From Chat-With-PDF To Enterprise Learning Transformation

What if your best trainer is the person who’s retiring next month—and an AI that turns their brain dump into accurate, engaging lessons overnight? We unpack how a scrappy chat-with-PDF experiment grew into an enterprise-grade course creation engine that slashes time, cost, and confusion while boosting trust and adoption.

Ro shares the early product choices that mattered: attacking hallucinations head-on, grounding content in source material, and designing guided workflows that mirror real instructional design standards. We dive into the moment enterprise demand hit—when a major manufacturer needed to capture critical know-how from a retiring workforce—and how AI made it possible to transform raw voice notes and messy documents into structured training with assessments, examples, and clear outcomes. The conversation widens beyond linear courses to the realities of learning at work: microlearning, deep search across knowledge bases, and just-in-time guidance that supports real performance.

Then comes the power move—joining LearnUpon. By pairing Courso’s AI-driven content generation with LearnUpon’s LMS strengths in skills mapping, role-based delivery, and gap analysis, creation finally meets distribution. Think on-the-fly lessons triggered by skill needs, rapid updates that keep content fresh, and spaced repetition that helps learners remember what matters. Ro’s philosophy grounds it all: AI should be invisible, human-guided, and relentlessly practical. It’s not a thinker; it’s a multiplier for experts, a backstage crew that helps knowledge scale beyond silos and survive turnover.

If you care about L&D strategy, instructional design, LMS integration, knowledge capture, and personalization that actually helps people learn faster, this story-rich deep dive brings clarity and a real blueprint for action. Subscribe, share with your learning team, and leave a review with one course you’d automate first—we’d love to feature your ideas next.

Transcript

Setting The Stage: AI And Courses

SPEAKER_00 0:01

Hey everybody. Really interesting, exciting topic on how AI is changing course creation with an innovator in this space. Ro, how are you?

SPEAKER_01 0:12

I'm good, thanks, Evan. How are you?

SPEAKER_00 0:14

I'm great. Thanks so much for joining. Really intrigued by this topic. Uh, topic is someone who loves consuming courses. Uh I never really thought about how they're created, particularly with AI these days. Before that, maybe introduce yourself and a little bit about your background. Bye.

Ro’s Background And Origin Story

SPEAKER_01 0:30

Yeah, sounds good. So I'm Ro. I'm the co-founder of Courso. Um, I have a background in product design, been doing that for about nine years or so. Um actually, at my last company uh in uh Amsterdam was a consultancy. I met my co-founder, who is a tech teacher. And uh we've been building Corsair for about two and a half years. And last year in November, um, we joined Learn Upon here in Dublin, which they're in a learning management system, um, which does you know the delivery portion of the learning experience for students.

SPEAKER_00 1:01

Brilliant. And when you started uh Courseau, what was the big idea? You know, what were you trying to solve uh back in the day?

From Chat With PDF To Learning

Prototyping And Early Customers

SPEAKER_01 1:09

Right. Um, so Coursaw actually didn't start with like the LMS angle or like workplace training or anything. It actually started a lot more just like for ourselves. Um, it was more of an experiment. So the very first idea I think came from Matias. Um so he was building back in like December 2022, the infancy of AI um chatbots and stuff, he was building a chat with PDF tool. Um I'd say was probably one of the first. Yeah, I think the Waiheba was actually probably the first that hit the market, which is quite exciting. Um I helped him with the product design there. And the idea behind it um was that he wanted to take um complex um contracts and legal documents and allow just like the everyman or you know, anybody to like break that down and really understand what was going on under the hood. Um, so yeah, he was kind of working with like reg and all that. And uh it was really it was a neat little tool that I I used a lot of as well. And after we built that, we were kind of trying to see or um yeah, we're trying to see if there was more we could do like on a bigger scale. And so we thought about education. Um and this came about when I was reading a really long form piece of content, really dry. And I was wondering if I could maybe ask ChatGPT to kind of break it down for me, make it more interesting, more fun, test me a little bit, you know, um, just make it actually engaging because I have a very short retention span. Um I tried it out with ChatGPT and like it actually worked pretty well with examples and everything. Um and I went up to Matt and I was like, hey, do you want to like try to build a prototype around this just to see what happens? Um and so we did that quite quickly, ended up getting customers actually quite quickly as well. None of this was really expected at first, but well um really the first iteration. Thanks. Yeah, the first iteration was really just to kind of see um uh yeah, if we could make something actually useful. So from Matt's angle, um he was really interested in testing the limitations of AI. At the time there were quite a few. So, for example, hallucinations were a really big issue. Um yeah, I'm sure you're familiar with that. You get all kinds of plausible output from LLMs, right? Um, because that's kind of what they're meant to do rather than be hyper accurate. So not really into like writing evals and like learning how to like do that really well. Um, another thing is increasing our context length, both input and output, because back then, again, um there was like very, very strict limitations on what you could kind of upload and what you would get out in terms of length. So we were just like, well, he was like trying to push the limits on all these different facets. And um, I think that was like really cool because it um it set us apart from all the other like um wrappers that were out there that were kind of just trying to take what's already available and just like make it like for a certain use case. Um and then yeah, from my angle, it was very much more like product first um as a product designer. So it being beautiful and usable and friendly was was was imperative, and um also building something that was actually useful that customers wanted.

