How Shortwave is Using AI to Take on Superhuman and Gmail

An Adobe Firefly-generated image of a laptop on a desk, showing the Shortwave email client on the screen.
"The AI Economy," a newsletter exploring AI's impact on business, work, society and tech.
This is "The AI Economy," a weekly LinkedIn-first newsletter about AI's influence on business, work, society and tech and written by Ken Yeung. Sign up here.

Eight years ago, I authored an article calling reports of email’s demise greatly exaggerated. The people I spoke with shared how there were still sparks of innovation left within this communication medium. That was all before artificial intelligence. And now, there’s renewed energy in finding new ways to leverage the inbox to improve productivity.

Andrew Lee is the founder of Shortwave, an AI-powered email client. He’s also an accomplished entrepreneur, helping build the developer backend service Firebase, acquired by Google in 2014. In an interview, Lee explains the goal of his latest startup and why he’s betting big on AI to help achieve it.

Justice For Email

“The company started at the beginning of 2020, largely because I saw something really cool, special, and beautiful,” he says. “Email, unlike every other communication app you have on your phone, it’s not one company with one app. If you want to have a group chat with somebody, you don’t have to get them onto your email app; you just have to get their email address, and you can put everyone together…I don’t need to ask, ‘Are you on WhatsApp?’ Or ‘Are you on Signal or whatever?’ I can send you an email.”

Lee cites two pivotal moments that convinced him email was a lingering problem. The first was the Hong Kong 2019 protests, during which he read about the Chinese government using WeChat to surveil the public. He said that if you can’t trust the major chat app, at least people could trust email. Secondly, the demise of Google’s Inbox app showed Lee that if Google “with their infinite money and billions of users on email is not willing to invest in email, we’re all going to end up on WeChat.” 

Realizing there was email potential, he reconnected with his former Firebase colleagues to ask if they’d be willing to “do justice to email” — they said yes. Initially, Shortwave tried recreating the core features from Google Inbox.”Let’s bring back some of the goodness…make these things fast…easy to use.” Eventually, the team said, “holy crap, these LLMs have gotten really good…maybe we could build some real features.”

Screenshot showing the AI autocomplete feature in Shortwave's email client. Photo credit: Shortwave
Screenshot showing the AI autocomplete feature in Shortwave’s email client. Photo credit: Shortwave

Shortwave launched its first AI features in 2022, with an email summary tool and an AI that helps autocomplete messages. Lee said the technology helped define the startup, giving it insight into how to stand out from peers like Superhuman and Spark. In fact, Shortwave is going after Superhuman because, according to Lee, from a business perspective, it’s “the most exciting one, and kind of the one to beat.”

Overcoming the Excel Problem

There’s a reason why we don’t see a plethora of email clients marketed on the internet, and that’s because of something Lee describes as the “Excel problem” or is perhaps better known as the 80-20 rule. 

“It’s so damn hard to compete with Microsoft Excel. There’s no small subset of features in Excel that covers most users, right? So, if you could do 20 percent of the work of rebuilding and get 80 percent of the benefit, there’d be a ton of startups out there,” he explains. “But the problem is, you have to build 80 percent of the product before you can draw a circle around enough users that you can start to gain traction because there are just so many little things you must do. And it turns out, that email’s kind of the same way. Even today, we struggle because there’s just little features that are missing here and there that people expect.”

Lee credits his past founder experience with making him aware of what users want besides a simple feature release. But that wasn’t the only thing Shortwave needed to tackle. It also needed to address its user experience. Lee concedes that in the email world, there’s not a lot of room to make drastic changes to the layout—it’s been tried, and “most of the best practices have already been adopted.”

Screenshot of Shortwave's AI assistant on a mobile device. Photo credit: Shortwave
Screenshot of Shortwave’s AI assistant on a mobile device. Photo credit: Shortwave

So, how did Superhuman grow its business and stand out? Lee believes it’s because they found niche design opportunities—inbox splits, he says—that moved the needle, and they had “really good marketing.”

Undeterred, Shortwave pivoted to AI, and Lee was surprised at the results. Until then, the company grew slowly with marketing efforts described as “hell” thanks to repeated rejection from potential customers. But after making the switch, it “reopened the Pandora’s Box of new things to try. For the past 20 years, everything you could try in email has been tried. But suddenly, it’s like a whole new set of things to try that are radically better. And it has given us a whole new lease on life and a whole new set of opportunities to go after.”

Taking On Superhuman

Shortwave ran through multiple ideas in an attempt to find product-market fit. Because it’s built on Gmail, Its first effort sought to make email collaborative, letting people see and collaborate. It’s Slack meets Gmail. The idea failed to garner adoption. A second thought was to pursue the consumer market, avoiding conflict with Superhuman, which owned the higher-end user. Unfortunately, Shortwave discovered that despite receiving a decent pickup of users, they weren’t the paying types.

