Incomplete emails in different languages with attachments in various formats in the inbox—B.i.Team aims to help companies with its AI assistant and hosted a webinar on Freshfin and AI-powered order capture to demonstrate its capabilities. AI can also be used in the purchasing process. There will be a separate webinar on this topic, according to Daniela Sternal, Account Manager at Freshfin, and Marius Wessel, Product Manager for Data & AI.
© B.i.Team Gesellschaft für Softwareberatung mbHMarius Wessel and Daniela Sternal during the webinar
"AI document assistant," Wessel began, "is perhaps a very clinical term, but it has its roots in the Microsoft Business Central environment. It essentially revolves around capturing orders in the sales area or purchase orders in the procurement area.2 Many companies receive orders through various channels, sometimes with incomplete information and not always via an electronic interface, he noted. "The entire process is still very often done manually, which is correspondingly time-consuming and error-prone, but it is necessary so that subsequent processes can continue. This is where we aim to bring about an improvement with AI-powered document capture."
The AI reads the information, digitizes the documents, and automatically matches them. "An automatic matching process takes place between the most important key data—namely, customer, supplier, item, and item variant, as well as delivery location and delivery time. Then there is the option for AI-powered document capture to automatically create orders in Freshfin or Business Central when certain thresholds are reached. Simply put, if the matching process has identified the correct customer, the correct item, or the correct item variant with a very high degree of probability, then the order is created automatically. If a threshold isn't met, there's the option to cross-check manually, but the information has already been digitally captured, and you can significantly speed up this verification with a single click."
Processed, under review, ignored 
© B.i.Team Gesellschaft für Softwareberatung mbHOften, companies have a general address for their internal sales department. "The assistant reviews the various emails and searches for specific keywords to determine whether it is an order at all. Then it checks the subject line, email text, and attachments, and searches the email address. This is important to understand where the assistant actually looks to first identify orders and second, extract the relevant information." In this process, the AI searches for, among other things, customers or suppliers, product variants, customer references, delivery date, delivery location, quantity, and unit.
Wessel then demonstrated a sample inbox containing three different folders: processed, under review, and ignored. "There are two stages. Once a certain threshold is reached, an order is automatically processed, and a document is created directly in the ERP." With collective inboxes, you need to know what has been processed by the AI assistant, "which is why these emails are automatically placed in the folders." The language of the emails plays no role in the processing.
Local matching takes place before AI matching, and the AI only intervenes if no match is found. Additionally, the system learns as it is used. "The system learns when manual assignments are made and remembers them." It also remembers affinities, for example, when customers repeatedly place similar orders.
After the email has been processed, everything can be reviewed manually. "I always have the option to view the original email again, go to the attachment via preview, or download the attachment if needed. You can go through the matches step by step and then save them." They can also be approved directly, thereby automatically creating a sales order in Business Central.
At the same time, the corresponding orders are created in the ERP system. "There you can see the customer and the details, including lines with items and their respective variants and quantities. If anything has changed, I always have the option to update the order again."
© B.i.Team Gesellschaft für Softwareberatung mbH
Customization
"I also have the option to check within the application: What has the application learned? And to verify that accordingly." During matching, the AI displays a match probability. "The question is how good the match is, but also: How automatically do you want an order to be created? We've set a threshold of 80% here, so the item match must be over 80% for the order to be created automatically. In the admin settings, you can individually configure when the AI should take effect and at which thresholds; such orders are automatically created, or from when they should be reviewed accordingly. Furthermore, it is also important that we mask the data—especially sensitive data—as soon as it is read so that it is appropriately protected. Of course, you also have the ability to manage and control access accordingly. This means, for example, that not everyone in your organization can access this admin area."
© Hugo Huijbers | FreshPlaza.com
Steffen Krause and Daniela Sternal from BiTeam/Freshfin at this year's Berlin Potato Evening: The software company has, among other things, implemented a customized ERP system at Agropa.
About the company
B.i.Team stands for "Beratung im Team"; a company that has been active in the software industry for many years. "We primarily focus on Microsoft solutions, specifically for the fruit wholesale sector," reported Sternal. "AI and data consistently play a major role here; this business segment is very important to our company."
For more information:
Daniela Sternal
Marius Wessel
B.i.Team Gesellschaft für Softwareberatung mbH
At the RaumFabrik 33c
76227 Karlsruhe
[email protected]
www.freshfin.de
