Getting ready for Now Assist / ServiceNow AI

Matt Dodd · ‎08-26-2024

Getting Ready for Now Assist / ServiceNow AI

Implementing Now Assist is a strategic move that can significantly uplift your organization's productivity and improve the engagement experience for your users and customers. To ensure a smooth and successful implementation it can be helpful to take several readiness steps in preparation. After many many conversations with customers on this topic, a common ask is "What kinds of things can we do in preparation for it?"

There are also numerous documents and articles written on the more technical aspects of this question, however not so many that talk about some of the other considerations — which is the purpose behind this article.

📌 Updates - What's changed recently

Since this guide was last refreshed, ServiceNow has shipped Zurich (Q4 2025) and Australia (Q2 2026, GA May 2026). The biggest shifts are in:

AI Governance — AI Control Tower has matured from a visibility dashboard into a full governance layer with capabilities across AI Discovery, AI Risk & Governance, Security & Guardrails, Observability and Value measurement.

Model Provider Choice and model selection — you can select and route between our model providers (Anthropic/Claude, Azure OpenAI, Google Gemini) to ensure you're powering your AI decisions with the best model for each job, as well as make specific model selections.

Knowledge Center - A huge value-enabler for leveling-up your organizations knowledge by scanning for duplicates, AI-assisted merging, generating of new content to address knowledge gaps and much more. This helps provide better quality and coverage for your human agents, but also for your AI Agents.

Official Now Assist Readiness Evaluation application - A store application that's available free, that can help scan your instances and provide actionable insights into many of the topics listed below in your instance!

Recent branding of our AI Conversational Experiences to ServiceNow Otto. Launched at K26. Meet ServiceNow Otto

💡 ServiceNow also publishes an official Now Assist Overview covering all of the basics around Now Assist definitions and components. It's a great companion to this article.

🏗️ Pillar 1 — Foundation (get the data and search right)

📚 1. Knowledge Management

This is one of the key areas that will help make Now Assist successful. Ensuring that the knowledge base is up-to-date and contains relevant articles that Now Assist can leverage for AI search tuning will help provide accurate and actionable answers.

Leading Practices: All of the usual leading practices around Knowledge Management are especially applicable to Now Assist/AI readiness. This includes:
- Retiring old and out-of-date content so that Now Assist doesn't surface it to your users.
- Ensuring a feedback mechanism exists to continuously improve the knowledge. Encourage users to provide feedback on articles, which can be used to refine and enhance the content.
- Scheduling regular audits of your knowledge to identify gaps, outdated information to be retired, and areas for improvement.
- For more comprehensive tips specific to Knowledge, refer to this excellent community post: Best Practices to use your Knowledge Articles with Now Assist by @Ashley Snyder.
Access to content: Think about the audience for your knowledge content and ensure that the various user personas, departments and roles in your organization are catered for correctly (i.e. roles and user criteria) from a visibility perspective. Consider the industry you are in and which types of users will want to find/consume different types of information. Unless configured otherwise, AI will potentially surface answers from all knowledge sources that the user has access to. Setting up and using User Criteria can help contain searches to the right areas applicable to your users.
Typically we cater for requester or fulfiller/agent roles quite well, but don't always consider other differences — such as faculty/staff vs students in an education institution, or shop floor vs manager in a retail store. Technically they may be different types of 'Requesters/Users', however your organizational processes may differ between these personas, and you want to ensure Now Assist gets the right answers to the right people.
Review your Knowledge Creation Process: Consider which users should be able to leverage the skill to generate new Knowledge using Now Assist. You could potentially expose this skill more widely to augment your existing Knowledge team and reduce the backlog, instead of (or in addition to) just reporting knowledge gaps.
Consider External Knowledge Sources: If you have alternative sources of knowledge information, it's worth considering whether they're suitable to leverage directly alongside your ServiceNow Knowledge bases via our External Content Connectors. These can also be explored further on the ServiceNow Store, noting there may be licensing considerations.
Leverage Knowledge Center: This new and evolving area of the platform is the best starting place to uplift and augment your organizations knowledge and covers the bases mentioned in the previous steps and a lot more. @rob_martoncik has written a great article introducing this here, which is definitely worth a look. Introducing Knowledge Center with Advanced Editing, powered by Now Assist

🖥️ 2. Service Operations Workspace (SOW) Readiness

Most agent-facing Now Assist skills are designed for SOW, not legacy backend forms. Assess your SOW adoption — and any related customizations — before starting your Now Assist implementation. This is a common blocker that's easy to miss. If your agents are still working primarily in classic forms, plan that transition as part of your readiness.

