The following are the outputs of the captioning taken during an IGF intervention. Although it is largely accurate, in some cases it may be incomplete or inaccurate due to inaudible passages or transcription errors. It is posted as an aid, but should not be treated as an authoritative record.
***
>> Good morning, everyone. It is my pleasure to be here at this IGF to this session. And myself co‑host here and Dr. Lamya Alomair from MCIT acting on the technology foreside and digital economy will be delivering very interesting (audio cutting) has the foundation of AI as well as qualitative and quantitative to what organization AI or what AI mature organizations are in that space. By way of introduction, my name is (?) and without further adieu, I will hand it over to Lamya Alomair who has amazing experience in the domain and will be going through the first section of this presentation. Go ahead, doctor.
>> (low voice) and the third stroll revolution with communication and electronic also help people to communicate both over the globe. So all the industry revolution help our physical ability. But with the third industrial revolution, there is one more ability that comes with AI. It improves our cognitive (?) so with AI, our cognitive ability is moving.
So all ever us said from the history, you can understand what you have to do. So I will just talk about very brief history about AI. Everyone thinks AI started very soon. No. AI started in 1915. Okay? When a professor, an internal was thinking about a computer thinking like humans. So this is the starting point. 1950. So he trained the computer with information and also he brings human and then he asked a couple of questions for this computer and a couple of the quo questions for a human and then he judged the answer for these two individuals computer and a human and then guess what. This judge could not distinguish from which one come from a computer and which is one comes from a computer. And this woman, Ellen Turner said that computers can think like this. Then in 1955, John MacAfee in a conference, he launched the name of AI.
After that, moving to the first robot, which was very modern. So we don't want to touch the chemicals. So they start with the first robot. So in 1961. After that, I just want to introduce you to the grandmother of chat DBT and MIT and in 1965, they have the first (?). It was simple, but 25 years later, we got our (?). So Aliza is the grandmother for our chatbots now. And then because like around 25 years, there is nothing moving in AI. In 1997, IBM had this program called D2. It is this program the champion in the sport. So since then, IBM started doing these changes and they started to know, okay. Let's do another champion. In 1999, the MIT also had one representative with promotional products. So they mirrored AI with the robot and is they got the first motion. After that, Google started doing the driving the car. And just to let you know how important the data from this and how to be also very mature when you use it. Come they use it, they trade it for only (inaudible). And then there was a guy who wants do drive the car, and it was missing it. This is where it comes in and maturity of the day. And then some of you know the jeopardy game. In 2011, IBM got the program the champion. And then they continued doing Google and they also do the defined another program in 2017, there is a break in the machine learning for this. They had most of the field of help here.
And then in 2020, open AI launched the beginning to see which one was (?) in actual languages which is the six trying to touch that. (?) and we need a vaccine. Which vaccine using AI. That's why we can't get the vaccine. (?) I propose for this structure, it can be promoted very quickly. It is like 10 years to have the (?), translated using that and thin haves structure and then try to find and rediscover. But I propose it is predicting the structure very quickly and then the discovery.
Also, we're very proud to say that in 23, we had (?) and so we (?) and also saying all of this history, we want to predict the future. I want to see what this can take and at the same time, we try to learn the journey before for having a new future. That's why today me and my colleagues engineer and trying to give you a sample of how to be mature in applying this.
So everyone is asking: What is the definition of AI? There is plenty of definitions, but let's say here, it's a system as mentioned? The beginning. Covenant system is a natural system that connects together taking all this information and data and translating it to an output something to understand, something we can get information from it. Something that decision maker can make. Underneath this whole information, the Artificial Intelligence again is a problem alga rhythm that is getting all this information and have something analyzing, coding and (?). This AI and Artificial Intelligence and after that, it is a machine learning, which is a subdivision of AI. The machine learning is how to teach all this information. So we get all the data and (?). Underneath this machine learning, um, (?) (low voice). And the machine learning is different learning. The team has machine learning and it is done differently. All of these have applications, mood base off the system and (?) (low voice) and all these depend (low voice) (?) two or three of these machines.
