344
Productivity & Workflow355
Automation & Workflow224
Software Development251
Marketing & Growth192
AI Infrastructure & MLOps174
Writing & Content Creation203
Data & Analytics141
Design & Creative170
Photography & Imaging156
Customer Support131
Sales & Outreach125
Voice & Speech135
Education & Learning131
Operations & Admin87
Superhuman is rolling out Auto Drafts 2.0, which suggests replies to important emails based on your past tone, plus two alternative versions.
In short: Superhuman has launched a new version of its Auto Drafts feature that picks emails likely to need a reply and writes suggested responses that can require little editing.
Superhuman, an email app, is releasing “Auto Drafts 2.0.” It is designed to spot messages that are important or time-sensitive, then create a draft reply for you.
The feature tries to match your writing style by looking at how you have replied in earlier conversations. It also offers two alternative drafts, so you can quickly choose a different option if the first one is not right.
In testing described by TechCrunch, some drafts could be sent with little to no changes, like confirming a meeting time or asking for more details on a work pitch. But it was not always accurate. By default, it sometimes suggested overly positive replies to pitches or agreed to meeting times after midnight.
Superhuman says the drafts improve as you use the tool, similar to how a phone keyboard learns your common phrases over time. Users can also add personal details in settings, and attach files or links, to help the drafts sound more like them.
Superhuman co-founder Rahul Vohra said that during testing, 40% of auto-generated drafts were sent within one day. He added that 60% of those were sent without any manual edits. Vohra also said Superhuman uses a mix of AI models, including ones from Anthropic and OpenAI, to write the drafts.
Many people feel buried in email. Tools like this aim to save time on routine replies, but users still need to double-check, since a confident-sounding draft can still be wrong.
Source: TechCrunch AI