Summary
The podcast episode titled "Email like a superhuman" features Loïc Houssier, Head of Engineering at Superhuman, discussing how AI and large language models (LLMs) are revolutionizing the email experience. Superhuman was an early pioneer in integrating AI into email clients, enhancing user speed and productivity by intelligently augmenting workflows in platforms like Gmail and Outlook. The rise of LLMs marks a paradigm shift from simple classification tasks to sophisticated, context-aware features such as summarization and content generation, driving dramatically increased user expectations. However, the AI output quality's heavy dependency on user prompt skill presents unique engineering and product challenges, necessitating systems that balance user customization with safeguards like automatic classifiers and system prompts. Superhuman employs a hybrid model where automated classifiers handle typical unwanted emails, while users can still craft personalized labels using natural language prompts, maintaining both reliability and flexibility. Recognizing that many users lack prompt engineering skills, Superhuman is developing prompt libraries and educational tools to improve user input quality, positioning prompt engineering as a pivotal user-facing skill. The episode also underscores the importance of rethinking traditional email workflows and UX designs, as AI fundamentally changes interaction patterns, requiring designs that preserve user familiarity while leveraging new capabilities. Additionally, the shift towards conversational and voice-based interfaces is anticipated to reshape user interaction paradigms beyond keyboard shortcuts, with current tools like ChatGPT and Whisperflow already nudging in this direction. Despite evolving interfaces, core mental models such as the inbox-as-task-list remain central, though AI will enhance how tasks are surfaced and prioritized. Rapid and disruptive AI innovation compels product teams — especially smaller, agile ones like Superhuman’s — to frequently pivot and reassess priorities, contrasting with slower processes in larger enterprises. Superhuman optimizes the trade-offs between performance and infrastructure costs by strategically switching between expensive and fine-tuned AI models, maintaining scalability and user satisfaction. Lastly, enterprise adoption of AI tools navigates the tension between heavy compliance demands and the pressing efficiency gains AI offers, with C-suite leaders increasingly pushing for adoption despite traditional risk aversion. Throughout, Superhuman emphasizes user trust, education, and delivering reliable, time-saving features that genuinely augment human creativity and decision-making in the modern email context.
Key Takeaways
- 1Superhuman established itself as a pioneering AI-enhanced email client by integrating AI capabilities early, significantly increasing user efficiency on platforms like Gmail and Outlook. Its focus on speeding up email workflows validated the practical value of AI in productivity tools, influencing subsequent entrants to adopt similar large language models (LLMs) and AI agents. This early market presence underscored the necessity of AI-first design in email clients and positioned Superhuman uniquely within the productivity software space.
- 2The advent of LLMs and AI agents has dramatically expanded the capabilities of email productivity tools from simple classification to complex, context-aware tasks such as summarization, analysis, and content generation. This evolution marks a paradigm shift, elevating user expectations toward smarter, faster, and more reliable AI-driven features in email applications. Superhuman has experienced this transition firsthand, moving from accelerating basic email handling to delivering truly intelligent assistance.
- 3AI output quality in email applications is heavily influenced by the user’s skill in crafting effective prompts, leading to variability that presents unique engineering and product management challenges. Superhuman addresses this via system prompts surrounding user inputs, automatic classification to handle routine tasks, and comprehensive user education on prompt engineering. Balancing user empowerment with output consistency requires innovative product designs that mitigate risks without constraining flexibility.
- 4Superhuman implements a hybrid AI approach combining automatic classifiers that reliably filter common unwanted emails with user-driven natural language prompts for customizable labeling. This methodology balances automation’s efficiency and consistency with the flexibility desired by power users, reducing cognitive load and risk from poorly structured prompts while preserving personalization.
- 5Prompt engineering is a critical user-facing skill influencing AI productivity tool effectiveness, prompting Superhuman to invest in prompt libraries and educational initiatives that facilitate prompt sharing and improved user competency. This elevates prompt engineering from a niche technical skill to a mainstream design consideration in AI-powered applications, emphasizing its role in achieving consistent AI output quality.
- 6Introducing AI features necessitates reimagining traditional email workflows because AI significantly alters user interaction patterns. Successful AI integration depends on deep understanding of existing human workflows, followed by careful, incremental evolution that balances innovation with preservation of user familiarity and usability.
- 7The evolving paradigm of user interaction with software, especially email clients like Superhuman, is shifting from traditional keyboard and mouse inputs to conversational and voice interfaces. Although not fully established, early adoption of tools such as ChatGPT and Whisperflow highlight a trend toward natural language interactions, hinting at a future where users communicate with applications conversationally to command tasks naturally.
- 8Despite AI-driven interface and interaction changes, core mental models such as viewing the email inbox as a timeline or prioritized task list remain fundamental. AI enhances how these tasks are surfaced and managed but does not discard these familiar concepts, ensuring cognitive continuity and easing user transitions.
- 9The rapid, disruptive pace of advances in AI technology, particularly in large language models and multimodal models, demands exceptional agility from product development teams. Smaller companies like Superhuman can rapidly pivot and incorporate innovations on a cadence of weeks, while larger enterprises face challenges imposed by bureaucratic planning cycles and slower decision-making processes.
- 10Superhuman balances advanced AI feature performance with infrastructure cost by employing an engineering strategy that initially uses high-quality, potentially expensive AI models for validation and then switches to fine-tuned or cheaper models once proven effective. They also use A/B testing to ensure features are addictive and valuable before scaling, optimizing computational resources without sacrificing user experience.