Next, (low voice) (?) to advocate AI above (inaudible) mature. When we go and see, notice the location and the (?). We have an example here. We need it to (?) which is with the customer and how we put the customer (?) and how we can (low voice) (?). We manage and maintain and maximize. (low voice) this helps us. The function of how we can (inaudible) with different organizations, different governments. How we're behind this and how we can (inaudible) all this can AI help again. The new business here, I know we have new (?) with new business that will be having a great impact.
So again, AI(?) I will try to explain one example today. So you ask me how we can have the best and the northwest profitable customers. Use the machine. (low voice) you have given some data and you will care. Among the data, who are the customers we can sell to? How we find the customers we can sell. Okay? Using the rule‑based systems for coding and identify the segments. And then how we can optimize inventory and optimizing it to lower costs. (low voice) (?). All these AI can support.
Now, let's see how this can be applied? (inaudible) production quality. Okay? We are planning about quality and we use AI and base the quality. So how do we analyze (?) pieces that they can get out. 60 languages. I want to use it so how we can analyze it. For example, there is another one used is to analyze in AI and so we have repairs. So using the AI will give you (?) it will make life easier. This is something. What I learned we need the right formula for having (?). That's why it is very helpful. That's why we have this one. You are preparing through use cases. Not many use cases. It is dime to value. We understand it ‑‑ it is time to value. We understand it. So (?) (low voice) (?) that is something that AI (?) (low voice) so this is the formula (?). AI you have to be (?). (low voice) will be applying many. And one spasm here is (?) density. Personally (?) what they mean by personalization, it requires different choices, different, ah, means of very important work. (?) (low voice) (?) for anyone else and I will keep this in my mind. It is very important. First of all, understand the (?) and come they should start ‑‑ when they should start and start with the important (?). We will continue with the schedule and I will wait till the end.
>> Thank you, doctor. So hopefully you've learned, you know, the foundational AI techniques, which are not really enough to apply AI within your organization. You need the first elements that Dr. Lamya talked about and we have done some analysis in the market. We went out and surveyed 650 organizations around the world who are adopting AI in general who excel applying the AI techniques and practices within the organizations. We try to even fetch out these 650 organizations to understand what are the common practices. We really, really make organizations and AI mature organizations. So out of the 650 that we surveyed, only 20% ended up to be AI mature organizations. I will go through a lot of, you know, the learnings and a lot of the teachings that we have gotten out of this survey. So basically what we have done is a practical and quantitative approach to understand what the organizations (?) in the main adopting AI.
So we need to understand one thing. When Generative AI, Chad DPT came out in the end was '22, it really made a big splash. It really made the needle forward when it comes to general AI adoptions. Why? Because a lot of expectations, a lot of value received out of applying generative AI within the organization settings in general.
So this splash had some sort of ripple effect on the whole AI adoption in the organization. So basically it did not focus only on Gen AI, but it touched every other AI technique that has been highlighted by Dr. Lamya in the previous section.
And you can understand the numbers that are being displayed in this. So we have done this survey almost every year for the past 8 years. And general was not really a technical or AI technique that we have highlighted as top 1; however, during last year, it came out of this and one of them is to move from no applications within organizations to the highest used AI technique within the organization. As you can see, you know, it compared to every prominent and famous AI technique whether it is related to machine learning, NLP, optimization techniques and others. So it came number 1 within the user organizations and it had the ripple effect on everything else that's been noted previously.
And to understand this a bit more, the remaining sections of the presentation is three parts. The part where Gen AI or the part that demonstrate where Gen AI has made an impact on the overall adoption, then the part that talks about the challenges related to overall AI adoption and particularly Gen AI specific and the last one is to district the common practices around AI adoption within the organization.
So the top 3 impacts where on were generated from Gen AI in particular was focusing on upscaling AI upscaling in the itself and not AI savvy for staff and associates within the organization. This is number 1. And the second thing is pushing the needle well it comes to AI adoption to next level. So Gen AI had a splash around AI adoption general. So because Gen AI is focused and can provide specific use cases, but it does not provide value in every single other case. You need to use a plethora of AI techniques similar to the ones that were talked about Dr. Lamya and the top AI within the previous slide. But most importantly and it became non‑negotiable. The AI Government system. It deals, you know, sensitive AI or sensitive data itself and becoming ‑‑ it became, you know, the central universe well it comes to AI policies around the world. Here in sawed I, for example ‑‑ Saudi, for example, we had personal data protection law that has a lot of firsts, mechanisms well it comes ‑‑ well it comes to data related itself. Aside from that, we have the AI outlined or the Gen AI outlined from Saudi itself to give insights in how you should use GenAI and AI. In general within your organizations.
