Episode Transcript
Kyle James (00:01.1)
Hey, welcome to the AI Chronicles podcast. I'm your host, Kyle James. And today we're going be discussing how an EdTech company called Course Agent is using AI inside of their own business. And we'll share the exact steps that you can take in order to implement AI for yourself. Now, before we dive into that, listen closely. Are you looking to implement AI inside of your own company? Or maybe just struggling to get your AI to stop hallucinating? Speak to GPT Trainer. GPT Trainer literally builds out and manages your AI for you.
eliminating hallucinations for good. Go to gpt-trainer.com. I promise you, it'll be the biggest time saving decision that you've made all year. Trying to set up AI on your own is like trying to build a house from scratch. Sure, you could do it, but the time of frustration, it's gonna take you to get it finished. It may not be worth it. It's a thousand times faster and safer to hire professionals. Schedule a consultation today. Once again, that's gpt-trainer.com.
Say hi with me on the show, Peter Murr, who is the founder of MeLearning, a company that provides online training services to over half the local authorities in the UK. And recently, Peter is also the founder of the startup company called Course Agent AI, a course authoring tool with AI at its core. Really excited for this conversation. Hey, Peter, welcome to the show. How are you?
Peter Murr (01:24.31)
Yeah, very well. Yeah, thank you, Carl. Thanks for inviting me on to your show. Thank you.
Kyle James (01:28.322)
Yeah, for sure. For sure, man. So give us some context here. So you founded course agent, like walk us through that background. They're like, how did you know you mentioned, obviously I mentioned the intro here, like me learning. So maybe share a little about that and like how you got to where you are today.
Peter Murr (01:31.446)
Yeah.
Peter Murr (01:37.142)
Yeah. Yeah. Yeah. So I'm the owner of a company called MeLearning, which creates online courses and builds online courses. It's been going for now 20 years now, but strong team and we've got a big customer base and we create kind of compliance courses for organizations. So.
A great team there that started about 20 years ago. And in terms of how CourseAgent got a firm, my background is writing and building courses using a number of different authoring tools. And I've used, I think, quite a few authoring tools over the past. I've always been a little bit frustrated about how long it takes to create courses, what skills, the amount of skills you need to create courses, the cost and time involved to create courses.
And then, you know, so I thought there must be a better way of perhaps doing this. And I always wanted to build my own authoring tool because all the authoring tools out there, was always that didn't do something and this that did something else. And it kind of was a little bit frustrating. So about two years ago, I thought I'd sign my own authoring tool and to create online courses.
But there's no point in building one. It's compete with the rest out there. It would have been fairly impossible. Now, I looked at AI about two years ago, and I thought, that's interesting. But actually, it's surprising. Even though it was only two years ago, it wasn't great. And then I looked at it again about maybe a year and a half ago and thought, hang on. Or a year just just to, I thought, this is amazing. I can see.
how perhaps the ideas and experience I've had over the years in terms of writing courses, creating courses, using the right training methodologies could be used to create online courses in minutes using AI, using the experience I have. So I decided to set up a totally different separate company called CourseAgent, a new startup company, with the idea of creating
Peter Murr (04:03.19)
online courses using AI at its core. So that was the idea. It started about a year and a year and a half ago. So my challenge was learning more about AI. didn't really know what I could do and experimenting with it, having finding something that I could use to experiment, a particular tool I could use to experiment with AI and all the different LLMs that
I could potentially use, working out the right kind of prompts to create the courses that I wanted to create with a fantastic outcome. So they were my initial challenges. so Course Station is just a startup company at the moment. kind of using, I've got a development company who's building the product at the moment. about three months away from Go Live.
And I'm currently working on the prompting as I speak. So I'm literally, I've got screens up working on different workflows. So, sorry, a little bit jumbled up that exploration, but that's exactly what I'm doing.
Kyle James (05:12.982)
Yeah, yeah.
Yeah, for sure. That's cool. like talk, talk to me a little bit about here. like you, um, you know, the course content, it's really, like, it's creating like e-learning content. And I think, you know, a lot of people spent, you can spend a lot of time. So was that like kind of the initial thing was like me learning, you were spending so much time building this content out. You're like, man, there's gotta be a faster way. And if so, like, I guess like walk me through, like how, how is the AI going to help, you know, for those course creators out there, like going for like.
