Episode Transcript
[00:00:09] Speaker A: Hey everybody. Today I'm going to publish a conversation I just had with the three founders of Arist. Arist is a fascinating pioneer in the corporate training industry. Founded by Michael Yaffe, Ryan Laverty and Maxine Anderson, three amazing entrepreneurs. Originally funded by Ycommodator, they have built what I consider to be the state of the art AI based, next generation training system for micro learning, mobile learning, learning in the flow of work and compressed learning in an amazing fashion. And what Arris does is deliver small bytes of learning on the platform of your choice, spaced over time, designed in a scientific way to take large amounts of content compliance information, regulatory process documentation. We use it for the JB Academy and give it to you in small chunks as you can consume it. The uptake and the adoption rate of arist learning is very, very high. Some of the world's largest companies are now using it for extremely high value, high impact training programs and it can be generated almost entirely by AI. And these guys are wicked smart. They're working on a lot of cool stuff and they are going to be, and they really are already one of the leaders in what I call the autonomous learning platforms. So here's the interview and I'll do a quick wrap up at the end.
Hello everybody. Today we have a very interesting podcast with the three founders of Arist, one of the most sophisticated advanced micro learning or other AI based learning platforms. I'll let them tell you more about it. Ryan Laverty, Maxine Anderson, Michael Yaffe. We have been working with these guys directly in our academy. We've had direct experience with their content. Those of you that have taken courses on your mobile, courses on the JBA, you've seen how the technology works. Thank you guys for taking some time. Since the three of them are such well integrated management team, I'm going to be asking them questions and they will decide who answers. But I think let's start at the beginning. Tell me a little bit about the formation of the company, what you guys set out to do and why you started it.
[00:02:41] Speaker B: Yeah, sure thing. So Aris is a bit of an unusual background for a learning technology company. We started initially as a way to deliver learning in war zones. So essentially, Ryan Maxstein and I were all focused on delivering learning in hard to reach areas.
And I was working with some students in the amenity conflict zone, trying to figure out how do you deliver learning in a place with no Internet access and no laptop access. What we started realizing was that everybody had access to a phone. If we could figure out a way to deliver learning by text message and WhatsApp, we could make learning dramatically more accessible. So essentially, about five years ago now, we created the first text message course, working with a few professors in Boston. And what we started realizing was that by delivering learning via text message, a few things happened. One is learning became way more accessible. You didn't need an app or a laptop or an Internet connection to learn. And two, learning came right to you. It was one click access, super frictionless. And three, we started seeing dramatically more adoption behavior change outcomes, far higher knowledge retention outcomes. And we sort of, by accident, started to realize that this model that we had delivered for accessible learning was actually far more effective from a behavioral point of view. Fast forward to today, we work with about 10% of the Fortune 500 helping them deliver learning fully in the flow of teams. Slack text message and WhatsApp.
[00:03:55] Speaker A: So the original target was a phone, and now you're just seeing it as message based learning on any platform, right? Is that the idea?
[00:04:03] Speaker B: Exactly, yeah. The goal is to make learning deeply conversational, hyper personalized to each individual, and as embedded into your existing day to day life as possible.
[00:04:13] Speaker A: So I think that's a great. Obviously, it's a hugely powerful idea. I've talked about learning the flow for a long time. Before it was mostly just an idea when I started thinking about it. It seems to me there are at least three problems, probably more than that. One is, what content do you deliver in this format? The second is, how do you deliver it in a way that people can get it without having to sit on their device and, you know, kind of hang on there all the time and ruin their experience? And the third is, how do you develop the content more quickly in a digestible form? I think you guys have opinions on all three of those. Let's start at the beginning. How do you, what is the content strategy for a message based learning experience?