You can see each one will impact the other. The AI adoption will entail you upscale it. The more organizations and care you need to do well it comes to AI governance itself. You cannot deal with know who of them in isolation from the other. They are really, you know, intertwined and backing each other.
Sorry about the delay, but the clicker here is not the best AI technology. But basically AI has a chain reaction well it comes. Gen AI has a chain reaction well it comes to overall adoption within the organization. The number here is very interesting. You know, since 2 years ago, AI adoption has doubled, has really doubled. We provide with understood that AI adoption has gone from 1 X to 2X in the organization. You apply multiple business units as well as multiple business processes. This tells you that AI is becoming more and more seen and organizations are serious about it.
In '22, but in '24, they have been scaling. And that scalability is related to the (?) we have been talking about.
And one of the things that adoption AI in general is embedding it within the general application landscape. So if you have a CRM, for example, it is not uncommon to see a Gen AI sort of capability within the CRM to help the normal to try and digest many of the processor related to a very complex CNM and let alone the data hosting within the frame itself. Gen AI became forcing sort of function well it comes to AI adoption. In fact, you see it more used and embedded applications comparing to isolated or standalone chat GDP and everything in between the is outside really the embedded application. So basically if you want to really for the key lesson that I'm trying to say here, if you want to really push the limit well it comes to Gen AI, AI adoption in specific, you need to think about adopting techniques within the general business applications in your organization.
And this is basically what we have seen. If you worked in engineering or if you worked in orchestrating enterprise application, the enterprise application strategy has moved away from only being compostable and only being usable to include the embedded intelligence. And embedded intelligence is nothing but infusing the functionality of AI modules whether they're Gen or normal within the enterprise application landscape itself just to make them more powerful, more impactful and easier to use for normal uses.
As I said in the beginning, Gen AI has really made a big splash where can it comes to AI. Verify that we need to stand humble well it comes to the general AI adoption. It is forcing us to mature, you know, to higher stages with the plethora of challenges that we're facing in the Gen itself.
And this is ‑‑ some of the challenges I will go through in that section. Before I go through, you know, the details of these challenges, I want to have very testing data point. You need to understand the journey toward AI maturity is not easy. It is very challenging. It is very ‑‑ it continues to be complex as we stand today and it is very costly to organizations. And, you know, one of the things that we notice from our survey this year is that AI projects on average never go to protection. They don't see the light at all. 52% is not an easy number. 52% is a very large number. You know? You take an AI budget. You have a number of months as we see in the next slide and then you end up throwing it because you cannot really realize how it is the sort of point when it comes to the cost and value of AI of the source scaling later on in protection. So 52, this means it is like tossing a coin. You don't know whether it is seen or fail ‑‑ whether it will succeed or fail.
The next point we'll talk about is a normal AI project. So in the previous survey this makes, you know, things or organizations in general far to take general AI with the organizations. They need to understand that in order to succeed in their ‑‑ in order to succeed in the ‑‑ there the scale‑up aspects well it comes to organizations, they need to focus in different things like you will see in the next session.
52 is baud failure and 8mons just to pilot and ‑‑ 8 months just to pilot and at the end, you don't know whether it will go to production or not.
The revolution that came before it, you need to understand the AI technique and the use‑case value for the organization. Without maximizing the end results of that sort of formula, you will not be able to actually say that we are AI enabled organizations at all. As you can see here, the areas are lack of trust when it comes to AI, lack of technology, lack of technological out especially for those that come from certain countries where Cloud is not really enabled there. Provided AI models and AI capabilities are better in Cloud environments comparing to local environments and definitely talent is a big thing. But mostly important, the estimating and demonstrating AI values is the biggest challenge for organizations. And it goes hand in hand with the previous two days that we described. 52% failed and it takes 8 months to roll out a project from file to production.
