Collibra in action
1. Collibra in action
Hello, everybody. And what you see here on the screen is our AI governance module. It's really important here that we give a comprehensive overview of all the AI use cases per status, per life cycle status in your organization. For example, I can see that there is a credit score calculation engine which is in the monitoring stage, but that here a customer support engine is being worked on for the sales department. That's really interesting. I also can make sure I don't duplicate my efforts by having a look at all the AI use cases in my organization. There is also a prominent button here, the register AI use case button. This button is accessible to everybody in the organization so that you can easily create a use case. So, let's, for example, come with a new model that needs to be built, the customer churn dashboards for fields. I'm gonna give it a description and that's how easily I create a new AI use case and that certain activities are being starts. Here it's important that we understand who is gonna do what type of activity. This can be tracked in the life cycle tracker and per stage in the developments of this AI use case we're gonna have multiple activities. For example, the business context assessments is our first activity. I could assign this to someone to fill in, for example, and that's gonna ask questions about what is the business problem that we're solving for, what is the value of your use case, what is the target audience, and so on. And whenever you want to add a decision that has been taken in your AI governance council, for example, you can easily add the decision records on a certain decision that has been taken and that you want to check on. So whenever we move to from one life cycle step towards the other, you might want to add a decision gate comment, adding the reason why you feel confident to move to the next stage. In order to see an AI use case that has been going through this whole AI life cycle, I'd like you to see the credit score calculation engine. Here you can easily see the information on why we're building this AI use case, what it's about. I could, for example, see all the business context information, what is the business case, the business value, the use case application, so this is customer facing clearly, but I also see the models that are connected towards it. What models are being used in order to build this AI use case. Furthermore, we can go go deeper on the data. What is the training data that is being used, what is the inference data and output data. Of course, knowing CIA, these are relationships, so you could easily go deeper and understand more about the data by hovering over it and going to that asset. Here in the data flow, you can't really understand which data is getting into what AIU case and how they are being interconnected towards each other. For example, the classification model is being used for the credit score calculation engine. But I also see that there is customer data that is used to train this AI use case. The output of the AI use case is now another dataset which is being used in other types of use cases. So it gives me a really good overview on how everything comes together. Finally, let's have a look at the model and agent register. Here I can get all the information on all my models across my different AI systems. I can see that I have hugging face models, open AI models, GCP models. I can see why we are having them, the owner, model type, and who is providing them. I also can see which of the models are not used, which models are not being connected towards a certain AI use case, for example. This can help to reduce costs because otherwise we would have models that are running longer than needed or just running in silence. And of course we have the agent register where I can see the different agents in my organization. So, here I can see my Microsoft customer support agent. I could easily see the description. I can get information on the instructions, how it has been built, and what the instructions are in order to take certain actions in this agent, what is the use case it's connected towards, and the model that is backing it up. I could easily go to this model and go back to the model register in order to understand all details from that side. All right, thank you for following along. This really gives a good comprehensive view on our use cases, models and agents that are working along each other in order to get AI done in a company in a proper and efficient way.2. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.