(Credit: this point was raised by community member Yogesh Shinde in the comments — a great field observation worth promoting into the body of this guide.)

🛒 3. Conversational Catalog Support

If you're looking to allow your customers to request catalog items conversationally with your Virtual Agent, it's important to review and understand the supported variables for the conversational catalog skills. For the full list of supported conversational catalog options, refer to What makes catalog items non-conversational in premium chat.

Once the Now Assist plugins have been activated in your instance, you can use the Conversational Catalog Overview Dashboard to see a comprehensive overview of which catalog items are conversational already and (for those that are not) a detailed explanation of what's required to make them conversational.

A good approach is to review the most commonly requested catalog items and target any remediation efforts there for some quick wins.

Victor Chen has written a great article on this topic here, plus an AI Academy session you can watch here.

📊 4. Incident & Case Data Quality

Incident and Case Summarization is available on records with more than 50 words, though you probably won't have much need to use it on those small records — you'll likely be able to read them by the time the summary comes back anyway. The real value is on complex cases where there are multiple updates in the activity log, status changes, re-assignments, and so on.

Once the incident/case is ready to be resolved, the Resolution Notes Generation skill will really help summarize what happened and how it was resolved. This is where the importance of data quality comes in — it uses the information in the record (and potentially related records) to create a summary. That information is then made available to other agents to help solve future issues, and is surfaced by other in-platform AI (such as the Predictive Intelligence similarity framework) and served to agents working on similar issues. The vision is that Now Assist can help improve resolution note quality for resolving future issues. If you want to get really fancy you can also go on to create Knowledge Articles via Now Assist to help empower your end-users and up-level your Virtual & AI Agents who can use this information to help deflect issues and "shift left".

🔍 5. AI Search & Tuning

AI Search is at the heart of Now Assist — it's used to gather the data points that help Now Assist respond to user queries. AI Search gathers the relevant information visible to the user and submits it to Now Assist to generate its response (a process known as Retrieval Augmented Generation). If you're not using AI Search already, it's advisable to review and activate it prior to attempting to implement Now Assist.

Once active, there are a number of tuning options to ensure it's working optimally and looking at the right sources of information. It helps to get familiar with the principles of AI Search tuning to positively influence the results Now Assist provides. Result Improvement Rules are a good place to start.

For some great additional reading in this area:

🛡️ Pillar 2 — Governance (control, trust, and compliance)

🔐 6. Security & Data Handling

ServiceNow maintains detailed documentation around our security and data handling processes related to Now Assist and our Generative AI capabilities. Many customers have their own concerns around adopting AI solutions and naturally have questions about these topics. ServiceNow aims to be as transparent as possible. Please refer to the following:

Data Processing for Advanced AI & Data Products FAQ
Overview of Advanced AI and Data Products: Architecture, Applicable Terms and Data Handling
Data Security Addendum
Data Processing Addendum
International Data Transfers FAQ
ServiceNow Responsible AI FAQ
ServiceNow Responsible Ai Whitepaper
Now Assist Guardian— worth getting familiar with the guardrail/safety layer that helps keep AI behaviour within approved bounds.

👤 7. AI Administrator Roles

Consider who is the appropriate person in your organization to manage and own the Now Assist capabilities. Your system administrator may or may not be the right persona. Many Now Assist capabilities have new roles associated with them too, so there's flexibility in how you empower your teams to leverage these new capabilities. You may also want to think about which users/roles you want to allow to use particular skills such as Summarization or Knowledge Generation. You can see how to enable and configure the skills in the Now Assist Admin settings.