We focus on GenAI in particular. Related to technical implementation can cause running GenAI and this here, we have seen more reliance, techniques especially for those that are (?) to cloudy computing. You need ops capability in place just to help you navigate the costs related to AI from something (?) and without it, it became also one of the top GenAI. Just with the GenAI literacy or AI literacy programs come to ‑‑ to resolve that sort of challenge as we would see in the next section.
So as I said in the beginning, well are many organizations and without that, there are many points related to their practices when it comes to AI adoption. We have high maturity organizations which came out to be around 9, 10%. Out of 650, we only have 65 organizations that are mature when it comes to AI adoption in the world. And we learned, you know, (?) when it comes to AI adoption and scaling AI adoption in the older. I want to highlight what action needs when it comes to our analysis to the data points that we connected. We need to understand that the common for the AI mature organizations applied or several businesses and processes. It is not a solution or model within production. We need to focus on customer activity, application and we need to focus on predictability and software for scenarios. This is related to the different business processes. You need to understand the maturity as well. There are many use cases. Most organizations are still in the beginning of the journey. They are still with one use case. The mature organizations build that and deployed already five AI use cases in production. And they're not in the stages of that deployment. These AI adoption models and use cases have stayed in production for at least two years on average, which is have generated a lot of revenue for the organization.
So, (?) we understood that AI focused on the bottom as you see on that paragraph of that diagram. So basically the bottom has the foundation of AI mature organizations relating the operated model that is being used to scale AI within the organization related to AI engineering practices and pipelines that should be put in place and definitely upscale because we learned in the previous surveys talent is a major issue. So upscaling is one way to resolve that gap in the organization as well as change management and definitely the government's aspect which is on trust and risk management and trust, risk and security. So the one on the top whether it is related to the AI use cases, AI trends, the next big thing and the AI models that are open influx when it comes to announcement, you know, this is something you should not be focusing on if you want to really mature your journey when it comes to AI. You need to focus on the foundational element as we will go through in a moment.
So as I said, you know, AI, you know, in the previous diagram, the ones on the top are really shy. You need to shy away from shiny objects. Delays the focus when it comes to foundational and fundamental components and helps you. Hopefully reach 2 years production when it comes to AI adoption. But if we talk about the scalable AI operated model and I am conscious of the type, so I will try to speed up. You know, when it comes to the operating model, scaling AI requires different AI operating model. Previously, central teams may have succeeded to maybe pilot AI in the organizations, but in order to really scale it, you need to think about hyper model where very central AI capabilities need to take place. They work in tandem and in collaboration with other business units within the organizations in a very specific sort of manner when you need to knowledge about that mature AI organizations. Distribute the AI budget across different sort of business project instead of being concentrated on one or two projects.
And this is one example of, you know, hybrid operating model. There are things related to AI strategy, AI architecture and, you know, some of the sent matter expert when it comes to AI domain, but you also have the edges, the business units and business processes where many of the innovation are being adopted and in order to adopt that, you need to upscale the team as we will see in a moment.
But there is no one size fits all. Every hyper model will benefit from one organization to the other. But at the same time, you need to centralize everything in one domain and think about, you know, what makes sense when it comes to hybrid in your organization. If you come from software and software engineer background, you will definitely relate to what I'm saying ‑‑ what I'm going to say now. In software engineering, you have the pipelines that helps you to, you know, develop and design specific components in scale to production. Set AI. You need to have a mechanism to help you manage AI design and deployment end to end with a very automated fashion in your organization and components in your organizations, you will not be able to scale every AI model that you adopt within the organization. So the main focus for AI engineering in general dedicate the AI team and should double down in AI engineering capabilities and practice in the organizations. Right? And you need to understand that these AI engineering practices and capabilities will help you even, you know, ready your data when it comes to AI adoptions. Traditional AI using different AI techniques that we talked about earlier.
And again, mature organizations double down and deploying AI solutions. If you focus on the solutions, you will relate to the compostable and reusable counts that you need to lay down if you want to become serious when it comes to AI adoptions in your organizations.
Now it is even more difficult to get it done. Okay. So basically AI design patterns very similar to software engineering design patterns. If you don't come from that field, they're nothing but Lego blocks you connect them like this or like that to come up with a specific shape that you have in mind, but basically AI design helps you to bridge the different use cases with the right AI solution architecture that you want to build for your organization. And they have one example in the next slide.
