Superhuman’s new auto-draft feature almost makes me like AI replies

Superhuman, an email client known for its speed and efficiency, has introduced a revamped auto-draft feature designed to generate email replies that sound more human-like than previous iterations. The update aims to reduce the burden of responding to a high volume of incoming messages by leveraging advanced artificial intelligence models. According to user feedback, the feature has already prompted users to rely on generated drafts with minimal editing, marking a shift in how individuals manage their digital communication. The new auto-draft function builds upon earlier attempts by Superhuman to integrate AI into email management. Past features such as instant replies and follow-up auto-drafts often produced responses that felt overly polished or robotic, leading to low adoption rates among users. However, the latest iteration appears to address these concerns by analyzing the user's tone from prior interactions and generating multiple reply options tailored to the conversation's context. During a trial period, users found that the system was capable of drafting responses to pitches, confirming meeting times, and declining requests for authored content, such as rejecting a request for a post on TechCrunch. While the feature occasionally proposed responses that were too agreeable or inconvenient, users could quickly choose alternative suggestions from the available options. The system adapts to user behavior, refining its output based on past interactions. After encountering issues with suggested late-night meeting times, the AI adjusted its approach, offering more suitable alternatives when similar situations arose. This learning capability suggests that the feature becomes increasingly accurate over time as it gains more data from user interactions. Superhuman’s co-founder, Rahul Vohra, shared insights during the testing phase, noting that 40% of auto-generated drafts were sent within a day, with 60% of those being dispatched without any manual input. Earlier versions of the auto-reply feature relied on less sophisticated models such as GPT-3.5, which lacked the depth and contextual understanding of the current implementation. The updated feature utilizes a combination of models from Anthropic and OpenAI, allowing for greater flexibility and intelligence in generating responses. This integration represents a strategic move toward enhancing the overall user experience by providing more nuanced and contextually appropriate replies. Following its acquisition by Grammarly last year, Superhuman has undergone a rebranding process and is now developing an assistant named Superhuman Go. This new tool is intended to operate across multiple platforms, maintaining continuity in user interactions by transferring context seamlessly between applications. As the use of AI in daily communication continues to evolve, tools like Superhuman’s auto-draft feature offer potential benefits for managing the increasing flow of digital correspondence. While complete reliance on AI for handling inboxes remains a point of caution for many users, the ability to generate quick, personalized responses could significantly enhance productivity for professionals dealing with high volumes of email traffic.