[00:05:03] Speaker C: Yeah, so a few different things that we focus on. The first and kind of the premise we started with was that arist courses are really good for things where simple concepts plus well practiced equals mastery. And so if you think about the types of things where you need to practice them every day, mastery comes from, hey, I understand this, but I'm constantly doing it all the time. Great examples include management, sales, any sort of broad communication use cases, because a lot of learning is communication. Something like a product rollout or an onboarding. Right. Arist is really, really good for that. What we found over the years that's actually really powerful is that we can actually get a lot more in depth to content than folks oftentimes think. And so a few good examples are, you know, if I'm teaching someone how to be a great manager, we can teach them how to give and receive feedback. We can teach them how to know good tips about a product. If I'm working in the medical space, we can get really in depth about a particular device or tool. One arist lesson only takes me about, you know, five minutes to go through, but it actually can cover about, you know, 20 to 30 minutes of video content. And so over a, you know, five or ten lesson course, we actually can cover a lot. The only things we don't use arist for are things that are highly, highly technical, highly specialist, where having kind of a one on one, immediate feedback interaction would be best. But for most other things, we can either partly or wholly put them into arist courses.
[00:06:21] Speaker A: Wait, hang on a sec. Ryan, you said something didn't make sense to me. So you had more video content than the person consumes, or can you go through those metrics again?
[00:06:32] Speaker C: Yeah. So just so about five to an Aris lesson takes you about five to seven minutes. And we found it's roughly equivalent to watching about a 15 to 30 minutes video in terms of how much content can be covered.
[00:06:46] Speaker A: How do you explain that?
[00:06:48] Speaker C: Yeah, absolutely. So if you. So if, you know, let's say you're an employee at work, you watch a 15 to 30 minutes video or PowerPoint presentation, there's a certain amount of content that's covered. One, the learner is passive. Two, we speak a lot slower than we readdeze. And then three, there's a lot of repetition to things because it's a one time interaction. But if we break those concepts down into bite sized pieces, make them all readable, interactable, I come to them at my own time, and then we space them out over time, it's actually a much more effective way for people to consume content. We can consume a lot more content in a lot less time.
[00:07:21] Speaker A: So there was actually a very shrewd instructional design idea behind this to take a, what most people would consider a 30 minutes, maybe instructor led thing and break it into these smaller chunks. Right. This is something that isn't obvious how to do.
[00:07:38] Speaker C: Exactly. So actually, that's why Maxine can probably jump in here, but that's why we actually use AI to break down a lot of the content.
[00:07:46] Speaker A: Okay, Maxine, let's talk about that.
So I've got a 30 minutes video or a 1 hour video of somebody talking about some important safety thing or management thing or process thing that somebody needs to learn. How do you break that down into small pieces?
[00:08:03] Speaker D: So what we do is over time, we've refined a course model that we have found to drive behavior change effectively. So we use that course model and thousands of courses to train LLMs. Initially chat GPT's model we actually released this AI content creation tool four months after they released their GPT model for public use. And what we did is we trained the model on our most effective courses for behavior change against different training type areas. So upskilling, or let's say it's like awareness training or reinforcement training, there's different models that you might use to create an effective course over text based on the type of training. And so our AI is really intelligent in creating courses for those types of training that ARIs is best used for. And so if you feed it content in some information, let's say in your example, a video that can be long form, or let's say you share a PowerPoint or a PDF about a product that you want to train sales reps on, let's say it will consider that information and parse out what's most important, and then it will put that into a message based course. It'll take the most important content from that. Consider our course model. Put it into a course into our product. It takes a few minutes to review, maybe add some edits or content text that the customer wants to do, and then they can send that out over our product really easily. So that's essentially the process and what our tool does. Because we launched it so early on, we've been able to iterate really quickly and we still get feedback from our customers. It's far ahead of most content creation tools. So speed was a huge advantage for us there in launching that quickly and iterating since then, and we've had awesome results.
[00:09:47] Speaker A: Wow. Okay. So you basically fed the system the current instructional design that you had been using, or the courses you had been using, and taught it how to make those kinds of decisions on its own. Is that general way of thinking about it?
[00:10:01] Speaker D: Correct. It's essentially an Aris course building expert.
[00:10:04] Speaker A: I know you've used it for some of our courses. Roughly what's the compression ratio like? Or do you tell it how long you want the resulting course to be?
[00:10:14] Speaker D: We usually will instruct it. It also depends on the training type. So, for example, if you do more reinforcement training, it might be eight lessons long, but those are spaced out over time. And so that would be different than an upskilling course where you're providing initial introductory information, let's say like the foundations of AI use case, that might be a five lesson course and so it can intuit how long to make the course or it can be instructive.
[00:10:39] Speaker A: How does it decide the timing of the spacing of the different modules?
