. The CI market is expected to grow exponentially along with that, and its top players include, Conga, HG Insights, Evisort and IntelAgree., Webber contends, has the unique tools to be able to take the industry to the next level. The differentiator is in the company’s semantic folding technique. This technology converts a piece of text, whatever that may be and in whatever language, into a bitmap — essentially a “semantic fingerprint” of the text. 

“It precisely converts text into a specific representation that makes it very easy to apply machine learning on top of it,” Webber said. 

Models built out from the technology can then be given examples of various documents. Based on those examples, they can search, extract and classify important information, even when this requires interpreting, or if the phraseology varies. Models can be used for any document-based standard, including contracts, lease agreements, quotes or commercial offers from insurance companies. has applied this technique, which is inspired by neuroscience, to its two newest products. Contract Intelligence 4.6 automatically searches, extracts, classifies and compares key information in documents such as contracts, insurance policies, financial reports and requests for proposals. Message Intelligence 2.4 automatically classifies and processes messages, attachments and unstructured text based on meaning. When used with Contract Intelligence, it can also extract and leverage buried information, Webber explained.’s technology can calculate similarities between texts effortlessly. With a termination clause, for example, models use information from previously provided examples to process what terminology and phraseology is used, and the context around that. It can then examine another clause and predict if it matches. 

Models could also be applied to search and classification, Webber said. For example, after processing previous examples of email or social media complaints or other feedback coming in at high frequency, models can create specific filters so that these messages are immediately forwarded along to the correct department.’s models are easy to train, tune and apply across different languages, he explained. They can achieve accuracy with as few as 20 examples of a certain type of clause, contract or other business document. And the process does not require a data scientist or AI expertise. It is wholly between the subject-matter expert and the automation model. 

“It needs so few examples that you could literally ask a subject-matter expert ‘Show me a handful of things you are spotting for in your contracts,’” and the system will pick it up, Webber explained. “We want to put the subject-matter experts in the driver’s seat. That’s where the future is. Models will be modeled around humans.”

This technique allows users to train their document extraction models and maintain their IP. From there, they can start a new type of document model without having to go back and build out a whole new system. The tool can adapt as an organization grows and evolves, allowing “much more strategic digitalization movement in a company,” Webber said. 

Models can analyze relevant information from a document quickly and accurately and achieve at scale what is otherwise difficult for other contract analysis tools, he added. This helps to increase efficiencies and benefits companies by saving time, money and manpower and reducing turnaround time. 

“There is an incredible amount of gain in terms of speed and quality,” he said. uses it too, he added — for example, to annotate training data for classified models.

Ultimately, there are significant implications for CI, he said. AI models are much less expensive than big language machines and can continue to be scaled with fewer computational obstacles, thus allowing them to quickly evolve. 

“There has to be a new orientation, specifically on the business AI side,” Webber said. “The main criterion for evolution is efficiency. In nature, if a creature isn’t efficient, it won’t survive.”

The industry had to wait for machines to become strong enough to be able to tackle such challenges, he added. It is “computer intense,” and the focus to this point has been on algorithms. 

But, he noted, “it’s not so much an algorithmic problem as it is a representational problem.”

A native of Vienna with a background in bioscience, Webber has always been interested in the link between how the brain works with language and what we then do with that information, he said. In beginning to tackle CI, he was inspired by the book On Intelligence: How a New Understanding of the Brain Will Lead to the Creation of Truly Intelligent Machines by Jeff Hawkins and Sandra Blakeslee. 

According to Webber, the company is getting closer to finding that answer. “We are a company working in efficient AI,” he said. “Now we have taken that technology to provide real-world tools. We have developed a highly efficient method of representing language.”

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.