What is Now Assist and Why You Should Care

Remember Virtual Agent? Of course you do. It was ServiceNow's chatbot solution, and for a while, it felt pretty clever. You'd spend weeks building conversation flows, training Natural Language Understanding models, mapping intents, defining utterances, and tweaking entities. It was proper work. You'd create intent after intent, train the model with examples, test it, tune it, test it again.

And when it worked? Brilliant. Users could reset passwords, check ticket status, and order laptops. All self-service, no agent needed.

But let's be honest about the reality. Building and maintaining those NLU models was exhausting. Every new use case meant more training data, more utterances, more model tuning. Subject matter experts spent hours helping you understand the variations of how people might ask for things. And even after all that work, the conversation still felt a bit rigid. Users had to phrase things just right, or the bot would hit them with the dreaded "I'm sorry, I didn't understand that."

It worked, but it was limited. Structured. Predictable. Very much built on decision trees and keyword matching wrapped in some machine learning.

Then ChatGPT Happened

November 2022 changed everything.

Suddenly, everyone had ChatGPT in their pocket. People were having natural conversations with AI. No intent training. No utterance mapping. Just pure, fluid conversation that actually understood context. You could ask questions in any way you wanted, and it just got it.

The world saw what conversational AI could actually be, and there was no going back. Users who'd patiently accepted the limitations of traditional chatbots now expected something completely different. They wanted that ChatGPT experience everywhere, including in their enterprise tools.

Every major platform saw the shift happening. Microsoft rushed Copilot into everything. Google launched Bard. Salesforce announced Einstein GPT. The AI arms race was on.

ServiceNow had a choice. Stick with an NLU-based Virtual Agent and watch users get frustrated by the gap between what they experienced elsewhere and what they got at work, or completely rethink how AI works on the platform.

They chose to rebuild. Properly.

Enter Now Assist

Now Assist isn't Virtual Agent 2.0. It's a complete rethinking of how AI works in ServiceNow. Instead of building conversation flows and training models, you're now working with Large Language Models that genuinely understand natural language. Instead of mapping every possible variation of a question, you describe what you want the AI to do, and it figures out how to do it.

This is generative AI embedded throughout the entire ServiceNow platform. Not just chatbots. Everything.

Think about what that means. Now Assist can read through 300 work notes on an incident and give you a proper summary of what actually happened. Not just field values. Not just the last comment. A genuine understanding of the entire situation.

It can write knowledge articles that sound human. Generate resolution notes that make sense. Answer employee questions by pulling information from your knowledge base and understanding context. Help agents respond to customers with suggested replies that actually fit the conversation.

And Virtual Agent? It's still there, but now it's powered by these same LLMs. So instead of spending weeks building intent models, you describe what your topics should do in plain English, and the AI handles the conversation naturally. Users can ask questions however they want, and it understands.

What Now Assist Actually Is

At its core, Now Assist is a family of AI capabilities built on Large Language Models and integrated throughout the ServiceNow platform. It's delivered through something called skills.

A skill is basically a specific AI capability paired with a use case. Incident summarisation is a skill. Knowledge article generation is a skill. Resolution note creation is a skill. Virtual Agent conversations are skills. There are dozens of them, covering ITSM, ITOM, HR, Customer Service, and more.

These skills are supported by the Now Assist framework, which sits on top of the ServiceNow AI Platform. You manage everything through the Now Assist Admin Console, where you can activate skills, monitor usage, and track performance.

There's also the Now Assist Skill Kit, which lets you build custom skills for your specific needs. Because whilst the out-of-the-box skills cover common use cases, every organisation has unique requirements. The Skill Kit gives you the flexibility to create AI capabilities tailored to your exact needs.

And critically, there's Now Assist Guardian, which monitors and evaluates everything the AI generates. Because in an enterprise setting, you can't just let AI run wild. Guardian helps ensure outputs are appropriate, accurate, and safe.

The Shift You Need to Understand

Here's what's fundamentally different. With Virtual Agent and NLU, you were essentially teaching the system by example. Show it 50 ways people might ask for a password reset, and it learns the pattern. It was training-based.

With Now Assist and LLMs, you're instructing the system. You describe what you want, provide context, give it rules, and the model uses its understanding of language to do the work. It's prompt-based.

That's why this entire book exists. Because the quality of your AI outputs now depends entirely on the quality of your prompts. Get your prompts right, and you'll have AI that genuinely helps your teams and users. Get them wrong, and you'll have inconsistent outputs, frustrated users, and an implementation that never quite delivers on its promise.

The technology is ready. ServiceNow has done the hard work of integrating these powerful LLMs throughout the platform. The models are trained, the framework is built, the infrastructure is there.

What you need now is the skill to use it properly. That means understanding how to write prompts that get the results you want. How to provide the right context. How to structure instructions. How to test and refine until you've got something that works reliably at scale.

Why This Is Worth Your Time

You might be thinking this sounds like a lot of work. Learning a whole new skillset, rethinking how you approach problems, and changing the way your organisation uses ServiceNow.

You're right. It is work.

But consider the alternative. Your users have experienced ChatGPT. They know what good AI looks like. If you stick with old approaches and don't invest in understanding how to make Now Assist work properly, you'll end up with AI that disappoints rather than delights.

On the other hand, get this right, and you're looking at genuine transformation. Agents who can handle more tickets because they're not drowning in admin work. Users who actually get helpful answers from self-service. Knowledge bases that stay current because creating articles isn't a massive burden. Incidents that get resolved faster because the correct information is instantly available.

The business case practically writes itself when you implement this well.

So yes, you should care about Now Assist. Not because it's shiny new technology, but because it's a fundamental shift in how ServiceNow works, and the organisations that figure this out first will have a massive advantage over those still trying to make old approaches work in a new world.

Right, let's look at what that business value actually looks like.