Peter Murr (05:23.914)
Yes. Yeah. Yes.
Peter Murr (05:32.32)
That's right.
Peter Murr (05:38.166)
That's right. Yeah.
Okay. Yeah.
Kyle James (05:45.1)
Ham spinning four hours a day for one course, so ham spinning 15 minutes. Walk me through that a little bit.
Peter Murr (05:47.636)
Okay, yeah.
Okay, so there is a way of creating courses. Okay, so I've got a workflow that, first of all, you need to know what the course is about. So in terms of the workflow that I've put together, just based on experience, what the course is about, who the target audience is, who the sector is for, the location to location.
Then you have to kind of build a course which says, OK, what is the skill level of the person or the people you're trying to address? But also then, also what type of course complexity do you involved in that? So you might be a highly skilled university lecturer that wants to learn something new about a subject in your field, but you don't know about it. So you don't want a highly complex course. So that was kind of trying to.
split out the skill level from course level complexity was a bit of a challenge. I've done that now. Then you work out, typically as many of the writing tone, but it's not just one specific writing tone. It's like a writing tone that has a primary writing tone and a secondary one underneath. So that there's a kind of a bit of nuance between all the sections in the course. So you might have something professional, but with a kind of a friendly bait or a kind of a
casual based it. And then you can, I'm not going to do it in the first one, you can actually sort of switch between primary and secondary as you're going along. Then you've got this, there's plenty to think about after that, because once you've done that, potentially have to, in this platform, you can either upload any kind of reference materials or source material, the training materials that you've got already. So you've already got
Peter Murr (07:49.366)
some kind of training materials or some sort of, if you're a subject matter expert, you might want to put up some information yourself. And then you can, with AI now, you can say, type of influence is that AI going to have on your uploaded materials? Do you want to have a minimal influence, a moderate influence or a high level of influence of AI involvement, using it as a base source? So then that's working out. Then you kind of work out then,
the course structure, say, OK, do you want it to be a short bite-sized course, you know, about 10 minutes or a 20-minute course or a kind of a high level 30 minutes with teachers? Enough, you know, if you're going online, do that. I I hope you won't talk too long, but people's attention span will be kind of limited. And then the one key thing about training is that people want to relate it to their work and to their jobs.
And therefore, want to influence some courses. It's very high level scenario level. So if you want a lot of scenarios in your course, or you want it theme driven throughout, or you want narrative driven throughout, you can select those things and go, OK, I'm going to mix it up. If it's a maths course, you don't want too many scenario levels. But if it's a kind of an induction course where you want to relate it to your workplace or whatever, then you can sort of. But then also the
The addition to scenarios is that I've added to this is that you can contextualize the scenarios even further. So once you say you want scenarios in your particular training program, can contextualize it and say, for this particular area of the business or this particular part, I'd like to focus on this when you produce a course, which is even more relevant to what the person's learning.
Kyle James (09:41.004)
So what do mean by text? tell me more about the contextualized part. Cause that's, think that's like a really important as it starts to like relating, you start relating to like, okay, this is how it relates to their, their scope of work or they're like their, their job. Like tell me like, is that like a specific, like, don't know. I'm just not grasping or grabbing it.
Peter Murr (09:45.77)
That's different, isn't it?
Peter Murr (09:53.024)
That's right. That's right. Yeah.
Yeah, that's okay. It took me a little while to work it out. Because I generating courses with scenarios and they were brilliant. mean, the output is just amazing. So, you know, with the various kind of prompts that you're putting in and the scenarios were great. And I've got bit of feedback to say, okay, well, how can we make these scenarios contextual to us? Because they'll produce great scenarios. But if you want specific to your
Kyle James (10:02.103)
Okay.
Peter Murr (10:26.87)
type of work or the type of thing you do or the type of the company does, you can narrow it down even further. So I kind of figured out kind of six different kind of categories for that. And that's kind of generic scenarios, which is what the platform can produce, everyday life scenarios, professional scenarios, academic scenarios, historical and contemporary scenarios. then it's good. It does get.
