DUE TO SOME HEADACHES IN THE PAST, PLEASE NOTE LEGAL CONDITIONS:
David Yakobovitch owns the copyright in and to all content in and transcripts of The HumAIn Podcast, with all rights reserved, as well as his right of publicity.
WHAT YOU’RE WELCOME TO DO: You are welcome to share the below transcript (up to 500 words but not more) in media articles (e.g., The New York Times, LA Times, The Guardian), on your personal website, in a non-commercial article or blog post (e.g., Medium), and/or on a personal social media account for non-commercial purposes, provided that you include attribution to “The HumAIn Podcast” and link back to the humainpodcast.com URL. For the sake of clarity, media outlets with advertising models are permitted to use excerpts from the transcript per the above.
WHAT IS NOT ALLOWED: No one is authorized to copy any portion of the podcast content or use David Yakobovitch’s name, image or likeness for any commercial purpose or use, including without limitation inclusion in any books, e-books, book summaries or synopses, or on a commercial website or social media site (e.g., Facebook, Twitter, Instagram, etc.) that offers or promotes your or another’s products or services. For the sake of clarity, media outlets are permitted to use photos of David Yakobovitch from the media room on humainpodcast.com or (obviously) license photos of David Yakobovitch from Getty Images, etc.
Welcome to our newest season of HumAIn Podcast in 2021. HumAIn is your first look at the startups and industry titans that are leading and disrupting ML and AI, data science, developer tools, and technical education. I am your host David Yakobovitch, and this is HumAIn. If you like this episode, remember to subscribe and leave a review. Now onto the show.
Welcome listeners to the HumAIn Podcast, where we deep dive into topics and augmenting humans, the future of work, developer tools and how to build a world of humans and machines. Today’s guest speaker is Thor Ernstsson. He is the founder of Feedback Loop as well as the CEO and founder of Strata. He’s built startups that have looked at architecture, at data and how to build stronger human relationships. Thor, thanks so much for joining us on the show.
Thanks for having me.
To start off, can you share with us throughout your career, how did you get into working with humans and machines? Can you start for the audience where you came from and what led you to found these ventures?
Happy too. It starts in the very beginning in rural Iceland. I grew up on the Northern Coast of Iceland, in a little fishing village. We’re about 450 people in technology there, which is a little bit different than how we think of it today. But, in a roundabout way, we ended up in New York, 20 years in the US and 10 in New York and absolutely love it here. And the reason is primarily that there’s so much creative energy around, exactly your topic.
How do you leverage technology to scale up what people are doing day to day? And, I started a handful of companies in this space. I’ve been in gaming, healthcare, and enterprise tech. Now, on personal communication and relationship building, the common theme is always: how do you leverage massive data to make recommendations of actions, of communication, of even healthcare options and things like that?
So with rally health, we were combining your clinical data with your self-reported and observed data to figure out recommended communities, content, challenges, things like that. Massively successful does a few billion in revenue, as a part of UnitedHealth Group now. Then in 2014, we started Feedback Loop to really help people understand if a product was going to be successful before you build it, which is obviously a big question.
It turns out we can use similar technology to access 110 million people on demand to figure out: Who you might want to target. Who you might want to talk to. We might want to interview who you might want to get to test your product, things like that, and then getting feedback in real time on any idea or question from really anybody. That was really cool. We got about a half of the Fortune 100 users. Leveraging AI at pretty much every stage, starting with the data entry from the person, translating it into a proper question, properly researching style questions on the backend. Then analyzing both people’s behaviors as well as, obviously, their answers and then presenting it in some sort of format.
Then, at the beginning of COVID, I realized that one of the challenges that we all had is, really, like immediately people throughout your career, and you want to stay in touch. But despite all of our best attention, it’s just really hard. It’s hard to remember. Everybody has manual hacks. They put together spreadsheets or just lists of names on a piece of paper or stacks of business cards.
Then from time to time, they might go through them. But generally we just let these relationships, whether, don’t stay in touch like we should, and COVID, sort of, put a fine point on that for a lot of us. So building a toolkit to really help executives, it could be anybody, but our focus is primarily on executives that have these non-sales relationships that are, and what we hear over and over, they’re our most valuable asset. Yet there’s no tools to manage it. So that’s what we’re building.