[00:10:44] Speaker D: That's based on our course, our learning model, what we found to be most effective for behavior change. So, for example, reinforcement after even, let's say, an upskilling course, we believe, let's say five days after, there should be questions that should be included to ensure that a learner is actually retaining what they learned, and then you might space it out even further. So it depends on the use case, but we've trained on different training types so that it knows what to do. Onboarding is a great example, right. There are certain checkpoints in onboarding that you'd want to include 30, 60, 90 days. It also will interact with our automation capabilities. So, for example, if there's activity in other systems, Aris works well when it's integrated with other systems like an HCM. So if we have indicators that someone's role has changed or someone was onboarded, we know their start date. We can use that as well to know when to space out learning and so we'll interact with that. Aris works best when it's paired with our AI and automation so that people get the right learning at the right time.
[00:11:44] Speaker A: So what happens to the instructional design team? Do they go home for a while and let the system do it by themselves or what role do they play?
[00:11:54] Speaker B: Yeah, so it's interesting. I think one of the things that we get really excited about is essentially elevating the instructional design team to be a lot more strategic. What I mean by that is that we had a client earlier this week that had 350 pages of Medicare documentation that they needed to convert into a course. Usually that would take the instructional design team 80 to 100 hours to build, you know, adjust, translate, et cetera, at least, right? So this course would not get launched for weeks or months. Essentially, this client was able to upload these 350 pages of documentation into arIs. The instructional design team essentially said, hey, we need this course for these eight different audiences in these ten different languages. And we needed to focus on these specific outcomes based on what the business is telling us and within about, I think it was like eight or nine minutes. All of the courses were done and ready to go and launched directly in the flow of work to the right audiences.
It takes the instructional design team from simply an execution function to a strategic function where you can really, they can essentially work with the business to solve business problems, work with the AI to actually create solutions that deeply solve business problems fast, and then use the data to then recommend additional iterations and adjustments from there. It's kind of what every instructional design team has always wanted to do, but has never really been able to do. And I think that's what we get really, really excited about, is I think we as an industry have not kind of been honest about the fact that so much instructional design time is really just spent in PowerPoint and in tools like articulate building programs that really aren't very strategic and not very customized and personalized.
[00:13:27] Speaker A: So what kinds of feedback do you get from customers about their interest in customizing it for their audience?
Generally believe that whatever it is you guys are generating has to be uniquely designed for their company, their audience, their demographics, their users. Or have you found that to be not necessary?
[00:13:47] Speaker C: I think it's a little bit of both. And so I think that's the beauty of the sidekick tool, is that you can upload your company's sales handbook, product specs, employee manual, pretty much anything, and it will customize that to you. I think what we've found, like Maxine mentioned, is that the AI runs on a few of these fundamental models. So one, it's been trained on, at this point, probably hundreds of thousands of arist courses. Two, it's intaking prompts from someone when they upload this content, audience objective topic, et cetera. And then three, it can either be used with no uploaded content or with the content that company has uploaded. And so the last piece of it actually is that, excuse me. A lot of companies will just output the AIH output and say, hey, this is good to go. We actually output it in a fully editable format and watermark everything that's made by AI until a subject matter expert goes through it. And we have a full course library that also has been built by our subject matter experts with similar content. We believe that most content is either out of the box, it's a video library, you have to use it. It's not customized, or it's PowerPoint articulate. I'm going to take two weeks to customize it. We kind of sit in the middle where we're going to build you a 90% done version really fast and then either use AI or leave it up to you to go kind of company ify it to the way that you want.
[00:15:04] Speaker A: So it sounds like we could use Arist to build new content based on documentation. Or videos or other sources. What about existing courses? Do people take Arist and add it to long legacy content? They already have to just simplify it and get it out there in a more compelling way?
[00:15:22] Speaker B: Yeah, 100%. I think for us, what we found is that many of our clients have taken long compliance courses or leadership development programs or existing in depth trainings. You can essentially just upload the course, whether it's the scorm file or the transcript or the slides into ARis. Click a button and ARis will dramatically simplify the experience and make it way easier to edit, way easier to deliver.
So it's been a very helpful way for organizations to gradually modernize their existing experiences as well.
