Adam Oliner is Head of Machine Learning at Slack, the popular workplace collaboration platform. Prior to Slack, he served as Director of Engineering at Splunk, where he developed and led the machine learning team for four years. Adam holds a PhD in computer science from Stanford University, and was a postdoctoral scholar at UC Berkeley’s AMP Lab. You can find him on LinkedIn here.
At a time when coworkers can’t tap their neighbor to ask a question or request a document, Slack has become a business essential. As Slack usage soars, Adam reports that companies are relying on the technology more than ever before to stay connected.
But employees aren’t just leaning on it for instant communication—they’re also leveraging it as a digital history of all company conversations. In fact, Adam notes that Slack is actually an acronym, for Searchable Log of All Conversations and Knowledge.
Whereas video conferences, or even employee memories, are ephemeral records of information, Slack is persistent by default. By leveraging the smart search functionality, employees can access all digital conversations conducted throughout the company.
Adam shares that AI is persistent throughout Slack’s platform, in both user-facing features and in the backend of the platform. Semantic search, autocomplete and automated recommendations throughout the platform are three of the most commonly-used AI-powered functionalities, all designed to make the experience more intuitive and user-friendly.
For example, any time a user begins typing in the search bar, Slack’s machine learning technology suggests relevant people or channels they may be looking for. More recently, Slack added a similar feature for recipient recommendations; now, when a user begins to compose a message Slack suggests additional coworkers or teams the user often communicates with. Another common example is channel recommendations; ranked by internal priority, Slackbot will suggest channels to join that may be useful or relevant.
These smart features are increasingly important as companies rely more and more on Slack as a central repository for company conversations and historical knowledge. Slack understands this need, and is actively working to make information more readily accessible to users. In fact, when Slack recently rewrote the front end of the product it enlarged the search bar and made it more central to the platform display, offering a constant reminder to users that they can easily retrieve valuable information at any time.
As companies increasingly rely on digital methods of communication to succeed, Slack is dedicated to continuous improvement of the platform to make it even more useful. AI is a critical component of this vision; as it learns from user and team data over time, it creates an increasingly intuitive product. Slack’s data scientists also leverage AI in the backend to make better business and product decisions, too.
In addition to Adam’s belief that AI and machine learning will make Slack even more useful, he shares insight on Slack’s other strategies for product improvement, including customer feedback, extensive A/B testing and executive intuition to continuously create a more seamless product.
While Adam acknowledges that there are certainly many challenges with today’s remote work, he also notes that constraints can breed positivity, like creativity and innovation.
In fact, Adam believes that even Slack—a pioneer in keeping employees connected online—would have previously been skeptical that a global SaaS company could function entirely remotely. But the pandemic has created a stress test, and many companies have discovered what’s possible from afar. For many, the time saved from commuting, or even international travel, has been a boon for productivity.
EPISODE 14: ADAM OLINER
Jim Freeze Hi! And welcome. I’m Jim Freeze, and this is The ConversAItion, a podcast airing viewpoints on the impact of artificial intelligence on business and society.
It’s hard to believe that it’s been a full year since our podcast first launched, and it’s even crazier to think about how much the world has changed since then. The evolving role of AI has been a hot topic for years, but it’s been fascinating to see how AI’s impact has expanded as people adjust to a world where home is our hub for work, life and learning.
Without even realizing it, we’ve all been interacting with AI—from the technology we use for work, to the platforms we use to sort our email, or to find recipes. That’s why we’ve dedicated this season of The ConversAItion to exploring AI at Home. Throughout the season, we’ll examine the ways in which AI has not only helped us get by, but also become a standard part of our everyday lives.
We’ll be speaking to a series of industry experts at some of today’s most talked-about brands, and together we’ll take a closer look at how they’re developing and refining AI to help people make the most of their time at home.
We have an exciting lineup of guests this season—many companies you’ll recognize, and various innovative applications of AI that might surprise you. We’re looking forward to it, and excited to have you tuning in.
On today’s episode, we’ll be speaking with Adam Oliner, the head of machine learning at Slack. Throughout the pandemic, Slack’s workplace collaboration platform has kept teams around the world connected and working efficiently. We’ll hear from Adam about how his team has spent years integrating AI technologies like machine learning into the platform, and how each of these AI based technologies and features have made Slack more useful.
Adam, thanks for joining us today and welcome to the ConversAItion.
Adam Oliner Thanks, Jim. It’s good to be here.
