Step 1: Create or choose an AI role
Start by considering the use case of the AI service you’re about to build. Is it an assistant to be interacted with via chat? Is it a developer assistant or a data-based service?
By creating a specific AI role at the beginning, you will be able to properly scope the AI’s access to the data source you will be creating and ingesting in step 2.
Remember: the more scoped, specific and defined the model is, the better accuracy the results of your AI service will have. It will also limit hallucinations and provide better governance.
- Name your AI role, add the description, select the AI profile, add any relevant tags and give your AI role an image for easier browsing later on.
- Depending on what kind of AI service you are building, you choose the relevant AI role profile. In the dropdown for AI profile, there are several predefined AI roles that are built on top of profiled models trained and aligned to perform specific tasks, rather than a generic, generalized model. Choose from the following AI profiles: assistant, tool, data analyst, operator and developer assistant.
- Once you've uploaded and ingested your data source for this service, you must select it under the Sources list.
- Should your service also include an action, select it here.
- Finally, select any relevant policies or rules that the AI role should adhere to.
Click create and proceed to step 2.
