Episode 36: Reimagining Instructional Design with AI: A Conversation with Ethan Webb of Mindsmith

In this episode of Demystifying Instructional Design, guest host Nicole Taylor sits down with Ethan Webb, CEO and co-founder of Mindsmith, an AI-powered course authoring tool. They explore how AI is reshaping the instructional design workflow, the evolution of Mindsmith from a college startup to a leading corporate learning platform, and what the future holds for AI-driven learning. From dynamic SCORM innovations to designing for accessibility and real-time collaboration, Ethan shares insights on building edtech with purpose, embracing disruption, and empowering instructional designers to work smarter—not harder.

Transcript

Nicole: [00:00:00] Welcome to Demystifying Instructional Design. I'm Nicole Taylor, your guest host for today's episode. I'm currently a student of Rebecca Hogue's in the Instructional Design Master's program at UMass Boston. I've worked as an instructional designer in higher education for the past seven years. 

Today I'm excited to speak with Ethan Webb, the founder and CEO of Mindsmith, an AI powered course authoring tool.

We'll be discussing how AI is shaping instructional design, the evolution of Mindsmith, and what the future holds for AI driven learning tools. 

Hi, Ethan. Welcome to the podcast. To start today's conversation, could you introduce yourself to our listeners? 

Ethan: Sure. Yeah. You kind of already introduced me from a professional point of view.

So Yeah. I'm one of the co-founders of Mindsmith. We're an AI native e-learning authoring tool. Aside from that, I mean that takes up a lot of my time being a co-founder of a startup, but I also love to ski. We're based here in Utah, so my [00:01:00] wife and I love skiing. We love rock climbing. We have a 2-year-old, so that keeps us on our toes. 

And that's pretty much it about me. I kind of fell into the learning space by accident as I think a lot of people have. But yeah, I'm excited to be here. 

Nicole: Your first interview on this podcast was two years ago. Since then, what has changed? Could you give us an updated overview of Mindsmith?

Ethan: Yeah. When we first came on this podcast, I think we had just started building, it was myself and my co-founder. We were actually still in college. We were building this company while students and so the product itself has come a really long way. I mean we had just started dabbling into AI, I think when we first started talking. We had kind of launched our first, in startups we call it an MVP, a minimum viable product. We had just launched and it was very emphasis on the minimum I think when we started on here, [00:02:00] especially looking backward. Anyway, so we've really leaned into how is AI changing the workflow of the learning designer and how can we design an authoring tool around that changing workflow has really been our focus. We've made a few other pivots. We are really focused in on mostly corporate learning. Most of our customers are in the corporate space. We're also more of a generalized tool rather than a microlearning tool. Kind of when we first started, we wanted to stick to micro learning, but since then we've decided we don't really want to pigeonhole ourselves into necessarily one type of learning. So we're a generalized authoring tool, AI native. We do a lot of really cool things, but that's kind of the core of who we are now. 

Nicole: When you first launched Mindsmith, did you always envision it as an AI powered course authoring tool, or is that just the natural progression of things?

Ethan: Yeah, that's a good question, and the answer's yes and no. [00:03:00] So what happened was when we first started building, when I first met Zach, he had been in the e-learning industry for a while. And he realized that authoring tools as an industry hadn't been disrupted for 20 years, right? And so he is like, there are some really cool things we can do. At the same time, this was in the summer of 22, there were some other startups starting to gain popularity in the AI space and finding huge efficiency increases. The most popular at the time was called Jasper. It was an AI service for marketers, and basically it was a wrapper on top of GPT-3. So it was a fine tuned GPT-3 model, which was the precursor to Chat GPT. And so we did some experiments at that summer of how could we maybe use AI, this new generative AI LLM thing to help learning designers move faster and we kind of determined that it was too expensive.

It was really expensive back then, you had to fine tune a model and it was just really difficult to do so it just didn't make sense. Well, a few months later, after we had [00:04:00] done our initial experiments, Chat GPT dropped, and all of a sudden, the out of the box models were outperforming these models that people had spent tons and tons of money fine tuning and training, and it was 10 times cheaper.

And so we're like, oh, that just makes sense. So we built a very basic AI feature on an afternoon. My co-founder just threw up something. It was basically a useless feature where you type in the title of your lesson and we just generate the content for you. Which , obviously when you're creating custom content, that is not helpful.