Jim Freeze We’re excited about having you. In particular, I was really looking forward to this episode because at Interactions, we are big Slack users and it’s been amazingly efficient for us during the pandemic. As mid-March hit, we went to all work from home and we would not have been able to function without Slack. So you’re talking to somebody who’s a big fan of Slack.
Adam Oliner Fantastic. Glad to hear it.
Jim Freeze I’m sure you’re very glad to hear that. So I’ll start off by just kind of asking a little bit about your background. How did you end up at a Slack? What drew you to the company?
Adam Oliner Well, I wrote a blog post recently called A Recipe for Success with Machine Learning. Around the time that I joined Slack, about a year and a half ago, I was excited by the opportunities to do machine learning there and by the high caliber of the people who I interacted with. The people there are very talented and very kind, and I’m at most one of those things, but they hired me anyway.
Jim Freeze And a sense of humor as well, I see, so that’s good. So I’m assuming most of your users and certainly many, if not most of our listeners, aren’t aware of the extent to which AI is operating within Slack. So would you mind talking a little bit about that and maybe give some specific examples of AI-powered Slack features?
Adam Oliner Certainly. It’s actually all throughout the product. I can give a bunch of examples first that are user-facing AI powered features. The ones that users are probably most familiar with are search and auto-complete. So anytime you use the search functionality or if you type an @ or a #, and it suggests people to tag or channels to link to, the ranking technology, there is using machine learning.
More recently, we added recipient recommendation in the composer. So if you add a couple of people that you want to send the message to, it’ll suggest additional people and cohorts that you typically communicate with. The channel sidebar has a scientific sorting that uses something that we call internally channeled priority. And we use that same machine learning technology for Slack bots suggesting what channels you should join, and when you join a channel it suggests additional channels as well. So recommendations like that are throughout the product.
We also use machine learning on the backend at Slack for our own internal use cases. One example of this is a spam filter, which we also recently published a blog post about, that we use to filter out invite spam in particular. So people would create fake teams and use the invitation mechanism as a way to send out spam. We also use a metric anomaly detector, which looks for performance metric anomalies in our backend services. And finally, we have a variety of data scientists who use it to make business decisions and product decisions.
Jim Freeze It’s obviously, as you said, very pervasive throughout the product. With so much potential for AI within the platform, how does your team determine what to focus on next? How does customer feedback play into that, and what are the other kinds of things you consider when determining what’s next for AI in the platform?
Adam Oliner There are a lot of mechanisms at Slack to gather and prioritize customer feedback. We have a fantastic customer experience team that triages and feeds these requests and feedback back to the engineering teams. And those do drive our prioritization, for sure. We also use data and experimentation to drive product choices. So we measure how our customers are using the product and where we think we can make the experience better for them. We A/B test features in live experiments to see whether it has the effect that we want, then use that to make decisions as to whether to ship or not.
We also have a user research team, so we’ll get volunteers to sit and try to perform some Slack task in front of us. And we’ll watch and take notes and realize sometimes that we have a long way to go. And other times where we’re thrilled that people accomplish the task and are very happy. We also have leadership with strong intuition and opinions about what things should go in the product. They’ve been thinking about this for a long time and have a strong track record of making good such decisions.
Jim Freeze It’s very interesting. It’s very similar to what a lot of tech companies do. When you’ve done some of the user feedback, are there things that you learned that were a real surprise to you?
Adam Oliner So I think it’s interesting, the tension between wanting to make all of our customers happy but also having to behave like a company. So to give an example of this, users had wanted dark mode forever and we really wanted it to give users dark mode. But what we couldn’t tell them is that we were in the process of doing a full front end rewrite. And until that was done and had shipped, we just had to say, “Yes, we hear your feedback. We’ll get to this as soon as we can.” And when we finally shipped it, everyone was very happy and that was great. But that’s sometimes an interesting lack of conversation, so to speak, by necessity.
Jim Freeze It’s a bit of a dilemma, I guess. And once again, very consistent with what I think a lot of tech companies deal with when they’re dealing with their customers. I’m curious, throughout the pandemic, have there been any interesting trends that have emerged about how people are leveraging Slack potentially in new ways that maybe you hadn’t envisioned pre-pandemic?
Adam Oliner I would say that the usage hasn’t necessarily changed a whole lot. We had a pretty wide breadth of companies from pretty much every industry using Slack before. I would say what’s interesting is the extent to which people now rely on it. It increasingly is a business critical tool that a lot of people require in order to continue functioning, as a business and as an employee. And that’s been interesting, that shift to increasingly thinking about ourselves as a business critical application.
