We’re Fil, Lambert and Matt, the founders of EasyEmail (https://easyemail.ai). EasyEmail is a Gmail plug-in that helps you write emails quickly. We train our software with your inbox and quickly suggest what you should write based on your previous responses.
We started working on improving how email is used about a year and a half ago, when Fil was organizing the MIT Fall Career Fair, an event for 6,000 people and 450 companies. He was sending 200-300 emails a day and feeling suffocated by the volume. We started chatting about some sort of a solution, and after a long journey (including a car crash on the way to our YC interview!) we finally have a working product and some happy users.
Our current product is a Chrome Extension helps you write emails with two main components: autocomplete, and hotkeys.
Autocomplete searches through every sentence you’ve ever sent in the past and suggests 5 sentences you might say at this moment. An example could be me typing “how a” and the autocomplete suggesting “How are you doing?” together with 4 other sentences. This feature turned out to be harder than we expected because users say things that start similarly a lot (I have 207 unique sentences starting with “how a”). The question is how to sort all those sentences so they’re most likely to choose one of the top 5. Our sentence-matching algorithm includes things like frequency, recency, and context from the email you’re replying to.
Hotkeys are a quick way to enter snippets of text that you repeat a lot, but that aren’t sentences, like a link that you send a lot, or pieces of text that you send often but don’t merit a new template.
It’s very exciting to work on this problem, because email is so universal. That also makes it very hard, because we need to satisfy a lot of different email users. There's also a lot of competition - most prominent is probably SmartReply by Gmail (those 3 buttons saying “sounds good” on your mobile app). The most important difference between us and them is that our suggestions are always personalized, since they come from your own mailbox.
This may sound like we’re trying to remove thoughtful emailing by just making our users repeat the same sentences over and over again, but that’s the exact opposite of what we’re going for. Our initial users tend to already send repetitive emails, and we’re just reducing the amount of typing they have to do. The goal is to give everyone more time to put into the non-repetitive parts of emails!
We’d absolutely love to hear your thoughts on the product and your experiences in this area. If you’d like to try out our product, it’s easier to go right away to the Chrome webstore: https://chrome.google.com/webstore/detail/easyemail-ai/giage.... Please let us know what you think!
I hope you take to heart the privacy concerns here. In case you haven't come across it, here's a useful thread from last year about Unroll.me's datamining and selling of customers' inboxes, and the general trust problem for services like yours:
https://news.ycombinator.com/item?id=14179077
Frankly I'm not sure what it would take to convince me to open up email text access to a 3rd party service at this point.
You have to convince me not just that you (Fil, Lambert and Matt) are trustworthy, but also that the three of you can't be bought out by someone less trustworthy (bearing in mind that your customers may well be worth more to a shady business than they are to you!), and also that you're competent to run a hugely valuable database without ever making a mistake (quoting tptacek in the above thread: in your place, "I would be terrified.")
I don't know how you would crack this, but on the risk-of-hacking side as a user I might be looking for technical reassurance -- say, that you're running entirely locally instead of storing my email on your server in the first place, or that you're only using temporary email access to store sentences with k-anonymity or client-side encryption, and that your privacy model has been published in detail for review.
On the risk-of-purchase side you could look at Keith Porcaro and Sean McDonald's idea of the Civic Trust, where a private company's data is owned by an independent trust that can protect the interests of your users, separate from the vagaries of your business model:
https://medium.com/@McDapper/the-civic-trust-e674f9aeab43 http://digitalpublic.io/
My feedback here may amount to concern trolling, since I'm over on the paranoid end of potential customers these days and probably not your target audience -- but I'm passing it on in case it's helpful ...
We're incorporating some of the things you've already mentioned (client-side decryption, better sanitization), but we'll be a lot more cautious moving forward. You brought up really excellent points, so thank you for sharing those with us!
It definitely isn't slow for all users! However, there are some that send so many emails that any incremental speed-up is very helpful. Plus, we're looking to start working on mobile, where it should be most useful!
If that's the case, that's a fascinating insight - thank you so much for sharing it!
I'm not suggesting at all this will be the case here but this seems related and might be an interesting read.
My second thought was the Cory Doctorow story which predicted that the first true AI will be the result of a machine learning spambot and a machine learning antispam bot getting stuck in an arms race.
This tech would make for an excellent means for spammers and malware authors to bypass spam filters easily.
1)Infect user. 2)Generate new content based on old emails. 3)Infect everyone. 4)Repeat.
Use same content, change attachment
Sadly I came across this before
There is a lot of knowledge hidden in support ticket replies but it’s didficult to know what other team members previous answers would be useful to crib from without guessing search terms. Heck you may even forget exactly how to find your own answer from a few months ago.
On top of that there are concepts we have to communicate all the time as building blocks of a reply, which is where it seems this service may thrive. Let me know when you need beta testers for the zendesk version :-D
1. As someone in sales, there are a truckload of emails I send out each day. Contrary to what a lot of people think, no, we don't send out templated stuff blindly. I think EE should be able to help me shave off about half an hour in think-time in a a day. Huge.