But we shipped this very basic feature just to see how people are reacting to it and it was incredible. We pushed it out to Reddit and LinkedIn and some of the other forums we were a part of and it like 10 X star new users and seven X star daily actives. And it was really clear that people were thinking about how is AI going to change the workflow of the learning designer?

And so that's where we really leaned into since. How can we design this authoring tool around this new workflow? And so I guess [00:05:00] from the early, early days, I would say yes, it was, but we had to time the market right for it to make sense to actually build AI as a core part of the product.

But as soon as Chat GPT dropped, it was December of 22, we were like, yeah, a hundred percent. It's time to go. So yes, the answer I would say yes. 

Nicole: How does Mindsmith leverage AI in course creation? And most importantly, how does it ensure user privacy and data security?

Ethan: Yeah, data security is very straightforward. Basically we only use out of the box models and we use their enterprise APIs. So none of the models we use whether it's from open AI or Google, they don't train on anything we pass to them and we don't train any models ourselves. So nothing you upload to Mindsmith gets used to train any models ever.

So it's actually very straightforward from a data security point of view. As far as how we do it, there's a really complicated engine in the background where we have to give the AI the right things at the [00:06:00] right time, especially when we're dealing with a lot of documents and a lot of content and the AI's having to contextualize all of it and help build the product. So there's a really complicated answer to that that I actually couldn't fully answer for you. My co-founder would maybe pull out a chart and we could do a deep dive into it. But we want to give the AI the right stuff, and we want to kind of guide you through that process of creating content with AI.

Nicole: Switching to a macro level overview, what would you say are Mindsmith's core philosophies? 

Ethan: Yeah, so the tool itself have a few core philosophies and then we as a company have a few core philosophies. As a company, some of those philosophies are from the very beginning we kind of realized that AI is kind of an arms race, so similar to how computing in the very beginning days of computing ended up being an arms race. And actually the computer components and the compute [00:07:00] itself ended up being kind of commoditized. We think AI is going in a similar direction where intelligence or artificial intelligence kind of asymptotes at free almost.

And we can kind of make bets assuming that future models will get smarter and will get better. So similar to Moore's Law where every seven years you have a doubling of the transistors you can fit on a silicone chip. It's a similar thing. We believe in AI that this arms race will produce.

So that's a core philosophy. It's that the out of the box models will continue to get better, stronger, more intelligent, and actually cheaper. So that's one of our philosophies and it's been really helpful for us. Some other companies will spend a lot of time fine tuning a model, which is really expensive or training their own model.

Which are both really expensive to get a 10% or maybe 20% efficiency boost, and then three months later a new model will make that irrelevant. So it's been a really good philosophy that for us held true. It also allows us to prioritize the long term [00:08:00] in terms of the the features that we build.

So we don't want to go after what's hot or what's new. But we wanna be purposeful about the AI that we're implementing into the product itself and the relationships that we're forming. So those are kind of the core philosophies of the company. I would say from a tool perspective, we're kind of agnostic about how much AI will, quote unquote, replace the instructional designer.

I don't think anyone can or even should try to postulate that. But we do think that the creation of e-learning content will become better and cheaper and faster as AI gets better and cheaper and faster and smarter. So those are some of our philosophies. 

Nicole: For our listeners who may not be familiar with the term, could you talk about what out of the box AI models are?

Ethan: Yeah, it's pretty straightforward. We don't train the model. So every time in Mindsmith, when [00:09:00] you're prompting the AI, it's a new prompt basically. So we just use the APIs that Open AI has and we provide it with context. And those context windows have gotten really large over time and as the AIs have gotten smarter.

So yeah, out of the box is just we don't train. 

Nicole: Now that Mindsmith has been in the market for a couple of years, what key lessons have you personally learned about running an ed tech startup? 

Ethan: Yeah. A lot of lessons. When we first started building Mindsmith, I didn't have much expertise in the L and D industry and I've really had to fight hard to gain that.

So talking to instructional designers, talking to learning leaders, reading about instructional theory and learning about instructional design has been a huge learning curve for me. As I've learned it, we've been able to incorporate aspects of learning design into the product itself. And that has been really helpful.

And as we've learned [00:10:00] more about the workflow, the learning designer, we've been able to incorporate that into really cool parts of the product. So that's been number one just learning about the industry, learning how people learn, learning some of the pitfalls of e-learning and thinking through how can we leverage AI to maybe fill some of these gaps to make the learning impactful and meaningful and good. So that's been kind of one, I would say also, I've had to learn how to do B2B sales, which I think has helped me in other parts of my life. When I started Mindsmith, the only sales experience I had was door to door sales, which is very different.

