Extract information from auto documents

For any company dealing with cars, extracting information from auto documents is error-prone and tedious — not to mention boring for whoever has to do it.

Though the world has gone digital, handling automotive documents is still painfully manual: scanning pages into a PDF, passing them to someone else’s desk, extracting information line by line. In some cases, one person has to look through the same type of document hundreds of times, typing long strings of cryptic numbers such as the vehicle identification number (VIN) into an ERP or centralized database.

When customers write in with common requests like making adjustments to documents — a change in ownership, for example, or a policy cancellation — this generates another stack of manual work.

We use artificial intelligence and automation to speed up every step of the process. This includes extracting relevant data from PDFs or images, then sending that data to the right place. In the end, this boosts accuracy and saves your company hours of valuable time.

Automation Hero can connect to many types of data sources, including email clients, an ERP, or a CRM like Salesforce. Our platform syncs up with your information, wherever it’s stored, automatically retrieving, say, scanned PDFs. Then an AI model specifically designed for auto documents — a car title, for example — aligns the scans and uses optical character recognition (OCR) to extract information like the make or model from relevant fields. Automation Hero trains its deep learning-based OCR models on custom domain knowledge, meaning it’s better equipped to extract detailed, relevant information.

Then the platform can perform hundreds of actions that aid in post-processing of data. These ultimately help boost the final accuracy. That information, along with image snippets, is passed to Robin, our personal assistant for attended automations. Robin serves up the information to agents, who can log into Robin then check the extracted information and do final cleansing where necessary.

Information from Robin then moves into a backend automation, where it is imported into an ERP or centralized database, via REST API or any other connector. For customer emails requesting changes to documents, we can use AI models to analyze incoming emails and discover the intent within them, and then create other models that
respond automatically or route the emails to the appropriate department.

Benefits include: Keeping humans in the loop (data passes through an automation assistant, Robin, for agents to review before upload), and a gamified interface: With Robin, agents view data in an interface that motivates them to review many documents in a short time. It’s also easy to view: Image snippets for all fields means humans can compare extracted values with little eye movement and without scanning a dense PDF. All this can save a tremendous amount of time: For organizations processing millions of documents each year, shaving precious minutes off of input and analysis saves hundreds of hours worth of work. The process is also more than 90% accuracy for most extracted fields.