SPEAKER_00 3:59

Um yeah, that that's that's amazing. So I I like that building something users actually wanted uh in business to business uh kind of sales, not always that obvious. So who ultimately became the the heavy users, the the power users, uh who were at Corso, who really uh engaged with the product?

Tackling Hallucinations And Context Limits

Enterprise Use Cases Emerge

SPEAKER_01 4:22

Yeah, that's a great question. And this is kind of how we like found our way into the enterprise world. So again, at first it was kind of for people to just like remix their knowledge, like get what was in their brains out into the world or whatever. But um, we started getting um this these customers that were a lot more serious, like pretty early on. So, one example was a pretty large manufacturing company, and their workforce was just about like most of them were kind of retiring around the same time. So they were a bit panicked about being able to teach the next generation of that workforce. Um things stuck in these people's brains. Um, and they didn't have a big enough team in the learning space to kind of like support that. And so they were kind of desperately looking for AI solutions, found us. And when we learned about their use case, we're like, hey, this is like kind of an interesting, it kind of clicked for us because we had been working for a while on like reducing hallucinations, and people were like genuinely very impressed by the accuracy of the output, like people were telling us all the time. Um, and of course, in an enterprise or like a B2B situation, like learning um accuracy is highly important. Um and we were at this point able to ingest a lot of content. So imagine um one of these people at the company could just like brain dump all their knowledge, like make a voice recording, um, and then or have 10 people make 10 voice recordings and like dump that all into Courso and get out like a structured um output. So we really kind of like laid the groundwork early on with our experimentation there. Um and then yeah, when we realized this gap in the enterprise, like the corporate learning space, we started working with instructional designers and people who actually um built courses for a living and basically like rebuilt Corsot systems um to follow like their expectations. Uh so that comes down to like the workflow. Um, you know, it's like a guided kind of process that anyone can use, including subject matter experts, people without any learning design background. Um, but also like within the systems, it's like built in these like pedagogical standards. So the output is actually as close to what you'd expect an actual instruction designer to make. Um and yeah, we always, of course, recommend that you kind of review the output. That's always important. But yeah, we got to a pretty good place doing that. So when we're like, hey, there's like a there's a real business here. Um, and instructors can do other things that are more strategic. Um, and it's insane cost savings for companies who are spending upwards like 10 months and like tens of thousands of dollars creating every single course. And then by the time it was out, it was like too late and you know, like not relevant anymore, or like you know, updates took forever. So, yeah, it just kind of solved all these problems. Um, being I love that.

SPEAKER_00 6:57

Wow, what what a great story. So things were going onwards and upwards, real customers, bigger customers. And what what was the idea behind joining Learn Upon? When when did that opportunity make itself?

