Making the most of information and data is one of the challenges facing businesses in today's digitized world. This information is often collected in the form of documents, whether physical or digital, and in the transactions they carry out internally and externally. 

The problem is that companies exploit their data through traditional tools and that only 20% of them are structured. In other words, most of the information they handle is in the form of video, images, scanned files, etc., which cannot be interpreted by software and computer programs.

For this reason, and bearing in mind that documents will continue to exist, it is necessary to find mechanisms with which to extract their content and value. But, what tool can help us to achieve this?

One solution is to incorporate Document AI, a service that leverages pre-trained Machine Learning (ML) and Artificial Intelligence (AI) models from Google Cloud. This technology can be applied to all sectors. However, to give some concrete examples, this post looks at the potential that the tool offers to energy enterprises.

There are many use cases in the utilities sector where Document AI can be implemented. However, for this post, we have chosen three scenarios that can be quite representative of energy companies: invoice data capture, document scanning and email categorization. 

• Providing a personalized price based on data from a bill (contracted power, consumption, etc.)
• Extracting data from very specific documents automatically, such as social vouchers.
• Classification of e-mails, to group them and redirect them to the corresponding department.

All of them are use cases that correspond to different forms of customer-business interaction. Let's look at them in more detail.

Capturing data from an electricity bill to provide a personalized offer

The traditional utility procurement process is often long and tedious. It usually involves several phases, starting with the search for information. Visiting websites to compare prices, requesting information through forms... Sometimes customers have to negotiate the offer, which involves exchanging emails or calls until the contract is finalized.

In short, all these procedures are characterized by taking hours or even days to complete the contracting process. In addition, customer service hours are usually restricted, which is inconvenient for the user. Other incidents may also arise, such as delays in the exchange of documentation. 

This slowness can lead companies to lose potential customers who were willing to enter into a contract in the first place. However, this process would be much faster with Document AI. For example, the starting point could be the company's website or interaction via a messaging channel. 

Hence, the customer could send the company an invoice from his current company. The customer would only need to send an image or a PDF file. Thus, Document AI would extract the relevant fields such as contracted power, CUPS code and other billing data. With this data and depending on the systems linked to it, the company could make an offer to the customer. If a virtual assistant is also integrated, it could be used to guide the customer through the remaining steps of the contracting process.

Automating utility-specific data mining with Document AI

In the contracting process, the customer must send several additional documents. This is not only for this use case: any processing will require some form or proof from the consumer. 

One possible scenario could be social vouchers, for which family book, certificates and other proofs are requested. In this case, Document AI avoids a manual process of reviewing these documents. This technology not only extracts the necessary information, as any Optical Character Recognition (OCR) would do, but also offers it in a structured way in a set of <key:value>, facilitating its storage in a database or CRM used by the company. 

On the other hand, applying Document AI in the utilities sector also helps to improve the user experience. Consumers would only have to upload an image or pdf and the system takes care of the rest.  

Another case could be the registration of a new dwelling. A common document required in these circumstances is the low voltage installation certificate. This file usually contains handwritten data and is usually of poor quality. 

Document AI makes it possible to upload this statement by transforming written information (ID card number, address, file number, etc.) into structured content. This automates a task that, until now, required the intervention of a human agent. 

Automatic email sorting 

One of the most important channels of communication between customer and company is e-mail. Thousands of users contact companies using this channel. There are many reasons why consumers send these emails. Therefore, the task of classifying these requests requires a heavy workload for contact centers. 

An email usually consists of a subject line, a body text and some attachments. Document AI helps to extract the content of all these elements.

For this classification, we use a classification model based on Natural Language Processing: AutoML. An algorithm that is in charge of recognizing and classifying the content of the attachments. For example, automatically "tagging" an attachment as "invoice", "photograph", "complaint/claim", etc.

Implementing Document AI as a solution in the utilities sector

Until recently, there were practically no tools with which to process all that data in unstructured format: images, physical or scanned documents... Now, thanks to the power and flexibility of the cloud, we can start to see services such as Document AI.

Document AI solution is focused on text and handwritten content as well as diagrams and more visual information. Moreover, this is a very interesting service to improve the user experience, streamline processes and extract information. If you want to consult or find out more about how to implement this technology for your company, please contact us so we can help you.