The company is now on its third idea, which might be paying off. It’s now targeting the business user, which makes sense since Lee says they’re the only ones willingly writing a check or putting in their credit card. Seeing a path forward, he made a fateful decision, “We’re just going right at Superhuman.”

He rationalized that Shortwave would catch up on the features with its rival but would rely on its AI to differentiate itself. “If you compare us to Superhuman, we have almost parity on the feature side where we’re about to launch some stuff over the next couple of months that should get us basically all the way there…But on the AI side, I think we’re kicking their butts. And that’s how I want people to think about us as the new upstart that’s AI-native.”

The AI Differentiator

Screenshot of Shortwave's AI assistant on a mobile device. Image credit: Shortwave
Screenshot of Shortwave’s AI assistant on a mobile device. Image credit: Shortwave

You might ask how Shortwave differs from Superhuman because both companies use AI. Lee states the variance is in the architecture thanks to work his company repurposed from the team’s time building Firebase. Shortwave has a server that “live-syncs data back and forth from the client.” This was done to facilitate collaboration early on in the company’s history. “But it turns out, this is useful for AI,” Lee explains. “Because if you want to do really cool things with AI, you need access to big GPUs and vector databases. These things are very hard to replicate on the client side.”

He elaborates further: “Superhuman, Spark, Missive, and all these other [companies] are client-side applications. They download your email to the device and run things on it, which before AI was a selling point…And now, you want to embed all your emails and run a bunch of GPUs to run searches. You can’t do that on the device effectively, and we can. So, we ended up in a situation where, unlike every other email client, we have access to your data on the server side. When we get the email, we embed it stored in a vector database. When you ask questions, we can spin up a big cluster of GPUs and throw a lot of horsepower at it.”

Lee cites two features created thanks to Shortwave’s architecture: The first is a Q&A bot in a sidebar that helps write messages, search for specific keywords or conversations in your email, or identify the last time you’ve spoken with someone. When viewing a demo, I was reminded of the sidebar launched on the Microsoft Edge browser when the Windows maker first integrated OpenAI’s GPT. The second is a GitHub Copilot-like autocomplete bot that provides personalized suggestions when you’re struggling to input details into an email.

What About the 800-pound Gorilla?

Companies like Shortwave and Superhuman are platforms built on top of Google’s email service. How can they compete if the Alphabet company invests heavily in AI in Gmail? Lee offered an optimistic outlook, telling me that AI is only added to Gmail because the team is given a mandate from CEO Sundar Pichai. “There is another team at Google building cool AI tech and shoving things they were told to. But I don’t think anyone is really thinking through how to make this a great experience.”

In other words, Lee doesn’t think Google is making AI a part of Gmail’s DNA, unlike what Shortwave is doing.

He also claims that Google’s mandated use of its technology disadvantages the tech giant. “Google tech might not be the best for whatever they’re doing.” 

The same goes for Google’s AI: “I think the models they use are not as good, the weights integrated into the UI, the AI is not as good. They don’t have any of the search tech. So Gemini can now run normal Gmail searches on your behalf as a plugin, but there’s a big difference between having the AI write a regular Gmail search and doing semantic retrieval—which is what we do—where you actually embed the emails and use AI tech for the search itself. And Gmail doesn’t have anything like that. I think it would be very hard for them to do that.”

Simply put, Lee doesn’t think Gmail is moving fast enough, and the team is not thinking holistically about using AI.

“I think the reason we’re going to beat Gmail here is because Gmail has a lot to lose,” he argues. “If they went hard at AI and they screwed up the [user experience] or screwed up privacy, they would nuke a business with two billion users. That would be really scary. So they are going to be really cautious here and will take little baby steps. But what we can do as a startup is say: We have a much smaller user base and can take the risk to say, ‘We might screw this up and screw the whole company.’ But if it works out, we can be ahead of these people and take the lead. And so I think we are betting the company on AI.”

Does such a risk frighten Lee at all? “We could be wrong! We might find that it doesn’t deliver the value people need and we go out of business.” However, it seems the reward outweighs the risk: “If we’re right, the right thing to do here is to go all in. And that’s what we’re doing.”

To reassure himself that Shortwave is on the right path, Lee tells me the team constantly uses its own technology and observes what their users do. They don’t use Slack or any other third-party software. “We’re dog flooding the crap out of this thing, and we’re living it every day and taking it one step at a time.”

Leave a Reply

Discover more from Ken Yeung

Subscribe now to keep reading and get access to the full archive.

Continue reading