[00:15:53] Speaker A: It blows my mind, you guys, what you've done. I mean, I really haven't seen anything like it anywhere. Can you give me an example who's a client that you can talk about that's using it just so people listening can get a sense of the scale and the potential here?
[00:16:06] Speaker B: Yeah, sure thing. So two clients that actually, I'm going to talk about, three clients that I get really excited.
[00:16:11] Speaker A: Okay.
[00:16:12] Speaker B: So one of my favorite examples is the state of California. The state of California both has a large frontline population internally, but also has, you know, they are responsible, in part, for the safety and wellbeing of hundreds of thousands of field workers across the state. Many people don't know this, but the state of California actually has an office dedicated to the safety and wellbeing of farm workers. Right. And field teams across the state. And one of the biggest challenges that they've had is, you know, how do you get training to these people given the limitations that they have, right. Limited devices, limited Internet access. And so they've essentially, for the past few years, have been using Aris to create a series of safety and disaster preparedness courses that they can push to these field teams all across the state in multiple different languages entirely by WhatsApp and text message. Aris is actually the only way to compliantly reach frontline teams on personal devices. There's some sort of magic behind the scenes that makes it so that employees can learn on their personal device while still maintaining wage hour compliance and without any costs to them or their employer. And so it's just been a game changer. Right. And so now, for the first time ever, you have folks in the Central Valley field teams learning on their personal device in a way that's designed just for them and their knowledge gaps. So that's sort of one example and another. One of my favorite examples is our work with HP and where we're starting to do a lot of scaled training across the enterprise, helping them build skills way, way faster in a much more scalable way, directly in teams chat. And I think it's been remarkable to see both how much adoption has increased when you push learning to the right people, in the right place, in the right time, but also real behavior change impact in a very short period. I think today, across enterprises that we work with, 92% of people prefer learning via Eris to any e learning medium.
And it's just, I think, telling people want to learn this way, they're asking for it. Enterprises just need to adjust.
[00:18:12] Speaker A: So a little bit more on the company and the platform, and then we'll kind of talk about the market and wrap it up. Now, on the back end of this, you guys are tracking all sorts of things, and there's a whole back end system. What can the back office folks, you know, look at? And how do we, you know, get access to all the information about how this content's being utilized or not utilized?
[00:18:33] Speaker D: Sure, I can jump in here. So we have pretty in depth tracking in the Aris application. We look at a few things. One is engagement metrics, which are traditional learning metrics that you can see, for example, are people completing courses. How are they engaging with it, with Aris, with each course, you have many touch points with each learner. Right. So you're able to ask each course, ask many questions. We collect a lot of information, how the employee feels, for example, about their manager, let's say if they're onboarding, they might share some of that information. And so all of that response data is tracked in Eris and can be looked at where we look at a few indicators as potential indicators of behavior change. So one of them is confidence lift. So from the start of a course to the end of a course or learning experience, how to have someone's confidence in what they're learning changed.
And that can be really insightful. And we actually use that information to tell the people who the customers use, Aris and the admins, if that learner might need more training on that topic versus moving on to the next thing in that upskilling journey, let's say. Right. We also have some embedded sort of adapted learning in. Let's practice this concept more in the learning experience. If that confidence has not changed to the amount that we'd expect it. The other thing we look at is application of learning. Right? So we're really focused on behavior change. So let's say you're taking a public speaking course, and we're prompting you to take certain actions when you're out working with other people on the job, and we're having you practice something in a meeting to get better. Public speaking might ask you the next day, how did that go? Did you apply this concept? And they might say, well, I didn't. And it's like, okay, why explain a little more about that? And we also measure that application in the product in terms of synthesizing that information.
A few things. One is AI insights, so we can use insights against learner information, how they're responding to say if there's competency gaps, predictive analytics for those administrators. But we can also send all that data back to a data lake or let's say an lms very easily as well, so that it can be looked at against other learning data too.
[00:20:40] Speaker A: So, wow. So it really is an adaptive system. I noticed in the course that we just launched, it was asking me questions about how confident I felt about the topics of AI. So those kinds of questions are leading the system to give me different forms of content next, right?