So thinking about what you built and scaled at Feedback Loop to where you’re at today on Strata, can you tell us how the work you’ve done at Feedback Loop has informed your decision on this new venture?
Absolutely. So what we were doing at Feedback Loop, the core of it is really you take a business question: Is this going to work, for example. Which is not a well-formed research question. So we have to translate it into the intent of the question. What you’re intending to do is assess functionality or competitors features or price point or messaging or whatever it is.
There is a best practice way to ask those questions, but generally people don’t know what it is. However, what they’ll do is ask the same kind of bad question over and over and over so that our system will just match it to the proper question on the backend. So if you’re asking: Will people want to use this? Then it will actually create a battery of about 10 questions on the backend that properly assesses the intent of your question. That’s exactly the same kind of work that we have to do as Strata, where you might say something in an email, like: let’s catch up next quarter. In many cases you intend to do that.
You do have intent to stay in touch with somebody. But in many others, especially if you’re a VC, you might say things like looking forward to staying in touch when you have zero intention of actually staying in touch with the person. So it’s an interesting challenge of being able to suss out on an individual level what they actually mean, and then helping them carry through whatever it is that they say anything.
Thinking about what you just shared there for, which is about relationships that we all have the best intention to continue to sync up and connect, whether it’s business or personal, not only on someone’s birthday. But staying in touch during the pandemic became a lot harder because it was all virtual.
Then people started doing exactly what you suggested. Let’s put it in the spreadsheet. Let’s put it in the calendar, but it’s a very manual process. There’s always been the classic “the numbers”. We talk about relationships that every human can manage. There was a classic story in the New York Times about Dunbar’s number. It’s a new study that said, on average, each human can only maintain about 150 relationships.
That’s from your friends, your family, your workplace, everything outside of that, it’s just so much for our mind to quantify. We have to put it somewhere. We have to augment it somehow. So what I’m hearing is that you’re building a new operating system, so to speak, to capture not only these 150 relationships, but more than that.
That’s exactly right. Because, even though you can only juggle in your mind, let’s just say 150, and the number is a bit fuzzy, but let’s say that it is 150. You interact with thousands of people throughout your career, and you go to a conference and you meet a bunch of great, interesting people that you want to stay in touch with. You have coworkers that you may have worked with five years ago, 10 years ago, doing either something really fascinating and you want to stay in touch, or they’re just friends and you liked interacting with them and you want to stay in touch.
But just like you said, you can’t. It’s just too hard for us to do as people, but it’s a trivial thing for a computer. So it’s a perfect example of where you can leverage technology, broadly speaking, to augment what we already want to do. We want to be better for our network, for our people. And we want to give back and want to be helpful or not make introductions. We want to be more mindful of what’s going on in their life. We want to remember things. We want to do all these things, but we’re limited by a number of things. Being able to just work in the background and monitor all this stuff, and they give you specific recommendations that you should do this, send this email to this person, you said you were going to do this and I’m going to do it, things like that.
You’re right, that you don’t want to do it just on these obvious occasions, like birthdays. I guess it is fine. It’s nice to get a happy birthday when it’s her birthday. But, you’re going to get a hundred of them or more. The individual relationship and connection kind of gets drowned out. So a good example is if somebody tells you they’re moving, like so many people did during COVID, or if somebody takes a new job, like LinkedIn will tell you to congratulate them with some sort of salary emoji. You don’t really need to reach out when they get the new job, that is almost too much, because they’re overwhelmed by everything.
But if you reach out three months later, how was the job going? That is going to be way more thoughtful, way more, even authentic. Because you’re doing it, because you genuinely care as opposed to when they’re overwhelmed and inundated. So there’s things like that are little things that are basically things that we all do or want to do that Strata helps them not just manage and track, but actually do. We will help them be better people for their network, for their relationships, and they’ll build deeper and more authentic relationships.
When you think about your product Strata, Thor, you mentioned that it goes all back to relationships and that you discover, as a serial entrepreneur, a former founder and founder, again, that your professional network has been important. But there are relationships that could be untapped that we forget about. In LinkedIn you might have more than 150 connections, I would assume, but do you regularly stay in touch with them all? So there’s the opportunity to unlock that network.