Jim Freeze That’s probably most interesting to your marketing team, I would think, in terms of how you position the product and the growth potential for it.
Adam Oliner Possibly.
Jim Freeze How do you think about the role of AI just in general growing over time? Whether it’s within Slack or just in our daily lives in general?
Adam Oliner Well, let me talk about the role in Slack first. I think that in terms of the impact that my team has at the company, some people think of Slack as just a chat tool. You can certainly use it like that, but you’re really not getting the value out of the product. For us, the value comes from having a persistent, searchable history of all of your company’s conversation and knowledge. And that’s actually in the name of the company, some people don’t know Slack is an acronym, actually a backronym for Searchable Log of All Conversations and Knowledge.
And so for people who don’t necessarily use search a lot, or make use of the fact that you can go back and review any decision that got made in Slack, any file that got shared. If you do it in Slack, it provides value to the company forever. If you do it in a video conference, by contrast, you better hope someone was taking notes, it’s ephemeral by default. Whereas, Slack is persistent by default. And a lot of what we do on my team and at Slack generally, is try to provide value to that data, and that information, and those conversations that have been put in Slack so that they provide value in perpetuity. And you see this in terms of attention prioritization, ranking with the kinds that I talked about, recommendations, and so on.
Jim Freeze It’s so interesting you say that because one of the questions I had for you, and you’ve kind of addressed it to some extent. Which is with remote work becoming the norm, I experienced and I think most people are experiencing almost digital communication channel overload. There’s so many different ways to communicate and I think people often find, I mean, just email as an example. I often find it a challenge to find emails that I either sent or I received, I don’t remember who I’ve received them from, I don’t remember where the content is. It’s difficult to find things and what you just talked about was I think, a key capability that many don’t think of Slack for. I think many think of it as a great way to communicate in real time and that’s really interesting that you just hit on that because that was one of my key questions. And do you find more and more customers leveraging the tool that way?
Adam Oliner Oh, certainly, more every day. And we continue to make investments to make that better and to make search a more central part of the product. For example, when we did that recent front end rewrite, you noticed that the search bar got bigger and then more central in the product. We’re trying to put it front and center to make sure people know that the things that they say can be retrieved at anytime.
Jim Freeze That’s just fascinating. So, a last thing I’m just curious about. Do you think there are any silver linings to today’s circumstances in terms of driving workplace productivity and efficiency? Is what we’ve all experienced over the course of the last six months and probably going to experience for another six, how’s that going to impact workplace productivity efficiency? And how do you think about that at Slack?
Adam Oliner There are various versions of the quote, “Constraints breed X,” where X is creativity, or innovation, or what have you. And I think that certainly applies here. If you would ask even folks at Slack, whether you can run a global SaaS business without an office, we probably would have said, “No.” We certainly had a bunch of real estate and we were working there, but now Slack’s global headquarters is Slack. And office space is expensive, so that’s a pretty major revelation. I think for some people it’s certainly true that not having the overhead of commuting, or walking between conference rooms, or traveling across the globe for a meeting has been a huge productivity boost. And that’s not universal, of course. And I say that authoritatively as someone with a toddler, but I think these re-evaluations of what is and is not necessary for a business, hopefully some of that will persist past this most recent crisis.
Jim Freeze It’s so interesting that you say that, our experience has been exactly the same. We were within probably a week of signing a lease on a new corporate headquarters, 55,000 square feet. We were going to spend millions of dollars in building it out and the pandemic hit. And we took a step back and asked ourselves, what are we doing? And we’ve gone from literally within a week of signing for a new corporate headquarters to now becoming a primarily work from home company.
And I totally agree with you, our productivity… I used to spend an hour driving to work, an hour driving home. It’s two additional hours of productivity that I didn’t previously have. And I think a lot of companies are experiencing the exact same thing. I would argue that Slack has been a great enabler of that, so this has been a great conversation. For a platform that I use on a daily basis, there’s so much going on in the background that I suspect our listeners didn’t know about and candidly, I didn’t either. So really, really appreciate your time today and thanks very much for joining us on The ConversAItion, Adam.
Adam Oliner It was fun chatting with you, Jim. Thank you for having me on.
Jim Freeze On our next episode of The ConversAItion, we’ll speak with May [ing] and explore how 1-800-Flowers is using AI to keep up with the growing demand for contactless gift-giving.
This episode of The ConversAItion podcast was produced by Interactions, a Boston-area conversational AI company. I’m Jim Freeze, and we’ll see you next time.