ServiceNow’s Now Assist brings the magic of generative AI directly into your workflows with ready-to-use skills. But what happens when your organization has unique needs that go beyond the standard offerings? That’s where you, the developer, come in, armed with the Now Assist Skill Kit (NASK).

Think of NASK as your specialized toolkit for crafting tailor-made generative AI skills right within your ServiceNow instance. It allows you to build solutions for specific AI use cases that aren’t covered out-of-the-box.

Configuration First, Customization Second! Before you dive headfirst into building custom skills, remember ServiceNow’s golden rule: favor configuration over customization whenever possible. Explore the standard Now Assist skills first – you might find that configuring their existing properties gets you exactly what you need. Why? Because custom skills require ongoing maintenance, testing, and updates, all of which become your responsibility, not ServiceNow’s. Build custom only when truly necessary.

Getting Started: What You’ll Need

Ready to build? Make sure you have the following in place:

• Instance Version: NASK was initially introduced with the Xanadu release and continues to evolve. Check the Now Assist Skill Kit documentation for version compatibility.
• Latest Plugins: Ensure the most recent Now Assist plugins are installed and activated on your instance.
• The Right Access: You’ll need the sn_skill_builder.admin role assigned to your user profile. Without it, you won’t even see the Skill Kit interface in your instance navigator. Details on roles are typically covered within the main feature documentation.

Inside the Skill Kit: Your Building Blocks

NASK provides a structured environment with key components you’ll work with:

1. Skill Inputs: Feeding the AI Context

To get relevant results, your AI needs context. Skill Inputs let you pull data directly from your ServiceNow instance.

• You define the type of data (e.g., Record) and specify the source table (e.g., the kb_knowledge table for Knowledge Articles).
• Pro Tip: Select a test record while setting up record-based inputs. This lets you easily test your skill against real data during development.

2. Prompt Editor: Crafting the AI’s Instructions

This is the heart of your skill! The Prompt Editor is where you write the specific instructions (the “prompt”) that tell the Large Language Model (LLM) what you want it to do and how to do it.

3. Prompt Augmentation: Making Prompts Smarter with Tools

Static prompts are good, but dynamic prompts are better! NASK lets you use Tools to fetch and inject real-time ServiceNow data directly into your prompts before they’re sent to the LLM. You can use:

• Flow Actions & Subflows: Leverage existing Flows built with Flow Designer or build new reusable ones (good practice!) to grab specific information (like a list of company policies).
• Scripts: For more complex data gathering or manipulation needs, write scripts (e.g., Server-side Scripts) to fetch and format information for your prompt.

4. Choosing Your Engine: LLM Selection

NASK gives you options for the brainpower behind your skill:

• Now LLM Generic Model: Use ServiceNow’s own powerful, general-purpose text-to-text model.
• External Third-Party LLMs: Integrate with other LLMs (like those from OpenAI or Azure). Heads up: If you go this route, you are responsible for all backend configurations, including API keys, endpoints, and ensuring the connection works. Check the documentation for specifics on LLM configuration within NASK.

5. Testing, Testing, Testing!

NASK includes built-in testing capabilities. Use records from your instance to try out your skill, refine your prompt, and ensure it behaves exactly as expected before unleashing it.

Bringing Your Skill to Life: Deployment

Once you’re happy with your creation:

1. Publish: Hit ‘Publish’ within the Now Assist Skill Kit interface.
2. Activate: Head over to the Now Assist Admin console. Find your custom skill (it might be under a specific ‘Custom Skills’ section) in the ‘Available’ tab and flip the switch to activate it. It won’t work until you do!

Putting Your Skill to Work: How Users Trigger It

How will people actually use the custom skill you built?

• UI Actions (Initially): In early releases, the main way to launch a custom skill was through UI Actions. For example, adding a button to a Knowledge Article form that triggers your “Compliance Check” skill.
• More to Come: Expect ServiceNow to add more ways to trigger custom skills in future releases. Keep an eye on the release notes!

Life After Launch: Ongoing Care

Your job isn’t quite done after deployment:

• Monitor Performance: Keep an eye on how your skill is doing. Use tools like Now Assist Analytics to see how often it’s used and if users find it helpful (acceptance rates).
• Improve & Maintain: Use performance data to make your AI solutions even better over time. Remember, maintaining the custom skill, especially testing it after platform upgrades, is your team’s responsibility.

Take aways

By mastering these components within the Now Assist Skill Kit, you can extend ServiceNow’s AI capabilities to perfectly match your organization’s unique processes. Just remember to build wisely, prioritize configuration, and plan for the long-term maintenance of your custom creations!