That’s exactly right. Most people, when they first think about it, they’re like: I want more out of my network. But when we interview, especially the more senior we can, and we interview people, what we learn is the same thing over and over. It’s not that they want to get something out of their network. It’s not that they want to know who they should reach out to for sale or for deal or for VC. You need to stay in touch with their LPs and stuff like that, but it’s really more about giving back.
It’s more that they want to help their people that look to them and they want to be responsive. So when they get an email from somebody that needs something, the activation energy is just too high of a hurdle. So what we have happened over and over people will declare effectively email bankruptcy. And they’ll just say, like: if you emailed me in the last few months, like, I’m sorry, I’m not gonna be able to get back to try again later, basically.
It’s kind of sad because these are the most important things and what we’ve heard over and over is that it’s not the school you went to necessarily, or the job you have or your network, or anything like that. That matters: It’s the people you went to school with, the people you worked with and the people that you can now lean on or have, or give back to in some way. And personally, as an entrepreneur, that’s by far the most important thing. You can get money from VCs, and that’s fine, but you can’t replace the people that you work with, and to get the right people, personally, it’s always about leaning on my network, which obviously benefits me if I’m doing something.
But also it benefits the other person. If you make an introduction for me, you may end up as a co-founder and we build a business together and everybody benefits. And that’s the other thing that I found so fascinating in this is that there isn’t a one way relationship. So whenever you have something like that, whenever you follow up from an introduction, it could be like the smallest thing you strengthen your relationships all around, and everybody does benefit. Even if it’s just from a casual, like: thinking about you, hope you’re doing well. Thanks a lot for this thing you did a couple years ago.
Emails are literally the vein of everyone’s existence. For me, I did something in the pandemic where I realized I just had email overload from everything. So I separated out my newsletter inbox from my people inbox. Like actual, real emails from real humans that are saying: Hey David, I need your time. And my non people inbox I’ll call my AI inbox, these are at the time of this recording, it’s almost 15,000 unread emails because come on.
You and I were building. Builder’s going to build. I don’t have time to consume all that content, all that content is incredible. I wouldn’t be subscribed to those newsletters if I didn’t love every single one of them. There’s only so much time though in time to get access to those relationships and ensure that we’re thinking with the data first and the AI first. So with that talking about Strata, can you tell me more how it’s data science focused or it’s AI first?
Absolutely. You just highlight a perfect example, people can’t actually track all the communication again. There are so many things that fall through. So what we do first is we start with a bunch of rules. So there’s heuristics around what might be important. It’s this sort of static analysis of your communication and your calendar of your stuff like that. And then what we learn over time is who’s important to you.
So what we’ll do is model, basically, all of your relationships, and show you not just who you talk to a lot and things like that, which is fine, but not that interesting. But rather if there’s an expected cadence, somebody regularly talks to every six to nine months, but you haven’t talked to him in a year or two. That’s most likely somebody that for one reason or another sort of failing.
If you act on it, you strengthen it. If you don’t act on it, then you choose to affect it and let that continue. From a personal standpoint, if you’re looking at it as a person, and you’re saying: then there may be a reason for this and maybe a reason for why the communication stopped. So what we have to do in the data is actually model explicitly.
So first learn through your behaviors, through reactions, things like that. Which is really cool and interesting, but then model all these exceptions and it turns out it’s your ex-wife and it didn’t end well, and you don’t want to talk to her again.
You used to talk to her a lot, but not anymore, or it might be a coworker and you worked with them a lot on a project, and then all of a sudden you’re not working together anymore. So the cadence changes. So what we have to do and what we learn is the data itself isn’t actually the same as what a person would interpret in us. Just because you did something, it doesn’t mean you should necessarily do it again. So now our models are all these funny metha models where we see what the data actually says, but then now we have to start classifying them by the human layer.
But somebody that you had a burst of communication with, and then not anymore, maybe it’s because they’re a lawyer or they’re a real estate agent or they are some sort of vendor and it may sound and look really friendly, but it wouldn’t take long for a person to sound on that’s not really a close relationship. So we’re very much data first and everything we do is around how do your model capture and make recommendations on the data itself without really knowing what it is? So how does a computer start to learn and think more like a person would, when it comes to analyzing now one of the most human things was just your relationships. And it’s really fascinating because the data is almost always wrong in the first pass. So we have to have these models around it to have it make sense.
