Document transformation startup Sensible Technologies Inc. today became the latest company to embrace the phenomenon that is generative artificial intelligence, as it announced the availability of a new tool powered by OpenAI LP’s GPT-4 that makes it simple to extract information from unstructured documents.

With Sensible Instruct, it becomes possible to use natural language to extract data instantly from any kind of document, including resumes, invoices, contracts, academic research papers, financial statements and reports, utility bills and almost anything else.

The tool, which is initially being made available for free, applies the capabilities of the most advanced large language models to the very old problem of parsing documents to obtain information as quickly as possible. Unstructured data search is a difficult challenge, but recent developments in AI have made it simple to solve. Sensible Instruct allows any user to ask in natural language exactly what they need from any given document and obtain instant results.

According to Sensible, the tool is unique in that rather than using the “generative” aspect of GTP-4, it relies on its strengths in pattern recognition and data parsing to extract information from documents like PDFs.

Examples of how it can be used include uploading an invoice, or bunch of invoices, to quickly obtain the due dates. Alternatively, it could be used to identify the liability limit of an insurance policy, account numbers from bank statements, addresses from passports and more.

It can also be used to extract repeating data elements from a document, such as the work experience or skills from a resume, a patient’s allergies from their medical records, or the line items in a purchase order. Finally, it can extract data from tables, such as the list of transactions in a credit card statement or the balance sheet of a 10-Q regulatory filing.

The capabilities go beyond natural language questions. Because it’s built using Sensible’s own query language SenseML, developers can deploy its parsers as application programming interfaces to extract information from tens of thousands of documents per hour, the company said.

Constellation Research Inc. analyst Holger Mueller told SiliconANGLE Sensible Instruct is a common application of generative AI that has real potential to make life easier for enterprises. “The availability and affordability of more capable LLMs like GPT-4 creates new opportunities for software vendors, and Sensible Technologies is taking advantage of that,” the analyst said. “It uses GPT-4 to extract the most relevant content from what can often be very long-winded documents, helping uses to find the information they need much more quickly. The result will be higher productivity and enterprise acceleration.”

Looking ahead, Sensible said it hopes to use the feedback from early adopters to improve the tool’s capabilities further and apply it to newer use cases. It also has plans to integrate with more advanced LLMs in future, as they become available.

Image: Sensible Technologies