This guide sets out what you need to do to get started with the Enterprise AI Management Solution. It should be read in conjunction with Enterprise AI Management: Metamodel and Enterprise AI Management: Purpose, Scope and Rationale.

Your first step is to deploy the Solution by selecting it from the dropdown list of Solutions on the Ardoq Solutions page and hitting the Add solution materials button. This will deploy the Solution to your organization, including all the relevant assets. The Solution uses the following Workspaces:

Any that are not already present will be created for you when you add the solution to your organization. At first sight, this might seem quite daunting, but if you have no need to use some of these workspaces (for example because you don't wish to describe your supporting infrastructure in Ardoq), you can leave them empty and the Solution will work fine.

We have seen how one Technical Capability, named Artificial Intelligence, is a pre-requisite to recording AI Systems across your portfolio. You need to add this to your Technical Capabilities workspace. And if you wish to classify AI Systems according to different AI technical capabilities, you will need to add some children to this component representing those different AI technical capabilities.

If you only wish to have a single Artificial Intelligence Technical Capability, you can add this manually to your Technical Capabilities Workspace. If you already have a technical capability model there, simply add a new Technical Capability component and give it the name “Artificial Intelligence”. You can place it anywhere in your existing model, but remember that any descendents of that component will automatically be identified as AI Capabilities (i.e. their AI Capability field will set itself to True). If you don’t have any other use for a technical capability model, and don’t wish to classify AI Systems according to different AI capabilities, just create a single “Artificial Intelligence” Technical Capability component in the Technical Capabilities workspace.

If you have your own AI technical capability model that you wish to define, follow the above instructions for Option 1, then add children, grandchildren etc. to your “Artificial Intelligence” component to build out your own AI technical capability model.

Ardoq has made a sample technical capability model available, which includes both AI capabilities and non-AI technical capabilities. To import this model into your organization, select Imports and Integrations from the home menu and select Frameworks & Resources. Among the resources available for you to import is one called Technical Capability Model (including AI Capabilities). Clicking through to complete the import will create a new workspace called Technical Capability Model containing the complete sample model. You now need to copy all the elements of this model into the Technical Capabilities workspace. The simplest way to do this is to open both workspaces (Technical Capabilities and Technical Capability Model), select all the top-level Technical Capability components in the Technical Capability Model workspace and drag them to the Technical Capabilities workspace. This will copy them to the intended destination workspace. Having done this, you can now delete the Technical Capability Model workspace.

You may already have a technical capability model in the Technical Capabilities workspace, but just want to add the sample AI capabilities to it. Or you may only want a technical capability model that consists just of AI capabilities. In both situations, follow the process outlined in Option 3 above, but instead of selecting all technical capabilities in the source workspace, just select the top-level capability named “Artificial Intelligence” and drag it to your Technical Capabilities workspace. This can be added in any part of your existing capability model, if you already have one. As above, you can now delete the Technical Capability Model workspace.

The capabilities you have just copied into the Technical Capabilities workspace will all now have acquired an AI Capability checkbox field. This is a calculated field that will automatically recalculate its value within an hour. If you plan to make use of the model before then, you should force the system to recalculate this field for all components in the workspace by hitting the Recalculate button for that field in the workspace editor.

If you’ve deployed a Technology Catalog with the Technology Portfolio Management Solution, which is part of the Governance, Risk & Compliance Suite, you can include your catalog in the scope of this Solution. This will enable the Technology Product components in your catalog to be identified as AI Systems in exactly the same way that Applications and Technology Services are.

This Solution introduces a new field, AI System, that is added to all components that represent software assets that can be identified as AI Systems. Deploying the Solution automatically adds this field to the Application, Application Module and Technology Service component types. But since the Technology Catalog is not automatically covered by the new Solution, you need to manually add the AI System field to the Technology Product component type in the Technology Product Catalog workspace. See What is a Field for details of how to add a field to a component type.

Once you’ve done this, Technology Product components will have their AI System field set to True if:

Note that there is no need for you to edit the Gremlin code of either of the calculated fields to achieve this. They already contain the necessary logic to be aware of Technology Product components and their role in the metamodel.