That’s right. That the data gets muddy and murky. There’s noise, and you want to discover the signal from that noise. As you mentioned, there might be a relationship with a coworker, hundreds or thousands of emails, but does that mean you want to get on the call with them or reconnect perhaps, or perhaps not.
So there is that changing nature of relationships. You and I previously spoke about that. For example, as you mentioned, the X, Y, Z, for that coworker, you’re no longer with. And now relationships have gone not only data first, but also digital first. We’ve seen that with the pandemic that you and I would maybe be recording this podcast in person or getting a cup of coffee, and here we are in the remote or hybrid world: how do you see the changing nature of relationships in the world and with the product?
So, the COVID and just in general, digitization of everything and making everything Zoom makes this problem much worse, because before you would get a coffee, you would see somebody in person, you have all these nonverbal cues, you have all these triggers and all those memories that are way more than what you have when it’s just pixels on a screen.
So you don’t get any of the other things that people rely on to build relationships. So it is kind of flat, single dimensional if you will, as a result. So, a couple of things are gonna happen. People will 100% want to get back in some sort of real world setting. I don’t know if there’s going to be events in the same way or meetings in the same way, but whatever it is, it’ll be some sort of face to face interaction because Zoom is just so limiting for all this stuff.
But the problem is going to get much worse, because you will still have more people digitally than you did before.There are people you work with or people that you’re close to maybe, and I know for myself, there’s a few groups that I’m a part of and there’s one where we meet every month on Zoom and we had a dinner two weeks ago. So we met up for the first time. It was pretty much when we all saw each other in real life. And in five minutes, I learned more about those people than I did in the last 14 months on Zoom. Because it’s not as structured. It’s not, you’re not limited to just what you say, the range of what you communicate and how you communicate. There’s so much greater that I don’t even remember a single conversation we had on Zoom, but I remember everything we talked about at dinner.
So being able to capture that, and really like helping me do more with it is the key. So as COVID goes away, hopefully soon, we’re going to have all these weird hybrid experiences we’re going to continue with. The fact that there are no tools to help in the same way that you have investment management. If you make an investment and it doesn’t matter what you generally hire a person to do it for you, or now you have tools like Wealthfront and Betterment and all those where you could do it automatically and make it easy and do it a little bit at a time.
That’s exactly the kind of thing we want you to do with your network. Or it’s like investing in giving back, doing it in a more thoughtful way and just doing the things you’re already doing.This is crazy that we’ve all been sort of locked in these zoom jails for the last 18 months, and it’s going to be a pretty interesting time as the world start to spinning again.
As we’re leaving the Zoom jails, so to speak, I’ve set a policy called “no Zoom Fridays”. So Fridays I take no video calls. I’ll still take calls, but it’s audio only. It’s a great way to mentally declutter, relax and detox from this concentrated effort that’s staring at pixels for hours upon hours. Whether we’re looking at audio conversations or seeing these pixels on this screen through video, there’s a lot to unpack and uncover, and these conversations are not simply: Hey, this is an action and next steps, but these repeatable conversations could have results or actions. And I want to dive deeper with that thought about Strata. Is it a tool simply to just return to inbox, like Google where they say:I snooze the email, or is there something more, are you uncovering insights based on actions?
That’s exactly right. We’re helping you uncover the things you should be doing, even if you don’t know what you should be doing. That’s kind of the key here is that it’s doing the thinking and the heavy lifting for you. You click to accept it. You can reach out. You can action it. You can say like create a task out of it, basically. So that if I say to you in an email, or if you just send many emails ago, like that you used to introduce me to other speakers or podcasts.
Then, great, in the moment as I’m looking at the email, I might think of a few people, but I guarantee you, as soon as I close that email and go to something else. It’s just going to go away. Nothing is going to bring that back to my mind unless I see it again, or something’s in front of me, but I want to do it. I would want to make introductions, because obviously somebody coming speaking to the audience here is going to be super relevant for them and for you guys. So the question of how do we streamline that process and go from what I want to do to actually doing it, and using Strata to close that gap.
That’s sort of the crux of what we’re doing. So if I say I’m going to send you something, then that effectively creates a task in our system that gets triggered, and after a week or two, it was like, you said you were going to send something. Are you actually going to do that? Yes or no. Right. Now you could go on and do that. Or if no, then fine, you snooze it until later, and you just forget about it. If I say, I’m going to introduce you to Mike. Which Mike? This Mike. Great, Here we go. Here’s the introduction email ready to go.