Designing For Instructional Standards

Speed, Cost, And Course Maintenance

SPEAKER_01 7:12

Yes. Um yeah, so this was kind of at another transition phase when we're realizing, okay, so courses, sure, but like did is it something we believed in in the sense that, okay, yeah, of course, if you're gonna do like a compliance training, it needs to be like a structured course and you need to get like a stamp of approval. But for a lot of like real learning, it doesn't happen within these like, you know, linear course outputs, right? That people were spending a lot of investing a lot of time producing. Um we were thinking what other formats there were, were we thinking about like microlearning, but also just like deep search, like searching within your knowledge bases, um, because a lot of time learning is just like on the fly, like you need to know something right away, but also in a you can still like structure it a little bit to be um to help you like memorize things or like be space repetition to help people and so on. So anyway, we have all these ideas on how we could um kind of revamp workplace learning. But the thing that we were missing was the delivery component, which LMSs do best. So LMSs understand what your customers, uh sorry, um what your what your learners um who they are, um, what their roles are, what your skills gaps are as a company if you're a really good LMS, like LearnerPone is. Um, like I mean, you you you would know like what the company's um objectives are and things like that. So um, and then the delivery component of everyone actually getting access to the knowledge that is created. So we wanted to, like our third kind of um, I guess, uh I wouldn't call it pivot, but I guess like new angle would be to um work closely with learning management systems, become kind of like a partner or like a plug-in or something where um we enable the on-the-fly, on-demand training um where it's identified by a LMS. Um so we started looking for partners like this. And we went to this conference. Um, actually, the first person or the first like group that we spoke to that was an LMS very large, they actually were like, Hey, we just want to, can we just buy you? And we're like, Oh, never considered that. Um, let's do that. Um so a few months later we started talking, and then um it was funny because we were like close to signing this Exoscivity agreement. Um and like right then, Sean, a product manager from um Learn Upon found us somehow and had a conversation with us, and then we we just thought it was gonna be like a partnership discussion. Um, but it really kind of like snowballed very quickly um because we told them we had something on the table and um then they kind of amped it up a bit and we ended up pushing the other one back a bit, having more conversations, meeting Brendan, the CEO, Des, um, this the CTO, and just like really connecting with them and their vision. Um it's crazy because we had all these like um like late night conversations where it was like that this the CFO there was there as well, the five of us, just like talking about vision, talking about how this could go, demoing the product for each other, and like just growing more and more like like both are all our brains are like kind of on fire, like with what we could build together. And it was it was really beautiful.

Rethinking Learning Beyond Linear Courses

SPEAKER_00 10:10

So oh, that is an amazing story. I love that. I'm just envisioning that meeting and how awesome it must have been. And obviously, Learnapon acquisition happened. What's it been like within Learnapon? What have you learned, pun intended, with the new organization? Obviously, bigger company, lots of uh of different threads going on, I'm sure. Tell us about that journey post-acquisition.

SPEAKER_01 10:35

Post-acquisition, yeah. I would say there, I mean, what's great is like, you know, you start off with the vision alignment. That was like imperative, and then the cultural alignment, you know. Um, so the fit was already there, the groundwork was laid. Um, I feel like those are the hard pieces. Um, so now when you're working through the tricky bits, which is like cross-functional alignment and getting, you know, getting getting a bit more enforcing from like different departments, those are things that I mean, it just takes a bit of time, but everyone is kind of like bought into the mission and um like we have all kinds of support. Um, so we were actually at like a bottlenecks um state when when we got acquired because we were starting to speak to really large companies that had like so many security questionnaires and um and now Learn Upon is supporting us, like the they're so wonderful. They're just doing all this in the background, and um yeah, and so we're able to actually build and work on the innovative part. Um, is great.

Why Partner With An LMS

SPEAKER_00 11:35

Yeah, no, and to have a company supporting you is is always helpful. Uh how is your your let's say philosophy about AI evolved over the last few years? It's moving so quickly. I'm sure you have your own stack of tools that you know and love, as well as designing uh dish ones. What do you think about the current uh your current approach and how it might change over time? Do you have like a philosophy around AI that you try to bring to clients and into your user experiences and you know uh design experiences? Uh or are you just taking it uh day by day, week by week? It's tough to you know, it's like a fire hose, drinking out of the fire hose. It's tough to keep track of what's happening out there in AI.