🧭 8. AI Data Steward role & AI Control Tower

The AI Control Tower (AICT) is now the central governance foundation for enterprise AI, and a key area to understand as part of your readiness. The AI Data Steward persona helps shape the governance aspects of your company's adoption of AI, as well as the choice of allowed Model Providers. (see next section)

👉 Practical takeaway: if you're deploying AI for the first time, validate your governance foundation before enabling any skill. If you're upgrading from Zurich with existing AI deployments, review any ungoverned integrations now, before they become enforcement violations. See the AI Control Tower documentation for more.

🧠 9. Model Provider Choice

You can now select and route between AI model providers — including Anthropic (Claude), Azure OpenAI, and Google Gemini — via AI Control Tower. As part of your readiness, decide:

Your default model strategy and which workloads (if any) justify a specific provider.
Any data-residency-driven routing requirements.
Which providers you want to allow-list for Now Assist Skill Kit and AI Agent Studio experiences.

⚠️ In-country / sovereign customers please note: Not all model providers are available for customers with in-country SKUs (e.g. Australia IRAP-Protected / SPP-AU, FedRAMP, NSC DoD IL5), self-hosted, or other restricted environments, and some Now Assist features may be unavailable in those environments. Check the latest availability before committing to a model strategy — refer to KB1584492 and KB0743854 in the Now Support Knowledge Base, and watch for updates each release.

⬆️ 10. Instance & Store App Versioning

Getting access to all of the latest and greatest Now Assist skills generally requires your instance to be on one of the most recent releases. Being on the cutting edge of this technology means staying on the latest release of the platform, so to get the most value you'll want to be current. Luckily, ServiceNow makes it easier than ever to get your upgrades done faster and smoother with each release.

Additionally, we recently moved to Upgrade Suites which help package all compatible applications into a larger suite to upgrade them all at once which can help to reduce compatibility issues and version mismatch issues.

🚀 Pillar 3 — Adoption & Scale (roll out and grow)

🔧 11. Instance Customization

Sometimes it's necessary to tweak the out-of-box configuration to meet business needs. If you've gone down this path, you may wonder how much impact it has on your ability to adopt Now Assist Generative AI capabilities. For example, say you decided to use a different field in your incident table, or added a different 'state' to your process — you may also want to check for changes to the out-of-box UI Actions or the resolution workflow, as these may affect how you adopt the new Now Assist skills. This is the kind of thing to be aware of across each of the tables in the platform that will have a Now Assist capability added to it.

To help with this, ServiceNow released the ability to make a copy of certain out-of-box prompts to accommodate these kinds of changes. The copied skill allows some basic tweaking of the inputs used and the role conditions that control when the skill can be used, plus the ability to add additional input sources (related tables or emails) if required.

For capabilities beyond basic configuration tweaks, the Now Assist Skill Kit allows greater freedom to create your own custom-built skills, or modify some of the out-of-box skills too. This is a complete topic of its own — there's a great article (Now Assist Skill Kit (NASK) FAQ) you can use to explore it further.

💬 12. Transition to Now Assist for Virtual Agent (from the legacy NLU)

As part of Now Assist, we've upgraded Virtual Agent from NLU-based version that has been available for some years. There are some considerations here as below:

Compatibility: Each Portal you're using can only have one Assistant assigned to it, which means you have a choice to make for each. Do you want it to be an NLU-based Virtual Agent (legacy) or a newer LLM-based Virtual Agent using Now Assist? If you're looking to transition from your current NLU-based VA, read on. If you're new to Virtual Agent, you have a clean slate to start with and can skip the next step.
NLU Topic Migration: If you've spent a lot of time building your own topics and conversations for your VA NLU deployment, you'll typically want to migrate these across to the new LLM framework. There's a Topic migration utility to help. I'd suggest you don't just blindly migrate everything, but evaluate carefully where its actually needed, as the LLM-powered assistant can often handle things better without needing a specific topic. @Victor Chen has written a fantastic article, Migrating NLU Topics to LLM, detailing why you may want to consider this transition, plus a comparison of the two technologies.
Tracking the value from deflections: Value is an important consideration for any new investment, and capabilities are emerging that allow you to analyze and assess the value derived from interactions with the Virtual Agent.
One or Many Assistants?: Although you can only have one Virtual Agent per portal, there may be a need to create multiple 'Assistants' (think of an assistant as a Virtual Agent profile, configured independently but using the same underlying LLM technology) associated with different portals. Each assistant can have access to different information sources and different 'topics', giving you flexibility in how you deploy across audiences and portals.
🎙️ Now Assist for Voice: The Australia release expands Now Assist for Voice with AI-driven multilingual assistance, authentication, and real-time insights. If you support phone/voice channels, this is worth factoring into your VA strategy.

