Editorial illustration for Superhuman's AI Feature Drafts Email Replies in Your Own Tone
Superhuman's AI Drafts Emails in Your Own Tone
Superhuman is rolling out a new version of its auto-draft feature, and this time the emails it writes don't sound like a chatbot wrote them. The email client, which has spent years trying to fix the overflowing inbox problem with AI, has tried this before with instant replies and follow-up drafts. Those mostly flopped. The drafts read like an overeager AI salesperson trying too hard, and most users, including plenty at TechCrunch, skipped them entirely.
This update works differently. Superhuman's system now scans incoming mail, flags which messages actually need a response, and drafts a reply modeled on how you've written to that person before. It doesn't stop at one draft either.
The feature spits out three variations, giving you options instead of a single take-it-or-leave-it response. After getting early access to the beta, TechCrunch tested it on real inbox traffic, including pitches, meeting requests, and the kind of authored-post asks that TechCrunch doesn't accept in the first place. The results, and the rough edges that came with them, say a lot about where AI email tools stand right now.
Email client Superhuman is launching a new version of its auto-draft feature that identifies important emails and creates draft replies that sound less robotic.
Why this matters
Superhuman's pitch is narrow and testable: does the draft sound like you, and does it save time without embarrassing you. That's a lower bar than the industry's usual promise of AI managing your whole inbox, and it's probably why it works better. Founders building on top of LLMs should take note: the winning feature here isn't "AI writes your email," it's "AI writes three options and gets out of the way," letting the human pick the tone that fits.
For developers, the interesting technical bit is the tone-matching from prior conversations rather than generic style transfer, since that's the part competitors have consistently botched. Researchers watching the assistant-email space should track failure modes, not just wins. The embargo-agreement example is a real one: drafts that commit to something in a pitch reply carry actual risk if a user hits send without reading closely.
Superhuman's own history of "robotic" auto-drafts is a reminder that tone matching is hard to get right consistently, and one good demo doesn't mean the failure rate is low enough for people to trust it unsupervised yet.
Related Reading
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
From Archives ChatGPT Returns to WhatsApp in Europe Using GPT-5.5
From Archives New York enacts first US state data center moratorium
Common Questions Answered
How does Superhuman's new auto-draft feature differ from its previous instant replies and follow-up drafts?
Superhuman's updated auto-draft feature generates email replies that sound natural and match the user's own tone, rather than sounding robotic like previous versions. The earlier instant replies and follow-up drafts were largely ignored by users because they read like overeager AI salespeople trying too hard, whereas this new iteration focuses on creating authentic-sounding responses that don't embarrass the sender.
What is Superhuman's core pitch for its new auto-draft feature?
Superhuman's pitch is narrow and testable: the auto-draft should sound like you and save time without embarrassing you. This represents a lower bar than the industry's typical promise of AI managing your entire inbox, which is why it works more effectively for users.
How does Superhuman's approach to AI-generated emails differ from the broader industry approach?
Rather than promising that AI will manage your entire inbox, Superhuman's winning feature is that AI writes multiple options and then gets out of the way, allowing the human to pick the tone that fits their needs. This user-centric approach of providing choices rather than full automation is what makes the feature more practical and successful compared to industry-wide promises of complete inbox management.
What problem has Superhuman been trying to solve with its AI features over the years?
Superhuman has spent years attempting to fix the overflowing inbox problem using AI technology. The email client has tried multiple approaches including instant replies and follow-up drafts, with the new auto-draft feature representing their latest attempt to address email management challenges.
Further Reading
- Superhuman introduces AI-powered categorization to reduce spammy emails in your inbox - TechCrunch
- I tried Superhuman's new AI email features—they might actually make me more productive - CNET
- I tested Superhuman AI—my full, honest review - TechPoint Africa
- The next evolution of Superhuman AI - Superhuman Blog
- How an AI tone rewriter transforms your email communication - Superhuman Blog