 I am knocking on a door and I'm closing the deal right there. Versus B2B, you have to be a lot more careful, a lot more considerate. It's much more longer term, it's much more about relationship building. So I've had to learn how to build those relationships. I've also had to learn how to be assertive.

 When we first started Mindsmith, I was giving away discounts like crazy, you know? I was always like, anything to close the deal, bend over backwards for anyone, even if it [00:11:00] was bad for the long term of the company, even the long term of my relationship with the customer.

I was like anything to get the deal done and now I can kind of push back a little bit and be like, I can actually teach you about how AI can be used in L and D and, we can tell you how you should buy this product instead of always bending over backwards to the whims of the people we're working with. I mean, we're still very customer centric. Any of our customers you'll talk to and we respond to customer needs very quickly, but I have learned to kind of hold my ground on certain things and it's been really helpful I think.

Nicole: That segues nicely into my next question to shift gears and talk about the user experience. What kind of customer support can users expect? 

Ethan: Oh yeah, our customer support is great. We're an early stage startup, which means you get direct access to myself, my co-founder, and our founding engineers.

So we have a live chat on [00:12:00] our product where you can chat with us directly. And we tend to respond within, usually within an hour, sometimes within a minute. And at the very longest, within two business days. And then we fix bugs like crazy. So here at Mindsmith, we ship code every day.

 There's an update to the software, whether it's a totally new feature or bugs that we're fixing. So it's usually by the next day the bug is fixed , as far as if you're experiencing technical customer support. So yeah, very hands-on I would say in terms of customer support. 

Nicole: How would a user export content out of Mindsmith into something like an LMS?

Ethan: One cool thing about Mindsmith, we invented this thing that we call Dynamic SCORM, and we've actually seen other authoring tools start to offer, which has been really fun where basically the SCORM file is a link back to Mindsmith wrapped in SCORM, which allows us to do some really cool things.

The [00:13:00] first is that, I don't usually mention this, but it's actually a huge thing. The file is tiny. So it will only take seconds to upload to the LMS. The other cool thing is that it's kind of in the name, but it's dynamic. So if you change something in Mindsmith, it updates in the LMS automatically.

So this is actually a huge operational headache for a lot of organizations, especially larger enterprises where if you wanna change something, you have to delete it from the LMS, often lose a lot of your tracking, change it in the authoring tool, re-download it, re-upload it, and it's just this whole long process.

But for us it's just a nice smooth connection between us and the LMS. You can also stage changes if you don't want to do 'em automatically. We can also do a lot of other cool things like, another huge headache, especially with multinational organizations, is managing SCORM versions for different languages.

So, in other authoring tools, you have to export 15 different SCORM files for each of their different languages. But in Mindsmith you can create a new version for each of the languages that are all saved within one SCORM file. So I could create the Spanish version. [00:14:00] We use DeepL to give you the first draft, so we translate it for you.

You go in, make any changes. And then you upload just the one file to the LMS and the file itself will read the device language of the learner and give them the version that matches their device language. So if their phone's in Spanish and you've created a Spanish version, it'll give 'em the Spanish version automatically.

We can also do other cool stuff, like we can track engagement metrics that don't typically get tracked in SCORM because SCORM is a very limited locked down standard. So if you want to see where people are dropping off in a lesson or how they're performing on individual assessment questions, stuff like that, we can surface that.

We can also do cool stuff like we can do AI grading and we can pass that grade back to the LMS. So we have a short answer tile where you can give the AI example correct and incorrect answers, and then it will grade the learner and provide that back. So, and then we can do other cool things in the future.

You know, if we want to do real time adaptive learning or if we want to provide some sort of AI role play situation, we can do all of that stuff with the dynamic SCORM. The [00:15:00] funny thing about dynamic SCORM is we actually built it originally on accident kind of as a workaround because SCORM is a huge headache to actually implement.

And we have it now the static SCORM, your typical SCORM package, but it's so annoying. And so, my co-founder was like, what if we just linked it back to us and wrapped it in SCORM? And that was our solution and it worked amazingly and we realized we can start to do some of these really cool dynamic things and our customers have loved it.