It's simple to choose because you'll get an explanation behind. And then you can add your own. You can actually type in your own scenarios. And then what will happen is it'll understand what the course is about. And then when you press a button, it will work out what type of scenarios will fit the course content. The particular type of course, if it's a historical one or if it's contemporary type scenarios, it will.
narrow it down to two or three of those and it will recommend a couple of types of scenarios and then you can also type in specifics about what in particular you'd like. I've made it more complicated than it should be.
Kyle James (11:33.154)
Yeah. Yeah. No, no, no. That's really cool. Cause I think the biggest thing is like the first couple, like, right. Like getting the skill level complexity to like the tonality, the training materials, how much influence would they like all that? like it's pretty step-by-step. Like I think like, as you get towards that last bit of like relating it to their work, that's where like it really plugs in. And it makes sense why I like the contextualized part where it's like, like how do we want to phrase this in a way that like makes sense?
Peter Murr (11:40.0)
Hmm. Hmm. Which is typical.
Peter Murr (11:52.064)
That's It does.
Kyle James (12:00.354)
to them because they can have all the data and the process and like, then if they don't relate it and connect it, then it's kind of in a sense of it doesn't work. It is dead in the water. So I can see why the contextualized part, you know, is so.
Peter Murr (12:06.227)
Is that it?
Yeah. Yeah. Yeah. That's the magic source, really. But prior to that step, which I missed out because we talking about scenarios, is you have to generate some suggested objectives before you do that. So you've suggested your scenario level. You've put in all your data. And then you say, OK, can you supply me some suggested objectives? Because that will dictate the scenario level.
as well. So once you've gone past to say, this is my level of scenarios that I want, this is the high, low, medium level, and then use the overall course objectives, and then that will drive what you will achieve if you're using a scenario-based course. So that drives that as well. And then once you do that, it will produce
Kyle James (12:56.684)
Yeah, for sure.
Peter Murr (13:07.446)
I've been doing a lot of testing on this because also what we've done in the platform is that's just the high level prompting. There's the other prompting we've thrown underneath which puts certain taxonomies into it like Bloom's taxonomy and training taxonomies underneath. So that when it produces a course, it comes out with the perfect introduction, the learning outcomes. So you put in objectives and the system works out learning outcomes for the learner. And then it produces
interactive slides, knowledge checks, key learning points, summaries, final course summaries, and it will generate quizzes. I haven't even talked about the quiz section because you can then define the level of quizzes in terms of the complexity of the answers you get. There's citations as well that gets generated that you can say, okay, where did you get this information from? There's another element that I've added, which I've
had to take out because actually what your listeners might know is that the more prompts you put in to AI, the more expensive it gets. suddenly, it's got to be a consolidation of code. But there was another step that I put in, which actually goes back on itself. AI is going back on itself, checks it against all the parameters you put in and the objectives, which you might be familiar with. And then it recommends
what you can do to improve the course as well. that's, again, another expensive, but it's effective. You're running it again and saying, OK, tell me how you can improve this. Well, it might benefit from explaining this a bit more. then again, so I can't in phase one, I've taken that out because it does add a bit of time. But it's a pretty good, pretty good way of actually making the course as per you possibly can. Yeah, sorry.
Kyle James (15:00.566)
Yeah, for sure. So, so now like you've been, you've been, no, no, no, this is really cool. I really like, cause I think like even going back to like people who try to create content, it's like where they start, like even like the complexity side of things. Like I, I saw a presentation, there's someone in my network and they were sharing it with me I'm like, and they were talking about AI. And I was like, this is really calm, like kind of complex. who, like my first, was like, who you sharing this to? Like I, I didn't like want to question too hard, right? Cause it's like friend and family. I'm like, but I just like,
Peter Murr (15:05.088)
and
Kyle James (15:30.222)
This is a little too, this is complex like for me, I can't imagine for somebody. like, I like how it has like the framework of like, okay, here's the success part of like building out a, you know, like an e-learning or some sort of like workshop, right? And then having that structure too. Cause I think a lot of people don't really have a structure. There's like, oh, this is what I want to talk about. And like, here's the answers, right? But like you're connecting piece by piece with each of the AI. But my next question for you on this one, Peter is just like, so,
Peter Murr (15:34.784)
Yes.