So make it as easy as possible to just do the thing I say: I’m going to basically be even more accountable using technology to do that as opposed to having to do all of them by myself, because when you’re staring at a blank piece of paper or an unwritten email, basically, that’s a lot harder to get started than if you have a few things ready to go. So if you start with something and you’re editing it versus writing it from scratch, there’s a lot of human behaviors like that just make it easier to make it even more fun and help you do the things you said you’re already going to do. So it’s all about insights and it’s all about actions and it’s all about streamlining the process and closing the gap between intent and actually doing.
I really like where you’re going with this store, because for me, I have, I don’t know if it’s only me, but I have this never ending to-do list. Of course I prioritize it and use Pomodoro techniques and all the tricks in the books. Optimize it. But even then it’s never ending. And at one point or another I’ll go into it and realize I was supposed to do this weeks ago. What happened? If only there was a system to then remind me for that, and it sounds like you are solving for a gap in the market that will be beneficial, both to consumers and enterprises, and there’s a lot of technology that’s enabling that. That technology sounds like it is looking at the core AI breakthroughs that we’ve seen this decade.
There’s been two core breakthroughs in AI. One has been around computer vision with video and audio. Though, Initially you’re focusing more on the text side, the natural language processing of the content and context that we’re seeing in the emails. If I’m writing hundreds of words or just a few words, is there any context there that can impact the actions and follow through as I should take. Share with us more about why natural language processing is so important for Strata.
For sure. There’s a lot of really interesting work that has been done that we can leverage in your right, that like building this from scratch even 10 years ago would not be possible. It’s everything from memory constraints on the actual servers. The fact that I can spin up a 90, it was a 96 or 92 core Amazon instance and just at the click of a button and trained a model. I couldn’t have done that before. So it would have been prohibitively expensive and improvely hard, actually, it’s just not wasn’t there.
So today, to give you a real-time example, it’s a conversation that I had half an hour ago, we’re training a model, figuring out when the conversation is over. So how do you know in an email that you don’t have the intent to continue the conversation. ‘Cause you don’t really say like: “Goodbye”, like you would at the end of a phone call, you just don’t respond.
Thank you for your email. Goodbye.
Exactly. So there’s lots of ways that email threads end, then we’re trying to figure out. Can we tell which ones are natural and which ones are effectively errors, where you were when you dropped the ball on something. It’s a fascinating problem. We have millions of messages to train on where you can see this. This ended and this didn’t, and then we’ve got to figure out, how do you know if it was intentional or not.
Then by spending the time to create a model like that, we can deploy that against everything else and milliseconds or fractions of milliseconds. So it’s really easy once you go through the heavy lifting of doing the training, doing the modeling, deploying, it is just something that just works, and it’s really cool to see these just random things. We have an idea, and then, a couple of days later, we’ll know if it’s helpful or not. So obviously it will be dog food at all in our own accounts.
Then we’ll see like: well that’s actually somebody that I really needed to talk to. That’s somebody that the system correctly classified as this type of person. In this mode, that’s absolutely somebody I need to talk to. So for example, one of the first ones we did is poker buddies. So people that the assistant tells who’s an actual friend that I would play poker with. And It was so close to over 80% accurate in who had got. It was really amazing because there were people there that somebody would make an introduction to me years ago saying: David likes to play poker. So keep them in the loop, and I just totally forgot. Didn’t do it. Then now, a couple of years later, it’s like a great reason to reach back out.
Now I feel like a bit of an asshole for having not done it, but the system can actually surface those things. So imagine if I’d had it back then and how many other missed opportunities effectively, and others like failures to respond or follow up or whatever it might be, there are hundreds of them. So when we build a model for these kinds of things It’s always really eye opening to see my own data come back to me in this format where it’s like: here are the things you should have done. Damn, it should have done that.
A lot of these great tools on the market that have solved for executive assistant tasks, such as ThoughtXAI, which was acquired by Visibo and Calendly, where we automate the calendar management through automations and triggers. So they’re solving for the executive assistant market. I get the sense with what you’re building for Strata, you’re solving for the chief of staff market. Basically Strata will be my own personal chief of staff to say: David, did you do this project? Did you follow up with this person? Keeping me in check so that I’m being accountable. What do you think about that chief of staff?