2. Organic-ish. Early days, I know, but I'm hoping that EE won't devolve to a glorified SmartReply with additional text templates.
3. There's a roadmap for pro-users. Which means at some point of time, there _might_ just be an Outlook/Polymail plugin(please say yes). Which is when I absolutely win. :-) Also, it's nice to be relatively assured that your data isn't likely to be sold.
Almost definitely a yes to an Outlook extension coming up - we're just making sure now that we're doing a good enough job on Gmail before we start porting over to another platform :)
Hey - any chance you could email me at filip@easyemail.ai? I'd love to set up a phone call to hear how we could be better for you. Thanks!
1. The link to the privacy policy comes AFTER you take all my data. 2. If I uninstall your extension you have it programmed to give me an exit survey, but it doesn't prompt me to revoke access for your app from my account? So you can continue to read all my communications. At the very least tell me how to remove it when I uninstall your extension, or fill out your exit survey. Don't just hide the fact that you retain access indefinitely. 3. You are appending your marketing to the bottom of every email regardless if it used your product or not. Thats crap, I don't want to spam every single person I communicate with. Let me opt in, or only append when your plugin is used.
I get its a prototype, I get the "potential", but it seems you put 0 critical thought into making the app safe.
1. We're sharing our privacy policy on our chrome extension webstore page - did you manage to see it there? If not, we should definitely do a much better job presenting it, so thank you for pointing it out.
2. So sorry, definitely don't want to hide anything! We'll push out the changes asap to show our users how to remove your permissions. We stop looking at your data once you uninstall (even if you don't revoke permissions).
3.Sorry about that! You can remove it easily in the "EasyEmail" control panel - just hit "Settings" and you can uncheck the button
4. We'll keep getting better! Thanks for pointing out that we're suggesting some generic sentences - this tends to happen when the user didn't send a lot of emails from a given email address (which means there's not a lot of data to learn from.)
Again, thank you so much for your feedback. We'll get right to fixing all the problems!
Yes, you need to remove permissions. That is absolutely imperative. Do you delete the data after uninstall?
Perhaps give an option checkbox in your initial walkthrough of the app to allow the footer only on emails using EasyEmail.
I retracted #4. The first email I got after the loading robot screen made me think the "learning" was complete. I got an email much later saying it was complete. The UX confused me. Perhaps instead of popping up generic message before the learning is done, you can show a message, still analyzing data. So I cannot in good faith comment on the suggestion quality.
We delete data the second we don't have access to users' account. However, we don't store any new data after you uninstall us. The reason is that there's actually no way of tracking who uninstalls us - Chrome Extensions are not providing any real support for it. We're looking into it, and will retroactively remove all data from users who uninstalled us.
Brilliant idea! We'll make the checkbox more visible! Got it, sorry about that. Please let me know what you think once you get some time to play with the extension! Thank you so much again for the feedback, it's incredibly valuable.
Or perhaps this is just the MVP and idea is to scale it further ?
It also seems like we can keep going with a Chrome Extension for a bit - Streak, MixMax or even Grammarly are good examples of companies who are focused around a Chrome extension.
What other platforms do you think we could focus on? Where would you like to see us most?
Great founders + (seemingly) toy/stupid/too hard ideas = how YC selects.
That's certainly a possibility, but for now we're just really focused on making the product really useful.
Privacy and security are extremely high on our priority list since email is such an intimate communication channel.
Perhaps you could have a privacy policy which clearly states no text from user emails will be looked at by humans.
It's possible, though not easy, for a limited form of client-side encryption to be used while still offering autocomplete; e.g. if each word is encrypted client-side with a per-client secret, an encrypted "next word" could be determined and returned without the server ever knowing the specific words and sentences it operated on (other than length). There are other caveats here.
> Privacy and security are extremely high on our priority list
The lack-luster privacy policy doesn't show that at all.
We're working on client-side encryption right now actually, which will further help us be better about your privacy.
The only people who can read email sentences are our researchers - and even then the sentences are not identifiable. They are obviously bound by very restrictive NDAs.
I don't think this would be practically possible. As these guys have to improve algorithm for better suggestion, someone has to look at the suggestion and the context to improve the algorithm.
To me, it sounds like a personal assistant that can automatically reply for you to fairly complex emails. We get SO MUCH email, maybe this will actually help us manage all that.
It is a very powerful idea, all the best to you.
That would make me redundant. 4 hr work week, I am coming.
Then again it’s a gmail plugin :P
We've got no plans whatsoever to sell anyone's data - we'll be a good ol' regular SaaS product for business with a free version for personal use.
We don't have a number for the percentage of usage for what the user specifically types, but it sounds like definitely something we should have, so thanks for the suggestion!
It sounds like a violation of user privacy. Data-collection and privacy is a fine line, and I'm pretty sure any numbers related to specific text users type in private communication being surfaced up to humans is across that line.
I'm fine with my phone keyboard suggesting next words. I would not be fine with a human looking at the data model for specific sentences and words, even in aggregate.
I'll make sure we review the privacy policy once again - you've been bringing up some extremely useful points, so thank you so much!