Peter Murr (15:50.262)
That's right.
Kyle James (15:59.79)
Yeah. I know you're still like three months out from getting it like fully, you know, locked in complete with the dev team, but like what types of like results have you been seeing so far from just from your own testing? Like, what, like, what was some of the things are you like, wow, I can't believe this, especially with your experience at me learning and your background on that, like kind of paint that picture for me, what that looked like on your side.
Peter Murr (16:04.374)
That's right.
Peter Murr (16:11.605)
Yes.
Peter Murr (16:16.373)
Yes.
Peter Murr (16:20.502)
Incredible. think the word's incredible. In terms of the output, at the moment, it's going to go into a platform where it selects templates, interactive templates and things like that. But at the moment, it's going into just kind of this, also almost like a PDF format at the moment terms of the output. But what I've discovered is that you can put these prompts in and AI can actually sort of improve what you expect.
The language is clear. I've been writing courses for nearly 20 years, online training courses for 30 years. And I'll give you an example. I wrote a GDPR course, and it took me three months. I had two other script writers helping me. I had a legal team. I had instructional designers to help me to create scenarios, or organize. Over three months, I put the same parameters into my platform. This is course agent platform.
it took one minute. I thought, aren't changing. You still need instructional design. still need, because I'm struggling with it, which is one of the questions I was, what is image generation? AI has, in terms of creating the right kind of images or searching the right kind of images for your content, it's proving to be a bit of a challenge.
Kyle James (17:21.218)
Wow.
Peter Murr (17:48.372)
because I interpret it. So I do have a solution for that. That's what I'm working on today. But whether, you know, but it's but it's changing all the time. The other thing is that, you know, I started looking at AI a little while back during the last six months. It has changed so much that my initial thoughts and plans are almost like thrown out the window in terms of also the design of the product.
design of the product. So it started off, this is what I'd like it to do. And the developers are building it. And I'm going back to it going, oh, there's a far better way of doing this now. And every time you go to an AI engine or we use something called Mind Studio. I don't know if you're familiar with that to kind of test it. So that is also an issue because we're building this product.
And you know in about a month's time what you're thinking might can be the screens or the interaction of the UX or the UI can be different. So that's all right.
Kyle James (18:56.834)
Yeah, yeah, yeah. That's, that's cool. I mean, like, three months to one minute, like, think I wanted, it seems like, seems like almost too good to be true, but like, obviously you've seen it before, but like how, like, you think of the reason why it was so, I mean, Jenny, you like ready to go, was trying to think of a way to phrase this, like three months down to one minute, like, is it because you did all the steps beforehand? Like making sure like what, and the
Peter Murr (19:02.9)
I know.
Peter Murr (19:19.263)
Yeah.
Kyle James (19:26.67)
I don't know. I'm just like, I'm kind of blown away at how quickly and how fat, like how good the output is it like, Hey, 80 % there and you got to go back in and like kind of fine tune it. Or is it because you were so diligent and you were so detailed with the prompting on the front end, like with everything that way it was like, Hey, because like you're very detailed with like step one, two, three, four, five. That's the reason why I produced like a ready to go presentation or output in one minute. Like, you know what I mean? Does that make sense?
Peter Murr (19:27.114)
Yeah, yeah, yeah, yes.
Peter Murr (19:41.8)
Yes, that's right.
Peter Murr (19:52.644)
Yes, yes, yeah, yeah, yeah, the prompting really helps, you know, the prompting really helps, you know, because I knew with the original course, I knew what it needed to do, needed, kind of, kind of had an idea of what was right. And then putting those kind of prompting into this really helps as well. So.
I think it's just the course writing. And I tell you, it's understanding how things are structured and where you start with a course. So all of that's in the prompting as well. So how should you start a course? Should you start with a question? Should you start with a scenario? What is going to engage a learner to finish the course, start the course, and get involved in the course? And what can they take away from it as well? And what scenarios? So all of that.