It’s a combination of things. So, it’s definitely chief of staff in that way, but, arguably, it’s more like a social secretary. So it’s like helping organize the most important relationships you have. So for example, if you’re traveling to Chicago, who should you reach out to there? Because I’ve started heuristics, so obviously people that live there, fine. Second, people you met last time you were there, fine. Third, people you’ve talked about meeting up with in Chicago. Maybe you will remember that maybe you have a super memory where you’re not limited by only 150 relationships and you can actually classify all minus like 30,000 people.
But I can’t. I have no idea who I may have talked to about meeting up in Chicago, but the system does. So then I can just configure it and go to Chicago. Here’s 30 people. I guess. Yes, no, yes. Great. And it’s not like you send out a mass email or anything like that. It’s just, then you get 30 one-on-one messages that it makes us super easy,
Like pre-written and then I can tweak each one. By the way, let’s meet up for a coffee. Let’s meet up for lunch. Let’s do this, let’s do that, and it just streamlines it. So I imagine you have a person and you’re talking to that person and the person gives you suggestions like: David, do you want to do that? Sure. Do you want to reach out to this person? Should you send this or should I say that? Then you basically just give your feedback. Yes, No. Maybe say this, maybe say that, and the system will handle everything else.
So chief of staff is a thing, but the much bigger part of it is really giving you recommendations on, basically, improving your relationships. I was staying in touch, but really around the context of what you’re doing. So one of the features that we’re playing around with, like the travel one, is a real one. So, traveling somewhere I can get a list of everybody should reach out very easily with no data entry for me, so I can obviously change it.
But the starting point is going to be a really solid list of people in Chicago, in LA and wherever. So that if I’m going there it’ll track all those things for me. And then, there’s some interesting, more advanced features that I don’t know if they’re ever well-built, but it’s really cool. So imagine if you get an airline reservation for a trip to LA, and then there’s a list of people in LA, but imagine if they are also Strata users. The possibilities there are actually even bigger because now it doesn’t even necessarily take me and you to email about something like that.
The systems can just coordinate, and seamlessly do all these things that would take too long. Even if we both have assistants that are doing it for us, there’s still so much back and forth that has to happen on scheduling and coordinating things on like pretty trivial tasks that if both people have Strata, it just goes away. So like, in theory, we’re not doing this yet and probably never will, but it’s really cool. If the system knows we’re both going to be in the same place at the same time. That could be a nice little trigger, and that’s a little nudge saying: you’re both going to be in London next week. You want to try to meet up?
As a startup, you’re right, Thor, that you have to prioritize different technology and feature requests that you build, and that sounds incredible. You and I being at the same conference in London or LA and just knowing it, but we will see, of course, how that may become a reality in the future. But thinking more about where you are today and where you’re going. Can you tease for our audience a little bit more on some of the product roadmap or some of the next steps that you’re seeing as you continue to grow and scale?
Absolutely. So we have a few products that we launched: the recommendations where you get three recommendations every week, plus memes and so corporate communication seems to be working. So that’s live now called Reconnect. So definitely go to Straddled that CC and sign up for that. Then we’re going to be launching the broader platform that I’m talking about that has all these integrated triggers, and nudges, and juristics, and patterns like travel, list building, list sharing, all those things that I suspect just about everybody who’s listening to this does right now, and it’d be great to hear feedback.
It’d be great to like, as we roll it out in the coming months. It’ll be a limited release at first, but obviously anybody that reaches out says they heard about it here we’ll get priority access.
Excellent. I cannot wait to play with the product. It’s at Strata, STRATA.CC. Today’s episode of HumAIn we’ve been featuring Thor Ernstsson, the founder of Feedback Loop and the CEO and founder of Strata. Thor, Thanks so much for joining us on the show.
Thanks for having me, again. It was great. I enjoyed it.
Thank you for listening to this episode of the HumAIn Podcast. Did the episode measure up to your thoughts and ML and AI, data science, developer tools and technical education. Share your thoughts with me at humainpodcast.com/contact. Remember to share this episode with a friend, subscribe and leave a review, and listen for more episodes of HumAIn.