So this example relates to what we call retrieval research model or RAG retrieval augmented generation with the general ‑‑ GenAI large language model. Basically the retrieval augmented generation happens to serve multiple use cases within the organization whether related to customer or operational excellence or other sort of scenario like employee productivity or others. But basically this component here could be reused across many other AI adoption techniques and scenarios within the organizations and you can crop it with the right model that you can employ for the specific use cases that we ‑‑ you want to really scale in your organization. But basically, you know, built that as a Lego component or as a useable sort of component will help you to adopt other use cases in your organizations.
And the third sort of teaching from the survey that we have done is the focus on upscaling and change management. So upscaling alone, you know, focusing on the AI associate should not be your only concern. You need to think about how could, you know, adopt AI literacy programs within the overall organizations that help, you know, every associates within the organizations understand the capability of AI and how they can use it in their continuous. This is where GenAI literacy programs can help. But definitely change management and we will see in a moment how change management techniques can help you maximize the value related to your organizations across different spectrums and domains.
So as we see here from that case study, basically this, you know, case study or the lesson that I want to highlight in this case study is that you need to be systemic when it comes to upscaling your associates within the organization. I mean, you should not focus only on the pros who use and reuse and create AI models or become really strong prompt engineers, but also focus on the bigger sort of group that may not really need advanced capabilities when it comes to AI, but they need moderate sort of capability, but the general, associate having online courses distributed across will help you to really each every single associates within the organization. Basically layers sort of systemic approach when it comes to AI literacy or adopt AI in your organization is very important to reach out more people.
And again, change management is very important and change management, you know, techniques will help you to maximize the business outcomes when it comes to cost saving, customer experience or even activity for the employees. The applications you will not be able to reach very high, you know, sort of impact when it comes to different values when it comes to business outcomes.
And the last thing they want to focus is AI governance in general. Trust, risk and security management is one of the frameworks that we often highlight and it is basically focusing on the facts that governance is being applied by diverse role. AI associate or savvy people will not help you to reach high AI governance. You need to think about different dimensions and different perspectives that needs to be put in place by the diversity or fraud. And the budget authority especially when it comes to AI privacy and security is very important. They need to be own bide central units that help you to adopt the governance make an itch related and business impact when it comes to breaching or enabling AI privacy is very important.
So again, the AI framework will help you prism and apply the governance mechanism utilizing different components in the tourism technologies and connecting with AI systems and the organizational governance practices in the ‑‑ that you have in your organizations.
And the last thing I just want to highlight is that AI adoption faces maturity. We focus on what is above the surface rather than build strong roots that will help you mature with time and reuse the components as you go.
And this is ‑‑ we have done everything that we have side and the time is (?) when it comes to AI adoption and you need to provide models for using the right formula described earlier and for lessons gone through for creating hybrid in that case and utilizing AI engineering and upscaling literacy program as well as investing in AI as well as investing in AI trust and security in general. This is really, really important to push the needle forward and become AI mature organization. Thank you very much. We'll stay around if you have any questions. Yeah. Go ahead.
>> (off microphone)
>> That's fine because the people on the web they need to hear you, but I can hear you now.
>> You talk about 10% (?) (off microphone)
>> On the next big thing. They focus on the shiny object instead of the fundamentals. The different foundational components in your organization without this, right, you may be successful in one AI use case. You may be successful in one AI specific business process or specific unit. But you will not be able to scale different models crease the organization. It relates to AI maturity. If you want to become AI enabled organization different regional and different business units, you need the foundation. So they do not focus on the foundation. That's fine. That's fine. We can hear.
>> We now see we have some sales. (?) (low voice)
>> Very great question. Upscaling is very important as I mentioned before. Your question is very great. How we can (?). I am (?) this question coming. Usually when you look to what is happening, even me I am looking at what is next. Keep continuing what you're doing. Yeah. I was reading that it says what do we expect. Really the future is faster. So we have to upscale ourselves in the area that is interesting. So we cannot be (?) (low voice) thank you for the question.
>> (low voice) (off microphone)
>> The survey for the workshop. Thank you very much. (low voice)