Under The Shell: How Butter.ai Was Born

Last week, Butter.ai celebrated the public launch of their service, an AI-powered tool that makes finding documents on shared work apps such as Dropbox, Google Drive and Evernote smooth as, well, butter. Congrats to the entire team!

Now that the dust has settled, it’s a good time for the first of what we’re calling Under The Shell, a series where we take a behind the scenes look at All Turtles’ studio companies to shed some light on entrepreneurship, product development, and AI as we see it.

Familiar problem, new approach

Butter.ai has been a work in progress for over a year, but the relationship between their team and ours goes back more than half a decade. Butter.ai’s CEO Jack Hirsch and All Turtle’s CEO Phil Libin worked closely together at Evernote, where the nature of modern work and productivity were very top of mind.

Keeping notes are only useful if you can find them again, so for a product like Evernote to work, search had to be really, really good. But from that insight emerged a frustration: no matter how good search within Evernote was, it was only somewhat useful if it didn’t have a full view of all the online space where your work happens.

That was 2013, and since then, the number of places where work happens has only grown. Today, the employees at a typical enterprise use more than 1,000 apps. There still isn’t a good way to search between them, and it’s a big, big problem. Jim’s preference for Dropbox and Janet’s penchant for Box when the team is supposed to use Google Drive only adds new complexity to the already siloed world of organizational knowledge.

Seeing the enormity of the issue, Jack left Evernote with natural language search specialist Adam Walz and teamed up with co-founders Juan Carlos Perez, Jordan Knox, and Danial Jaffry. They set out to create a service with a singular purpose: search multiple platforms to help you find that needed file, no matter where it lives. They named it Butter.ai, after the Rick and Morty robot who was built with a similarly singular purpose in mind: pass the butter.

Early insights, technical challenges, and privacy considerations

With this problem in sight, they conducted surveys and studies, and sought to learn everything about it. A blog post that outlined their intentions garnered a positive response and attracted thousands of people to their waitlist — a perfect audience for Butter.ai’s exploratory research.

From their probing, they discovered some key insights that illuminated the true cost of the problem. The first was how much of a timesuck finding documents really is. The team found that there are two types of requests, each requiring different investments of time. This first was a search for yourself (“where’s that Acme presentation I did months ago?”), which survey respondents said took about six minutes of jumping out of context and back in again. The second was a search for others (“Hey Jordan, can you send me that Acme presentation you did a few months ago?”), which took nine minutes on average once all the social niceties and context switching were accounted for.

The second insight was how expensive the cost of these distractions were. As a request ripples across an organization of knowledge workers, the price tag adds up. Butter’s calculations showed that, distractions could cost an organization tens of thousands of dollars in lost productivity per year. The opportunity was there, but getting there wouldn’t be challenge free.

For one, bridging multiple cloud services was technically ambitious. The backend infrastructure required to ingest large volumes of content was significant, and whereas the cost of acquiring a new user with very little data was negligible for a cloud service like Dropbox, it was potentially prohibitively expensive for a new Butter user with gigabytes of data spread across several services. For another, user data and privacy posed a potential barrier to use. Asking a user to trust Butter.ai with their digital content was a tall order, so the service needed to be structured in a way that respected user privacy.

The Butter team started with the privacy hurdle and found helpful guidance from an unlikely field: medicine. HIPAA compliance had already outlined some practices for the protection of patient privacy that would prove to be helpful guideposts. Butter.ai would be HIPAA compliant and the team would build in the necessary data and server hygiene required to do so from day one.

More difficult were the technical requirements of actually building a HIPAA compliant system that scaled well. The Butter cofounders began hiring engineers and got to work, building an architecture that denied human access to any of the data itself.

As Butter was planning their product, they decided to develop it with a Slack-first approach. Slack was a service they knew well and loved, and it provided the social graph that their service needed. Jack, already with backing from General Catalyst, also got in touch with Slack Fund and got buy-in from two of his most critical business partners there. His relationship with Phil Libin brought the Butter team to the All Turtles studio, where they came to finish and launch their product.

A first for All Turtles

Butter’s launch marks the first product launched out of the All Turtles studio. In many ways it’s a fitting first product, because it meets the criteria we use to determine what to work on:

  • The problem is real. Cloud work apps facilitate collaboration, but their proliferation represents a real cost to organizations already struggling to find and share files that are scattered on many different platforms.
  • The tech is accessible. Slack is a great entry point to solving this problem. It’s easy to use and widely adopted in offices around the world.
  • The solution is practical. As an always-on Slack assistant, it’s right where the conversations around your work are already happening and requires little thought or effort to use.

Butter was also intriguing because it began to answer in very specific ways the broader question of how AI will help enterprises. For one, it improves the quality-of-life at work by reducing distractions and saving the cost associated with them. For another, Butter.ai makes it easier for individuals to find specific documents and files — where’s that thing I need? — while providing the whole team with a better awareness of what’s happening at work. Lastly, it welcomes non-uniformity by acknowledging hey, people work differently. Butter.ai makes it easy for organizations to respect people’s workflows and preferences. Go ahead, use the tools you want to use.

Butter.ai also begins to poke at how organizations can stay collaborative even as they become larger. Until now, to be constructive on a project, it was pretty much necessary to be connected to the team working on it. Butter.ai helps everyone at a company stay abreast, no matter what time zone they’re in. People across teams know when their work overlaps with work happening elsewhere in the organization. As a company grows or begins to feel the strain of a distributed workforce, that context makes all the difference.

All of this is something that digital assistants today simply cannot do. For all their attention and mainstream adoption, Siri, Cortana, and Alexa lack any real work context outside of your calendar. Butter.ai has that. And what that context makes possible is really, really exciting.

Congrats again to the Butter.ai team. We can’t wait to see what’s next.