So it kind of happened by accident, but it's been a huge innovation in the e-learning industry 'cause SCORM is so ubiquitous. Lots of LMSs haven't adopted some of the other standards like LTI or an XAPI can be a real headache to implement. And so you get the ubiquity and ease of SCORM with all the benefits of other standards.

So that was a long answer to your short question, but it's a feature that's really exciting to me 'cause it's actually outside of the AI, but it's a really cool thing that we've done that our customers just go bananas over. They love it. 

Nicole: You touched a little bit on open-ended [00:16:00] questions.

What types of assessments and activities can you create in Mindsmith? 

Ethan: Yeah so, we have assessments and then we've got interactives. So assessments are very straightforward. It's multiple choice and short answer. That's all we have right now. I could see us getting into some other cool stuff, we can add grading for some of our other interactives.

But that's all we have for assessments right now. Interactives, we have all the interactives you would come to expect, like drag and drop sorting, drag and drop matching. We've got hot buttons. Our scenario feature is really cool one where you can actually generate a branching scenario with AI.

So you tell it this is the character that learners interacting with. Maybe it's an angry customer. This is my learner. They're a retail employee who has to overcome objections and I want you to follow a happy path and when they say the wrong thing, loop them back or something, and the AI will create a first draft of that.

And we have this really nice, if you haven't gone into Mindsmith yet, we have this really nice editing interface for these branching scenarios where if you've used Figma, FigJam or [00:17:00] Lucid, you can see the different paths that a learner might go down. We draw them out for you and you can edit the scenes natively in this kind of canvas structure.

Anyway, that was kind of tangential. We have branching scenarios and all of the interactives that you would need. 

Nicole: Switching over to accessibility. Accessibility is essential for inclusive learning. How does Mindsmith ensure its content is accessible to all learners? 

Ethan: Yeah we've designed accessibility, just baked it into the product as we've built.

So everything out of the box is WCAG AA accessible. So it's out of the box, you can tab through it. We've got screen reader accessible, all of that stuff. The one thing you can break is our themeing. We allow you to not do options that are within WCAG AA contrast ratio standards, but we'll yell at you.

We'll be like, hey, this is not within accessibility standards. Maybe consider switching your theme. And then we have a VPAT and all that stuff. But yeah, we wanna make sure all learners can [00:18:00] experience Mindsmith to its full potential. So we wanna make everything fully accessible. 

Nicole: In the course authoring tool market, how does Mindsmith compare to competitors such as Articulate Rise? 

Ethan: Oh yeah. That's kind of a spicy question. If you work at Rise and you're listening to this, don't be offended. No, I'm kidding. Compared to Rise, we switch over a lot of Rise customers. The biggest thing is the AI.

So Mindsmith is an AI native tool, and so every part of Mindsmith is thought through with an AI lens. It's kind of a classic disruption theory where Articulate has this very old code base and they have a product that people are actively using and switching the way that people use the product is actually very, very difficult, much more difficult for them than it is for us.

And so November of 23. They were like, we are gonna do AI stuff. They announced it at DevLearn or [00:19:00] something, and they're like, Q1 of 24, we're gonna have our AI features. And Q1 came and went, Q2 came and went, Q3 came and went. And then they're like, oh, now we're having an AI assistant.

And the assistant can convert a tile into a different tile type and we've had that for a year and a half. Our assistant can do all of that and more. Plus we have this entire workflow built around AI. So the way that we've thought about it is how do we help you formalize your AI usage for creating e-learning content instead of being ad hoc, instead of feeling like an add-on, really building it from the ground up.

The other thing is, of course I mentioned this earlier, but dynamic SCORM is actually a huge reason our customers buy from us. We actually built a Rise SCORM importer. So for our customers who are switching over, they just upload their Rise SCORM files and we'll turn them directly into Mindsmith courses to make it really easy to switch over.

And we've got a few other authoring tools in our backlog of switch over , that we're [00:20:00] building. So, the biggest thing, AI .Number two is dynamic SCORM. Number three is UX. So Mindsmith is a lot more intuitive. There have been a lot of changes in the way we think about building great products in the last decade or so, and Mindsmith has been able to take advantage of a lot of those things.

So you'll notice things are a lot more drag and drop. They're a lot more wizzywig. Our branching scenario feature is designed around Figma and Lucid and these just really nice new modern tools. And then we have other stuff, we have real time collaboration so you see the cursor of the person you're working with.

And that's actually been a huge unlock for a lot of our customers where now, people can work at the same time. Instead of having siloed instructional designers working individually, now we can collaborate between instructional designers or also bring other people to the table.