[00:20:55] Speaker D: Yeah. So right now it informs administrator mostly, and we're incorporating more adaptiveness into the actual learning experience as well as AI gets better. One thing is that Aihdena still does hallucinate sometimes that's slowly getting better and going away, but because of that, we are careful about how we incorporate AI into the learning experience from an adaptive point of view.
[00:21:17] Speaker A: Okay, well, listen, you guys, this is really fascinating. You've probably read the piece I've put out on autonomous learning and how fast I think the learning industry is going to change. I think you're really one of the pioneers in this space. One more quick question for the three of you. Whoever wants to answer this, or all of you, we have been talking about reusable learning objects and learning on demand, and adaptive learning and micro learning and learning in the flow of work for at least ten years, and no one seems to have really nailed it quite like you guys have. What did you, why do you think what you've been able to do is so unique?
[00:21:57] Speaker C: Yeah, I think the, it's a great question. I think the biggest thing for us is that if you go look at pretty much every type of e learning, micro learning, learning in the flow of work, most people mean learning in the flow of work to mean, hey, it's kind of short and I can access it somewhere at some point if I go through an obstacle course first, I think for us, what's exciting, the thing we always remind folks of is the average person checks a learning management system or HRS system once every six weeks. They check Microsoft Teams or SMS or WhatsApp messages once every six minutes. And so that's a 1600 x attention difference in how much time we're spending in Microsoft Teams or SMS or WhatsApp versus a tool like a learning management system. I think what's really exciting about that, it actually took us about three, three and a half years, technically speaking, to be able to get the course model right and the delivery mechanism right. We've seen a lot of kind of potential copycats pop up and then die out because it is actually really complex to send these things, send them in a compliant way, send them in a way that creates a really compelling learner experience. And for us, that's the biggest thing is meeting people where they are. I think in addition to meeting people where they are, there's also, we've talked about the right content in the right place and the right time. I think that right content, a lot of that we can build with AI now. And that was always too difficult and had too much overhead to personalize to the individual or to make it actually compelling. The right place being SMS or Microsoft Teams, which we seem to be the only one who's really technically cracked that, and then the right time being the most exciting thing coming up with AI, because now we can link to an HRS system, a CRM, like Salesforce, over 100 other tools, and decide, okay, Josh wants to learn about having a career conversation. He's got a meeting coming up. Here's the right time to prompt that. And I think really, that's where the space is going, is a place where we can do that. But right place in delivering an SMS and teams is really what has been a big unlock.
[00:23:54] Speaker A: Well, I mean, the most obvious question, at least to me, is, is this going to appear in the copilot in Galileo, in all of these chat bots? Is that the next stage here?
[00:24:05] Speaker B: I think so. I think we may end up in a situation where Aris lessons, you know, because they are designed in such an effective and research direct way, will end up in whatever tools people use. Right. Whether that's the chatbot that they use on a regular basis, et cetera. I think one of the things that we've realized is that in the future, and the future is going to be way closer than we expect. Right. But in the future, organizations will have what we call a poll platform. Right. So platform where learners can go in and pull knowledge. I think Galileo is a terrific example of that. But organizations also need a push platform, a platform that proactively pushes learning in moments of need. Because the reality is that most learners don't know what they don't know. And it's the responsibility of organizations and organizational learning teams to actually push learning in critical moments of need and identify what those moments of need are. I think that's why we think enterprise L and D teams won't go away with these generative chatbots. On the contrary, that people will start to realize that push learning matters a lot more. And that's really what we're here to do, is to help organizations push learning in critical moments. So need and use internal data. And I think to Ryan's .1 of the only reasons why, one of the other sort of core reasons why Aris is able to do this is because we've really rethought this from the ground up. When you deliver learning in this format, everything changes from the way that you create content to the way that you deliver, to the way that you analyze. And the end result is you actually drive performance impact and you're able to have far, far faster delivery, analytics and far better learner outcomes. But it requires sort of this ground up approach that we've taken over the past few years.
[00:25:34] Speaker A: Well, you guys, I want to thank you so much for sharing this with the audience. I am just completely astounded at what you guys do. I want to thank you for your partnership with us and really congratulate you on building something that's really powerful, really unique, and really, to me, just designed for the future of where all of this is going, for learners, for producers of content, for organizations as a whole, for our daily lives. So congratulations and thank you so much for joining us today.