With this change in place, any Application, Application Module or Technology Service that is a deployment of a Technology Product that is an AI System will automatically be recognized as an AI System (i.e. have its AI System field set to true). In effect, they “inherit” their AI capability from the catalog entry, and do not need a separate Is Realized By reference from an AI Technical Capability to be identified as an AI System.

You also need to add your Technology Catalog workspace to the Solution’s two Reports so that they are included with the reports’s scope. You do not need to change any of the gremlin code in either report, just add the workspace. Edit each report with the Report Builder, click in the box headed Which workspaces should your data come from?. Start typing Technology Product Catalog, and the option will appear in a drop down list. Select it, and it will be added to the other workspaces from which the report’s data will be drawn, then click Save changes to save the new version of the report. Do the same thing for both reports in the Solution: AI Capabilities and their realizations and AI Systems.

If your Technology Catalog is populated with technology products that have not yet been deployed (representing technologies that are available to you, but not yet in use), you are able to use their AI System fields to plan your adoption of new AI capabilities, and the technologies you have at your disposal to realize them.

You may also wish to update the Viewpoints provided with this Solution that you use so that AI Technical Capabilities that are realized by components via the Technology Products they deploy are shown in views. Below is an example of the paths you can add to your selected Viewpoints:

In the space of just ten to fifteen minutes, you have successfully deployed and configured the Enterprise AI Management Solution. Well done!

Now comes the part that needs the knowledge of you and your colleagues: identifying which of your current IT systems implement AI technical capabilities, and should therefore be classified as AI Systems. If you have a Technology Catalog, having deployed the Technology Portfolio Management Solution, you need to identify the Technology Product components that realize AI technical capabilities. If you are not using a Technology Catalog, you need to connect the AI technical capabilities directly to your deployed Application, Application Module and Technology Service components.

Either way, you will need to review your portfolio and determine which components among them realize one or more Artificial Intelligence capability. You can choose to conduct this review in one central operation using Ardoq to work through all the relevant components, or alternatively distribute this task among the owners of, and experts in, your applications and technology services.

One way to manage this review with Ardoq is to open the portfolio workspace(s) and the technical capability map together. If you have a Technology Catalog, that means opening the Technology Product Catalog and Technical Capabilities workspaces. If you don’t have a Technology Catalog, then you should open the Applications, Technology Service Repository and Technical Capabilities workspaces. Start by selecting the Artificial Intelligence component in the Technical Capabilities workspace, and select a Dependency Map view. Now expand your Applications or Technology Product Catalog workspace so that all its components are listed in the left-hand panel. From hereon, we’ll follow the process using the Applications workspace, but the same process applies if you’re wiring up your Technology Catalog. You should see something like this:

It is now a simple job to work through the list of Application components and decide if any of them, or their constituent Application Module components (if they have any), realize one or more AI technical capability (of course, this assumes that you have the knowledge in the room to know which AI capabilities are realized by each component in your IT estate!). If you recognize one that does, select the capability from the map in the main window with a left click, then right click and choose Create Reference. This allows you to drag a link from the capability to the Application that realizes it. Clicking on that Application will open a Create reference panel, allowing you to create your Is Realized By reference. Remember to hit the Create button to save the new reference. You can work through all your Application, Application Module and Technology Service components in this way. When you have finished, you will have successfully classified all your IT Systems according to the type of AI they realize.

An alternative to reviewing all your applications and technology services in one centralized review is to distribute the task of wiring them up to AI Technical Capabilities to their owners and experts. Two simple surveys have been provided to facilitate this: Review Application's AI Capabilities and Review Technology Service's AI Capabilities.

Two corresponding Survey Broadcasts are also included with the Solution which you can use to automatically distribute these survey to their corresponding owners and experts. For applications, survey links will be sent to all people who have an Is Expert In reference to the application. If there are no experts for an application, the survey link will be sent to the application's owner. For technology services, the surveys will be sent to their owners.

The default message sent to applications owners or experts is shown below. This can be easily changed to suit your organization's communications style using the Broadcast Builder.