Notable Quotes
"Well, I know this is kind of interesting because I know Superhuman, I think, is one of maybe this sort of first really integrated AI first kind of engineering tools that I remember seeing. And of course, the AI space has advanced a lot in that time."
"Superhuman was almost the only one supercharging Gmail and Outlook. We were the only one on the space making people faster going through their emails and all of that."
"And now with LLMs, a bunch of the perceived quality depends on your prompt. So you have users that are prompting with different skills or different level of skills and the outcome of that prompt may be perceived as low quality. But that's something that is really hard to control."
"We have those auto labels. So automatic labels that will basically flag your emails. And based on the label, you can decide to skip your inbox altogether."
"And one of the things that he talks about is how if you rethink a process that was very human and manual before, often the way that you would make that an augmented or machine driven process is very different from what the original human process would look like."
"So, like what will stay, what will be slightly different? I'm pretty sure that the conversational aspect would be a strong paradigm. Like right now you don't talk, whether it is like through your keyboard or through a mic, you don't really talk to your system. You don't talk to the application. Maybe you start talking with ChatGPT because they have this nice voice interaction. Maybe you use Whisperflow or like this type of tools to basically write your email or like write in Slack and your messages. But you're not exactly commanding the device to do things as you talk just yet. But more and more people are doing so. I probably talk to my computer now more than I type, interestingly."
"One thing that I do believe will stay though, to your point, Daniel, and I would talk about email especially. The concept of inbox, like the concept of having like some sort of like a timeline of things that you need to go through and get rid of the stuff that are top of mind. Some sort of like a task list to some extent will stay. Now, how it will be surfaced, how you will go through it will dramatically change over time. And we're already like seeing this."
"What we've seen, especially with AI, is like the rate of those disruptive innovation is mind-blowing. I would say before AI, to some extent, like the technical innovation where maybe once a year, once every two years, like you have something that is like brand new and like, holy shit, I need to use this. And pardon my French. But what is interesting with like LLMs, like every two weeks or three weeks, if you're not on Twitter, you're not on Hacker News, like you can miss like the new big stuff."
"So being close to a, I would say, a close-knit team that is talking like basically on a daily basis to make sure that you're making the right decision is key. And by the way, just for listeners, you may have heard MCP in there. If we did an episode explaining what MCP is. So anyone who's not familiar with it, you should jump back a few episodes and hear that out. It'll give you some context around that."
"You know, how do you, when you get kind of granular on the product, how are you starting to think about that now? So, you know, I think the most important thing is that there's a hard way to bring a data online tool for you to think about those ways that you're on the machine. It and the results are just mind-blowing like the users find it like so addictive because it's relatively accurate and they win a lot of time. Like it's just about like winning time. Our users are mostly CEOs CXOs on the sales side as well some consultancy firm they leave like basically day in and day out like in their emails so every 10 seconds that you can make them win in their day is a huge win for them given like the amount of emails that they have so this is like one of those features that is super effective even if it sounds simple."
"Not everything is like black or white; there are nuances of gray now in terms of perceived quality. So you need to have more of a statistical approach in terms of understanding the impact of one model versus the others. Of course, we have internal evals and all of that to do our own testing with our golden data set, but the reality is we have a diverse set of customers and everyone is different, so we need to have a broader perspective than just relying on our own data set."
"The perceived quality depends now — we're in a world with way more subtleties with LLMs. Setting the right expectation, basically explaining that the way the feature can be built and sometimes failing because the feedback is not great might not be because it's not well implemented but maybe there's more to it. There's too much latitude offered to the end user, maybe some work on the prompt side is needed. That's something that hit me early on where the perception of the feature was like, 'This is terrible work, like it's not working; people are complaining.' Guys, what have you done? And the work was done properly, but the perceived quality of some of those features can be completely different based on those new aspects."
"Moving to enterprise is pretty heavy. You need a lot of features, a lot of compliance, a lot of things that are not directly improving your product but improving the confidence of those companies that you are the right partner to work on those. There's a shift now, especially with enterprises and Fortune 500s, where the risk associated with lesser compliance from a small company is completely counterbalanced by the cost and opportunity cost of missing out. We see a definite push from CXOs and their security teams for these AI and productivity tools basically saying, 'Hey guys, you need to make it work because it's improving so much the efficiency of the C-level and the company at large.'"
"The human part that is hard to replicate is creativity, the ability to define and detect patterns and stuff like that. The rise of LLMs is helping us get rid of everything that is mundane. I can give you one example: as part of every interview process, I used to write a thoughtful debrief that took about 20 to 30 minutes after each interview to basically put the pros, cons, and areas to dive in. Now, we are pretty much all using meeting minutes generated from meeting transcripts, taking just three minutes to produce what took half an hour before. This is replacing all the mundane work, allowing me and my engineers to focus on brain power and understanding the user, which is the core of their job."
"There's a lot of mental load that comes with using AI tools: managing context switching, guiding models in different ways, and reviewing outputs because sometimes the AI can make crazy mistakes that a regular engineer wouldn't. At the same time, it's definitely saving a ton of time for our engineers so that they can focus on the core of their job — understanding the user and determining the smartest way to get things done. These tools aren’t perfect; they’re ‘basically an intern’ that you need to supervise, but the efficiency gains are undeniable."