If you want to have a SME, a subject matter expert, in Mindsmith helping you create content, we can do that as well. So there are a bunch of things, but we are thinking through modern e-learning authoring. If you are [00:21:00] comparing the two tools, yeah, I invite you to just try both and decide on which one you like more.

Nicole: Looking to the future, where do you foresee Mindsmith going? Where would you like the company to go? 

Ethan: Yeah, this is an interesting one. So there are a lot of directions that we can go and I actually don't know fully the future of Mindsmith. I know what we're building for the next six months and I know some really cool things that we can build in the next year, year and a half, maybe two years.

I guess I can talk about a few of those kind of longer term things that we've envisioned. One is we would like to include AI for the entire workflow of the learning designer. So right now we come in after you've done your needs analysis, after you've collected your content from your subject matter experts, and maybe even have a vision about what you want the AI to produce. And then you jump in, you tell us your vision. You edit the storyboard, you generate the [00:22:00] content, you have an assistant to kind of edit the content and that's where it ends. But I could see us reaching or kind of thinking through other parts of that workflow.

So, one might be an AI quote unquote agent. I know that's a buzzword, but an AI assistant would maybe help you interview a subject matter expert or interview other stakeholders to gain a needs analysis because maybe e-learning actually is not what you should do to try to solve this problem.

So, kind of an AI on that first, and maybe collecting documentation, categorizing it, talking to people. Maybe it generates that first draft for you, kind of works in collaboration with the instructional designer. And then after you've released the lesson into the wild or shared it on your LMS, we'd love to have some sort of optimization on the other end.

So, an AI that looks at maybe some of the data analytics and is like, this is an anomaly. People are not performing very well in this assessment question and it can AB test maybe the question itself. So maybe the question's a bad question or it can AB [00:23:00] test the content. Maybe it's not being taught in the right way and it can do that in real time.

One of my observations as I've talked to learning leaders is that sometimes L and D feels like they don't have a seat at the table, and it's been traditionally really hard to prove ROI. And I think that that step can help with that quite a bit, where we kind of turn an L and D team into taking on almost like a product management function where they're doing AB tests, they're proving that what they're sharing works instead of this was based off of these 15 theories, and I'm pretty sure it's going to work because we're using learning theory. No, what is it actually doing?

How is it actually impacting people? So would love to do AI on the other side. So that's kind of the first answer I would be interested to see how we can build AI into other parts. The other stuff is bringing AI to the learner directly. There are a lot of tools doing some of this stuff, AI role plays, adaptive [00:24:00] learning.

The biggest hurdle is of course, hallucinations if you're generating content in real time for your learner. So that's a problem that's really exciting and fun and I think that we could solve it as well. So those are some of the grand visions. Maybe we do build an L and D employee.

 If the AI gets smart enough and we free up L and D to do more consulting stuff or take on more higher level vision stuff and we are the e-learning employee . I don't know. But that's kind of some of the things we're thinking about. There's a lot more, but that's some of the things that are exciting to me.

Nicole: This has all been very insightful. To tie into my final question, how do you see AI influencing the future of instructional design in general?

Ethan: Yeah. I wish I had just a solid answer for this. Similar to what I was just talking about, there's a lot of directions that this can go.

For instructional design, what I can speak to is the technologies that instructional [00:25:00] designers use. I think there's a possibility of disruption on the distribution and presentation of content. So we've kind of started to see the rise of the headless LMS and a lot of people are like, SaaS, at the end of the day is just a database with a front end to present the database.

And as AI coding agents get better, maybe that becomes more commoditized. So I think maybe people are going to rethink how content is distributed and measured and tracked, and maybe it becomes more in the flow of work. And I think AI helps with a lot of that. And it obviously helps with , personalization is kind of the obvious answer.

I would hope that AI does a better job at really following learning best practices and principles. That's something we are actively thinking about as well, is it helping you [00:26:00] practice or is it just checking your retention and memory. And what do tools look like that are helping people put things into application?

So, there might be a popup of more practical L and D applications, I think would be really cool. Yeah, that's some of the things that I'm thinking about. I don't know at the end of the day, what the future of instructional design is. I know that our instructional designers love moving faster and AI has empowered them to feel less like order takers.

So that's what I'm excited about. I'm excited to empower L and D organizations. 

Nicole: Thank you Ethan for this insightful conversation and thank you everyone for joining us today.