Exploring AI Skill Kit
Summarize
Summary of Exploring AI Skill Kit
The AI Skill Kit plugin for Otto allows ServiceNow customers to create and activate custom prompts and skills that extend Otto's generative AI capabilities beyond its base system skills. This enables greater flexibility to tailor AI-driven workflows and interactions to specific use cases within the ServiceNow platform.
Show less
Prerequisites and Considerations
- Users building custom skills should have a solid understanding of prompt engineering and large language model (LLM) behavior, including the probabilistic nature of output and variability across models.
- Effective skill development requires testing and refining prompts against representative data from your ServiceNow instance rather than relying on single examples or assumptions about prompt interpretation.
- Defining the use case and target persona for the skill is essential before beginning development.
Getting Started
- Update Otto plugins, such as the ServiceNow Otto for ITSM plugin, to the latest release via the Application Manager to access AI Skill Kit features.
- Assign the snskillbuilder.admin role to all users who will develop skills using AI Skill Kit.
User Roles
- AI Developer: Creates and refines custom skills and prompts, configures skill settings, and requires the snskillbuilder.admin role. Experience in prompt engineering and LLM fundamentals is necessary.
- Otto Admin: Reviews and activates published skills to make them available across configured touchpoints. Requires the admin role.
AI Skill Kit Workflow
- Define the provider: Choose and understand the benefits and limitations of the LLM to be used.
- Build the prompt: Design prompts based on a clear understanding of your Otto instance’s architecture and data sources.
- Test the prompt: Use AI Skill Kit’s built-in editor to iteratively test and refine your prompts.
- Deploy the skill: Publish your skill to ServiceNow Otto Panel, Otto Context Menu, Virtual Agent, Flow Action, or UI Action.
Benefits for ServiceNow Customers
AI Skill Kit empowers ServiceNow customers to design tailored generative AI solutions that seamlessly integrate with existing workflows, enhancing operational effectiveness and efficiency. Key benefits include:
- Custom skill creation tailored to unique business needs.
- Prompt editing and configuration for data-driven AI interactions.
- In-editor testing and iteration to ensure skill accuracy before activation.
- Flexible deployment options within the ServiceNow platform.
Next Steps
To deepen your understanding and effectively apply AI Skill Kit, explore ServiceNow documentation on configuring and using the AI Skill Kit for guided instructions and best practices.
Use the AI Skill Kit plugin for Otto to create and activate custom prompts and skills for Otto.
AI Skill Kit overview
Use AI Skill Kit to create custom skills when base system Otto skills don't fit your needs. Custom skills enable you to have greater flexibility with Otto's generative AI capabilities.
Before you build a custom skill
Because you write and refine the prompts that drive your skills, you should be comfortable with the fundamentals of prompt engineering and with how a large language model (LLM) behaves.
- How an LLM produces output, including its probabilistic nature and the fact that the same prompt can produce different results on different models.
- How to write, test, and refine a prompt based on the output it produces, rather than on how you expect the model to interpret your wording.
- The use case you want to solve and the persona you're building the skill for.
Effective skill development depends on testing the prompt against representative data from your instance and refining it based on the results, not on a single example. For the full set of guidelines and the phases of building a skill, see General guidelines for AI Skill Kit. For help defining requirements and outcomes before you build, see Scoping the skill.
Get AI Skill Kit
To use AI Skill Kit, you must update your Otto plugins in the Application Manager. For example, update your ServiceNow Otto for ITSM plugin to the latest available release.
You must also assign the sn_skill_builder.admin role to anyone who uses AI Skill Kit.
AI Skill Kit users
| User | Description |
|---|---|
| AI developer | Creates new skills, writes and refines the prompts, and configures the skill settings. This user must have the sn_skill_builder.admin role. Building an effective skill requires prompt engineering experience and an understanding of LLM behavior. |
| Otto admin | Reviews and activates published skills so that they're available at the configured touch points. This user must have the admin role. |
AI Skill Kit workflow
The following diagram shows the user journey for AI Skill Kit.
- Define the provider
- Understand the benefits and potential downsides of each large language model (LLM) that you're considering using.
- Build the prompt
- You must have an understanding of the architecture of your Otto instance and be able to define where input data should come from. You should also have an understanding of LLM fundamentals to build an effective prompt.
- Test the prompt
- AI Skill Kit enables you to test your prompt from the editor.
- Deploy the skill
- AI Skill Kit enables you to deploy your skill to ServiceNow Otto panel, ServiceNow Otto Context Menu, Virtual Agent, Flow Action, or a UI Action.
AI Skill Kit benefits
AI Skill Kit enables you to design your own custom generative AI functionality that is then easily deployed into the ServiceNow platform. Custom skills can augment workflows with generative AI to increase effectiveness and efficiency.
| Benefit | Feature | Users |
|---|---|---|
| Create custom solutions by building a custom skill or workflow. | Create a skill | AI developer |
| Create and edit prompts for skills and configure where you want to bring in data from to augment your prompt. | Create a prompt | AI developer |
| Test and iterate on your skill before activating it. | Test a prompt | AI developer |
What to explore next
To learn more about configuring and using AI Skill Kit, see: