Messaging is a mess. The SMS messaging inbox on mobile devices hasn’t changed in over two decades. Ever since the launch of SMS, it has been a simple chronological list of plain-text messages. Admittedly, the simplicity drove its mass adoption; it was simple and easy, anyone could use it and literally every mobile user did. That was fine when the average user was getting a small number of messages. But not anymore, now that users are inundated with a substantially higher volume of messages.
Mobile messaging is used by enterprises worldwide to communicate with their users. Banks send transaction notifications; e-commerce companies send package tracking updates; airlines and travel services send schedule updates; hotel, cab and food delivery services send booking confirmations; and so on. More enterprises have come to rely on mobile messaging to update customers and to increase customer engagement, as well as to reduce transaction fraud and customer support costs. Meanwhile, consumers have come to rely on these notifications as way to stay in control of their life—as a way to manage shopping, banking, travel, payments, etc.
The simple inbox is no longer able to cope with the high messaging volumes. It has become cluttered, dense and disorganized. One way to solve this problem is to make the messaging inbox smarter—to enable it to automatically organize messages, to reduce clutter, to reduce density. Now, we are seeing the launch of a solution to this, which automatically classifies and visualizes messages in the inbox by embedding this function inside its messaging app. As new text messages arrive, the artificial intelligence (AI) module first classifies it based on its content into one of many folders, such as personal, promotional and transactional. Then, the AI module extracts key entities from the message and displays it as a card with simple, clear info. It’s as if your personal assistant has organized and highlighted the key content requiring your attention. The AI adheres to strict user privacy constraints by doing all computation on the device—no user content is sent to the server.
The AI trains using millions of sample enterprise messages and learns to recognize patterns in the syntax and semantics of these messages. It learns to classify messages into the right category (e.g., banking, e-commerce, travel, etc.) and sub-category (bill due, balance, credit, debit, etc.) of a multi-level hierarchy. The AI is further trained to identify and extract key entities (e.g., amount, due date, account number) from the message content. With this type of AI rigorously field-tested and extensively trained, it is able to accurately handle a wide variety of message content. Lastly, the AI model is further optimized to do all of its computation solely on the device without sending any data to the server.
This approach of embedding AI into messaging apps works well with the status quo; it seamlessly works on existing text messages requiring no change on the part of enterprises, mobile operators or consumers. It benefits every player in the enterprise-messaging ecosystem: enterprises benefit from higher read-rates and conversion-rates of classified cards; mobile operators benefit from the increased efficiency of text messaging further extending its useful life; consumers benefit from the reduced clutter and delightful experience.
Classified cards represent probably the first, and long-overdue, major innovation in mobile messaging. Given how heavily used messaging apps are, it’s surprising that it’s received such little attention for so long. This also represents the first step of a long journey; the AI will indeed get smarter over time and expand its capabilities. This is the first step in making messaging smarter and taking the mess out of messaging.