Note also that both broadcasts have been set up as single broadcasts, to be used in a one-off exercise. It is a good idea to keep this data up to date with a periodic review, and we recommend you do this by adding a Technical Capability question to the Review Application technical details Survey. You'll see details of how to do this in the Make this sustainable section at the end of this article.

If you are using the supplied Ardoq AI Technical Capability Model, it is worth noting that it is divided into two parts: AI Functionalities and AI Technologies. As indicated by their classification, those capabilities that are in the Functionalities capability sub-tree can be used to describe what the AI System does. Those in the Technologies sub-tree describe how the AI System has been implemented. You might choose to use both together, for example to describe a chatbot both in terms of what it does (realizing the Conversational Systems Technical Capability) and how it does it (realizing the AI Agents Technical Capability). You might find some components are best described with one or the other part of the model. For example, a Large Language Model, such as those supplied by OpenAI, might be integrated with different applications that use it for different functions. You could represent this situation by having the OpenAI Application realize the Large Language Model capability, while the Applications that integrate with it might each realize individual functionality capabilities that best describe what AI functions they offer to their users (e.g. Text Generation, Image and Video Generation, Conversational Systems, Knowledge Representation and Reasoning).

You will recall that these component types now have a calculated Checkbox field, AI System. Within an hour of finishing this exercise, that field will automatically be set to true for all the components you’ve just “wired up” from AI capabilities. If you’re keen to see the results before waiting, you can force Ardoq to recalculate the AI System field in each workspace by hitting the Recalculate button for the field in its corresponding workspace editor.

Your graph now knows which systems in your estate directly realize AI capabilities. You can analyze this from many perspectives using the Presentation, Dashboard and Viewpoints provided with the Solution. But you should consider whether you have any other Applications that integrate with these systems, and which might therefore be considered to be, themselves, AI Systems, by virtue of their consumption of AI capabilities via those integrations. This might be particularly important if your organization is subject to regulatory requirements regarding the use of AI systems, in which case you must be careful to ensure that you’ve marked out, and are governing, all systems that fall within the scope of the legislation concerned.

The Solution comes with a Viewpoint that will help you identify such Applications. It shows you all Applications that integrate with AI Systems (Applications, Application Modules, Technology Services). Open the Viewpoint Non AI Applications that integrate with AI Systems and select all Applications. You will see something like this:

The top row shows all your non-AI Applications that are integrated with AI Systems, and below, the details of the AI Systems they are integrated with, and the AI Capabilities that those AI Systems realize. You should consider each of the non-AI Applications in turn and determine whether they should really be classified as AI Systems as a consequence of their integration with AI Systems. If you decide that any of them should be classified as an AI System, you will need to add an Is Realized By reference from a relevant AI Technical Capability to each of the Applications concerned. If you are using the sample AI Technical Capability Model provided with Ardoq, we suggest you choose one or more technical capability that best describes the function it provides from among those listed under AI Functionalities.

You’ve just completed a one-off exercise to identify and classify all your AI Systems, enabling you to make use of the assets provided with this solution to manage your organization’s adoption and use of AI to maximize effect (while minimizing risk). But how will you make sure this information stays up to date? Will newly adopted Applications also get classified correctly? Perhaps updated versions of existing applications will introduce new AI capabilities that need to be recorded. You need to put in place a sustainable approach to maintaining this information in your repository.

If you’ve adopted the Foundation approach to managing Application ownership, you will be using the Survey Review Application technical details. This is sent regularly to the owners or technical experts of all Applications. Now that you’ve adopted a Technical Capability model, we recommend that you update this survey to include a question that asks which Technical Capabilities are realized by the Application concerned.

Now every time your application owners or technical experts are asked to complete this survey, they will also be confirming the set of Technical Capabilities that each application realizes, including any AI capabilities.

And if you’ve launched the Broadcasts associated with the Foundation Application Ownership process, any new Applications added to the system will automatically be picked up and have the survey sent regularly to the appropriate colleague, ensuring that your knowledge of AI Systems across your estate remains up to date. For more information about the Foundation application ownership process, see Process Playbook: How to Automate Application Ownership.