AI Explanation

Hi , Dimitri .

Um As you suggest , I thought it would be valuable uh to show you exactly um what this A I is being used for currently and maybe give a little idea of what we might use it for and answer some of the questions you you put there too .

Um So our core product um at our organization is , is building these custom elearning modules .

So the typically our client will send us materials like a cybersecurity policy uh or uh like a classroom um lesson and then we'll convert it into uh these elearning modules .

This uses a , a tool called uh storyline , articulate storyline .

So , elearning offering tool and then typically goes on a learning management system um uh like Moodle or like um success factors one of these .

And so that's the most common thing that we do .

Um So then we end up with a module that looks a little bit like what you got on the screen .

Now , this is our own internal module that we use to test this capability .

So if you imagine we've been given a onboarding presentation , we've gone away and created this elearning .

I'll just mute that and So usually um in a typical module , you know , we'll have navigation , they'll go through each of the sections , there'll be an assessment at the end which I'll come back to and so on .

Now , the key key differentiator for this module .

So this is all standard so far .

But as you might see in this bottom right corner here , if I point to it , um this is where we embedded our A I avatar so we can put whatever we want here , it's obviously custom .

So when we click on it , this loads up the I frame which goes to the bot .

Um So this is just a um ka I do capitec.com .

So right now , it's literally just an Ira in bed .

So it's no more sophisticated than that .

Although ideally , I'd make it perhaps a little bit more um integrated so that um the performance is perhaps better , although it , it , it's fairly quick to load up there .

So now I can come in here .

This is obviously the legacy one that we had .

So I haven't changed this because I didn't want to break anything .

Uh So I can ask this , um how much leave do I get ?

And then it'll go back , same framework that you've seen already and it'll come back um with the lead policy .

Um So this was our , I guess minimum viable product to prove that we have this capability .

Um So we're looking to improve , of course , the responses be able to quickly replicate the um the app itself .

So for a new client , they've got some policies and procedures , we can train it like we're now doing for the Cyber uh example .

Uh And then um being able to roll that out inside initially , their elearning modules , we chose to do it inside of that learning modules to start with because this is what we already have delivered .

We can Upsell this to current clients , um sell it to uh past clients who already have modules in place .

And so it made a lot of sense to pick it in here and we can control this .

And it's a nice little bit of pretty simple IP for us , that's kind of the first phase later on .

What we want to do is have a level of abstraction , I guess from the modules .

And so then integrate that into other delivery mechanisms that could be directly on the learning management system .

It could be through other communication channels like um Slack or teams or whatsapp even um being able to , you know , they can ask the questions to the bot and then we can even later on , uh I haven't really talked about this .

It will have a space repetition product .

So we can ask some questions every day , every week and so on to reinforce learning .

So it all comes back to the efficacy of this learning .

So right now , this is just an A I uh tutor as I'm calling it .

So we come in here and it supports the learner , they can ask the bot questions and hopefully give some good responses in their native language as well , which is vital for where we are here in the Middle East .

So we can respond in Arabic , hopefully um and uh support the learner to increase learning outcomes .

That's how I'm positioning it , the next kind of capability that we'll have .

So if we go to like the end of the lesson , I don't think I can skip ahead , but we might um I , I mention this here , we instead of it being a tutor where we ask it questions , then it might ask us questions .

So we have it embedded on the screen and then it has a the initial question that asks us either a quiz question or a scenario question we respond and then or the learner responds and then they uh and then the the scenario kind of evolves hopefully within tight constraints uh to then um again , support the learner .

So all of that in one of the requirements you've seen is going to be tracked as well .

So then we can know exactly what the people are asking the bot um or asking the tutor as well .

So then we have full track record , we can improve it , we can give um we report to our clients about any gaps in the learning and the knowledge and so on .

So all this kind of feeds around , um , that component .

Now , um , I , I'll , I'll address this here as well .

Cos you mentioned about , um , using the , um , um , not being concerned about open A I using the data .

Now , that's absolutely like what , what you've written there , of course .

Um , now , but with our clients , some of them are like government , um , some of them are semi military almost .

Um So A US P for us is to say your data , even if it's in a vectorized data , but in this Rag model never goes to a third party .

So even if open A I or whoever is not going to train their models on our data , we don't want to send them the data because uh we lose confidence and about the um II I guess the security of that .

So for us to say , OK , in our modules , it's on our servers .

The Rag uh A I , not that I , I say that the tutors controlled by us , it's on a language model that we control .

So your data never goes to a third party .

That's the reason to have the private LLM .

Um So I , I won't necessarily invest in that .

Um until we've got our first kind of client , we'll set that up as , as um as I kind of mentioned .

And so that way I can with great confidence , say to a client , your data is completely secure .

It never goes to anyone else .

Open A I or , or anyone , anyone um beyond our organization .

Uh So I hope that explanation is helpful uh and nice and visual .

Please let me know if you have any other questions , happy to jump on a call as well .

If that's more suitable or happy to make another video recording or voice or text or whatever is easiest for you mate .

Thanks very much to Metro and I hope I'm pronouncing that right as well .

So , um uh yeah , I look forward to working with you soon .

Cheers .