Countless emails are received every day, especially in customer service and sales. AI solutions significantly reduce the effort required to read, forward and reply to these messages. One of the most popular AI applications implemented by MicroNova is based on a solution for categorizing data - in this case email content and attachments.

AI is used to convert unstructured data into structured data. Structured data, in turn, can be used to trigger automatic actions. In practice, this saves a lot of time and significantly reduces costs: sales employees no longer have to manually forward incoming emails to the right contact person and suitable offers can be created in just a few seconds using so-called "1-click actions".

In order to process incoming sales emails with maximum efficiency, the messages are divided into defined categories such as quote request, order or test license using an AI function. Relevant information such as company address, order number and product is extracted from the email text and attachments in a structured form.

ChatGPT from OpenAI can be used as the basis. The technical implementation is carried out via Microsoft Azure. To ensure full data protection, the GPT4o and GPT4o-mini models are used via the "Sweden region".

At defined intervals, e.g. every 30 minutes, incoming sales emails are retrieved by the system, summarized and returned as structured data in JSON format. JSON is a standard structure for machine reading and further processing. For this purpose, fields are defined that are to be taken into account, such as category, summary, product and offer or order number.

For correct processing, the AI takes on a defined role and receives context through a system message. This defines the AI's function and specifies its task - for example with instructions such as "You are an employee in the sales team", "You are responsible for categorizing emails" or "Distinguish between orders, requests for quotations and other message types". In addition, the AI is provided with examples to simplify the categorization of email content.

The AI receives specific instructions on how to carry out a task via a user message. This can be, for example, the categorization of emails according to the pattern of a certain function definition and taking email attachments into account.

Specific function definitions show the AI clear framework parameters and contain all the fields that the user needs. This allows the AI to precisely structure emails and automatically classify them into the correct category, which significantly reduces manual effort and optimizes the data flow.

An example of how AI is successfully used in practice could look like this:

The sales team receives an email with the subject "AW: Quote AN241234 - ManageEngine Endpoint Central" with the text "Hello, please send me the quote in euros instead of US dollars."

The email is categorized as an offer request with the subcategory "Euro offer" and the product "Endpoint Central" as instructed by the AI.

The message is tagged "Request for quotation", which automatically moves the message to the corresponding folder.

Also very helpful in practice: if a customer writes that their request is urgent, the AI automatically sets the priority of the message to "high".

The fact that emails are automatically converted into a structured form and assigned to the right person alone saves a great deal of time. In addition, the structured data obtained in this way provides the ideal basis for further automation.

With so-called "1-click actions", a suitable offer can be created simply by clicking on the corresponding button. In the example above, clicking on the "Create Euro Offer" button automatically determines the current exchange rate, creates a new version of the offer and converts it directly into euros. This reduces the time required by the sales team from around three minutes to one second.

The error rate is around one to two percent and is therefore comparable to manual categorization. As a rule, we do not recommend full automation: the best results can be achieved when AI does the preliminary work and a team member takes over the final check. This ensures correct execution and optimal customer service.

Can we help you optimize your processes with the help of artificial intelligence?