I think is based on that original experience of what's worked because that course really worked, people liked it and then figuring out, okay, how can we do this in this platform and create something. But it in use in multiple ways. I had a 13 year old boy who's learning Mandarin, came and said, I'm learning Mandarin, how can I help? So we just pull the prompts into the platform. Suddenly had this very engaging interactive
course for him to learn Mandarin from his level, his base, for every individual, this platform can build something that will make you learn something a lot quicker, all kind of, you can put in all your requirements and it will create something specifically if you want it for you, not just for a great big organization. So it can have multiple, multiple users. I've only discovered that by actually talking to people and they're saying,
might be useful for me, I can see that in my job.
Kyle James (21:47.662)
Yeah, right, right, right. And so as we start kind of transition, finish up here, but like what, what are some of those kind of, I know you mentioned one of them, like with your work on like how the image generation as like that, I think everyone out there who's done the image generation, it is more of a challenge. It's gotten better, but it's still a challenge. Like what are some of those upcoming AI initiatives for course agent and like, where do you see AI playing maybe some of the biggest role in the next steps in your operation?
Peter Murr (22:05.098)
Yeah, it is.
Peter Murr (22:16.15)
Yeah, so that's what I'm going to be working on. So image generation is the biggest headache, I suppose, at the moment. But I think we've got a solution for it. The next AI is going to be, so once the course is built, and this might be in this release or might be in the next release, is that once you've created a course to automatically generate a refresher course on it, then automatically create a course that has a different
audience and a different tone as well. So the ones it's created you go and then also to have, so the refresher course also looks at updates to it might be policy or the law or to new ways of thinking. So it kind of looks at the course and goes okay what else is out there that has changed over time. So what it means for organizations is that if you generate a course and a year later you've got updated because all online courses
become obsolete usually after about two years or three years or one year because the law changes or policies change or ideas change. So rather than having to create a new one, you just press a button and say, could you go and check out there, come back with any changes, suggest any recommended changes and identify where they are if you want to change the course or if you just want to recreate the course with the new changes, you can. So that's the future of this.
Kyle James (23:39.822)
Wow, that's so cool. I go back to like three months of building to one minute and then like, can't imagine like refresher. mean, I don't imagine refresher would take three additional months unless there's like so many more changes, but definitely at the very least, like having some of the new content like transferred over, maybe take like two to three, four weeks long, but there's multiple courses. Like that could be, that could be someone's full-time job.
Peter Murr (23:50.324)
Yeah.
Peter Murr (24:00.406)
That's right.
Peter Murr (24:05.777)
If you've got, because I'm familiar with this, if you've got 250 courses or a thousand courses and you have to update them, that takes a lot of time. Take a large organisation, you've got all these courses, you've got actually that's update, that's update. But if you can press a button and go, could you just update this course now, please? You know, and then also bring it out to your staff as well. It means that...
The majority of time is spent usually in training is updating the course material we've already got. So having it in a platform that enables you to do that is invaluable to a lot of organizations. So that's it. Yeah, Sean.
Kyle James (24:46.094)
Yeah, yeah. For sure, and as you start wrapping up Peter, where can people that you'd recommend maybe learn more, learn a little bit more about you and maybe a little bit more about course agent that you'd recommend them checking out?
Peter Murr (24:57.62)
Yeah, yeah, I think at this time, just contact me through LinkedIn and you know, I'll see if I can get back to you. I've got a website up. I'm spending zero time on the website. It's like a brochure site and that probably needs changing as well. Courseagent.ai. That'll be updated soon, but I'm totally focused on AI and getting the AI working at the moment. So it's like my full-time job. So yeah, contact me through LinkedIn. Peter and...
and see you out.
Kyle James (25:28.526)
Cool. Awesome. Love it so much. Thanks, Peter. It's great having you on the show today. Appreciate you sharing your perspective within the course content creation and on the AI side of things. I it's really cool what your team is doing over the next couple of months. I'm excited to see what happens as you start fine tuning even further. And thanks again for everybody for listening in. Remember, if you're looking to implement AI into your business today, please don't try and do it yourself. The time is just the AI could cause. It may not be worth it. Schedule a call with GPT trainer and let them build out and manage your AI for you.
Peter Murr (25:41.088)
for your time.
Kyle James (25:57.302)
Once again, that's gpt-trainer.com. Signing off for now, have a great rest of your day. Looking forward to seeing everyone on the next episode of AI Chronicles.