👩‍💻 13. Developers & Creators

Good news — there's a wide variety of capabilities to make your developers and creators more productive. Check out the latest release notes for the most up-to-date skills your creators can leverage. Many of these are accelerators and may not require the same planning and consideration as other areas before deployment — but it's well worth considering which areas to target first once these super powers are enabled. You may also want to check out how you can start vibe-coding with Claude Code and the ServiceNow SDK.

A couple of additional resources to get started in this area:

Now Assist for Creator Welcome Guide
Usage Tips/Use Cases/Example Prompts for Now Assist for Code
Now Assist for Creator Product Hub — check here for lots of other great content.

🆕 Watch this space: Build Agent has evolved significantly through the Zurich and Australia cycles — from an IDE-only preview into a Studio-native experience that can now create AI agents, skills, and agentic workflows, generate and run test coverage, and support MCP client connections. If your creators are keen, this is one of the fastest-moving areas of the platform.

🔀 14. MCP & External Tooling

MCP (Model Context Protocol) Client is now generally available. This lets your AI Agents and Build Agent reach approved external tools and data sources, opening up a much wider range of agentic use cases. From a readiness perspective:

Inventory which external tools and data sources you'll want to connect.
Confirm they're approved through your governance process — under Australia, unapproved MCP servers are hidden from AI Agent Studio.
Treat MCP connections as part of your governance and security review, not an afterthought.

🤖 15. AI Agents & Agentic Workflows

This is a whole topic on its own. Once you've nailed the basics covered here, you can start thinking about deploying AI Agents and leveraging agentic workflows.

If you're interested in getting started, identify some relevant use cases for your organization. A few approaches:

Explore the out-of-box options — review the AI Agent agentic workflows and see if anything resonates or sounds similar to what you need. This is a quick win — use as-is, or clone and configure to suit.
Leverage the AI Agent Advisor to mine your instance for your common use-cases to address and suggest the best available AI Agents and Agentic Workflows that can be used to map to those use-cases.

📚 Additional ServiceNow Resources

Docs – Release Notes: Now Assist landing page (Australia)
ServiceNow Community: Now Assist & Generative AI Product Hub
ServiceNow Expert Services: Put AI to Work with Now Assist
ServiceNow University: Now Learning Courses for Now Assist and AI Agents
Now Create: Now Assist Implementation Guides
ServiceNow Community: Getting Started with AI Search
ServiceNow Community: Knowledge Management User Group (with Video)
ServiceNow Community: Comprehensive Resources list for Now Assist Implementations

💼 Impact Services

Customers subscribed to ServiceNow's Impact offering can leverage a number of Accelerators to help get ready for Now Assist and AI Agents. Impact customers can view all available Accelerators from the Accelerator catalog.

So in summary, there are a lot of factors to consider when embarking on your Now Assist journey, and some of the topics above are considerations you may choose to go deeper on in preparation. I hope you find this useful — and please continue to add your own field learnings in the comments below. 🙂

priyaradhak · ‎08-28-2024

Great article! Thanks Matt!

Yogesh Shinde · ‎01-06-2025

Great article! One of the key aspects is the adoption of Service Operations Workspace (SOW), as most of the Now Assist Skills are designed to be used through SOW rather than backend forms. Therefore, SOW and its related customizations are crucial to check during the pre-assessment or instance assessment before starting the Now Assist implementation.

Additionally, it's essential to properly implement and adopt AI Search and Virtual Agent. AI Search is the most critical and widely used functionality since it is accessible to all users, unlike Agent skills which are limited to agents. I plan to add field learnings to this article and encourage others to do the same. Now Assist is an excellent product!

AshishKM · ‎06-03-2025

This is great summary about preparation on Now Assist learning.

I landed on this page from "Now Assist ITSM" course's best practice section.