Courtship And Acquisition By LearnUpon

Life Post Acquisition: Scale And Support

SPEAKER_01 12:22

Um there's obviously like yeah, a lot of new things all the time. I think I try to keep my finger on the finger on the poles, but you know, I do keep track of like what's happening on a you know um broad spectrum. I think Matias is a little bit more in-depth. Like he tries every new, every new tool, every new um uh model that comes out. He like really like, like I said, he pushes like the limits of AI. And I think again, that's part of the reason why we're you know successful is because we've always been like a step ahead. Um for me personally, I don't use AI for designing much yet. I obviously I see the potential. Um, I think I just I like being able to do it by hand, and so I'm just I'm pushing that a bit more until the tools get a bit more stable. I have no doubt that in the near future I'll be doing most of my designing with AI, but right now I still actually just enjoy it. So that's what I'm doing. Um, I'm very open to it. Um, I think also that AI is being treated um very much like a focal point right now, and you know, a lot of AI for AI's sake um mentality. And I do think that AI is more of like a I mean, it it should be like a backdrop kind of thing, you know, like a word process and enable you to it just kind of disappear into the background. Um it it shouldn't be something that people are constantly kind of shouting about, I think. So that's I guess that's my general philosophy on AI. And that tends to be more average than remarkable. So you really need to still have the expert there to like guide it. Um, it's not a thinker per se. Um great at analogies. Um I think though, on the other hand, I want to talk about like um my philosophy when it comes to learning. I think that's the you know, the use case that I'm the in the digging into the most. So, like, you know, zooming way out and like back to when we started Corsair, and um getting a like a little cliche here, but like democratization of knowledge. I mean, this from the beginning was real, yeah. Yeah, right. So many people have all this knowledge in their heads, like knowledge holders, not necessarily they want to share it, but they just like they're not that's not their forte or whatever. They they they're not meant to be creating trainings or whatever it is. Um, so just being able to help um people who who have this knowledge disseminate it. Um, and I think again, like way, you know, expanding it way out, um, if we think about like elites with academia and how knowledge divides contribute to like class and wealth divide as well. And how the knowledge that you're getting um is based on like how good of a school you go to or if you can go to school, things like that. So definitely dissemination of knowledge. Um and I think if you zoom into like the workspace context, um the learning context that I'm currently in, um, obviously taking those expert insights and turning them into learnable skills, uh, very, you know, used, very great use case for AI because you don't have to come up with anything new, um, which AI is not, I don't love AI for. Um, it's just kind of like rejigging things, right? And remixing. Um yeah, and then I think personalization, of course, is a big topic, but basically I've always thought uh interest-based learning um and curiosity-based learning is how we typically learn. When I was younger and all my friends, like we would go on Wikipedia and go into rabbit holes, you know, classic. I think many of us did this. And nobody told us to do that. We just like wanted to click hyperlinks forever and ever and ever. Um, and but that does come from like a real source of interest, right? So tagging on to I think existing neural pathways of learners as well. So things that people already care about and already think about a lot, um, is a great way to learn because it helps you remember it, like, you know, tethered to something you already that is already a thing in your head. Um, and then like I said earlier, I think it's great at AI is great at analogies and connecting things. So imagine, like, yeah, I mean, I think Google started kind of doing this recently where like for kids, you have examples based on soccer or something if that's what you care about, but like you're learning math. I don't know, really cool.

Philosophy On AI: Invisible And Human-Guided

SPEAKER_00 16:10

Um so yeah, I know it's been been fabulous that they're learning. I mean, in the in in the old days, you know, tutors, you know, and uh by the rich and famous could hire tutors to one-on-one tutor their children. And of course, most people can't afford or have access to tutors, but AI is becoming those personalized tutors uh that we never could have dreamed of had access to. So exciting the work you're doing. What are you excited about this year? Any events or travel, or is it more heads down getting product delivered to customers? What's on your radar? Next week or months, any events, for example?

SPEAKER_01 16:50

Um yeah, so I mean at Learn Upon, I think we're really heads down with building the integration. So they have a really great learning management system that has um skills pathways and um also skills taxonomies that help you like really, you know, define those learning gaps um within a company, right? So this is the perfect application for a Corso to come in where we can generate the learning on the fly. Um we are building up all the functionality that's on Corso right now into that into their platform. Um they're also keeping Corso separate so you can still like buy a license or whatever and connect it to your existing um LMS with um with uh Swarm or yeah, export it basically. So that's kind of where my head's at. I'm trying to think um about our roadmap. We have so many things that are now finally um we're we're finally able to action on our roadmap, is like we have like 300 tickets and so many exciting, exciting features that I've been like itching to do for like some of them are like two years old, and I'm like, please come and do this. Um and how we've been able to do it earlier. We've been like, you know, first to market, but what you know, whatever. I'm happy with with the pace um because we've really focused on quality and product. So yeah, I'm I'm excited to see evolve.

SPEAKER_00 18:07

Yeah, no, it's a wonderful story, wonderful journey. Thanks for sharing it. Uh, I know this isn't your normal format, so really appreciate you stepping out of the uh the office, as it were, and sharing the history and the vision. And uh it's it's such an exciting journey. Congratulations to you and the team. Thanks so much.

unknown 18:26

Yes.

Democratizing Knowledge And Personalization

SPEAKER_00 18:27

All right, and thanks everyone for listening and watching, sharing the episode. Also check out our TV show now, Bluebird TV and Fox Business at Tech Impact.tv. Thanks, Ro. Thanks very much.

SPEAKER_01 18:38

Take care.

SPEAKER_00 18:38

Bye. Thanks. Bye bye.