From Development to Production: Ask ServiceNow's AI Experts Anything

[ryanyong]

Hey everyone, I'm Ryan, an engineer on ServiceNow's Forward Deployed Engineering team. We're a team of experts and builders who've gained deep AI deployment expertise at tier 1 technology companies and research labs. We're certified experts on the ServiceNow AI Platform, with direct access to Product and the ServiceNow C-suite.

What we do
We partner directly with customers facing the most complex challenges to co-innovate agentic AI use cases and ship them to production on the ServiceNow AI Platform.

Why this thread
If you're a ServiceNow customer sitting on AI entitlement you haven't fully put to work, this is your chance to ask the Forward Deployed Engineers building these solutions in the field. Bring your hardest questions about moving from unused credits to AI running in production, and getting the most value out of your existing ServiceNow investment.

Ask us anything. We are the subject matter experts, and we will get you an answer.

We'll be answering live from August 3rd to 7th. Post your question anytime between now and 8/7, upvote the ones you want answered first.

[johlee]

On behalf of a few customers - 

What are the best practices for exposing NAVA on an external website as a customer-facing self-service chatbot? How should we think about structuring our knowledge and permissions? 

Use case: we have NAVA set up to answer questions for internal employees using our internal knowledge base. We want to expose NAVA via. API to our external help website so that our customers can get their questions automatically. 

[Dustin Gannon]

I would break this project into a few parts.

The first thing I hear is that you already have Now Assist Virtual Agent working internally.

What I don't know is the sensitivity and sanctity of the knowledge currently available to that experience.

So before I expose anything externally, I would get the dataset right.

I would consider using AI agents to parse and evaluate the entire knowledge base and potentially score, classify, or refactor that information into another knowledge base that you are certain is appropriate for customer-facing or public consumption.

Step one is making absolutely sure that the dataset you expose externally doesn't contain something that's going to put you in a security incident—or a news headline.

Once you've done that, the rest of the problem becomes much more normally shaped.

Configure the appropriate AI Search profile, configure Virtual Agent, establish the correct permissions, and expose the experience through whichever external portal or interface makes sense.

I might actually create three logical knowledge sets:

• One for internal-only knowledge.
• One for public/customer-facing knowledge.
• And one for shared knowledge that should be available in both internal and external experiences.

Then include that shared knowledge base in both search profiles.

That way, when your knowledge managers create information that genuinely applies to everyone, they can write it once and essentially make it available to both chatbot experiences without duplicating the article.

[Maulik176]

I'm a ServiceNow developer focused on custom app development and PS implementations (client scripts, business rules, Flow Designer, UI Builder, Service Portal) rather than core ITSM/CMDB. As orgs move from unused AI entitlements to production agentic AI, where does a custom-app developer plug into that journey — are you seeing FDE engagements pull in developers like me to build agent tools/actions on top of existing custom apps, or is that work handled separately from custom app dev? And from a skills standpoint, what's the fastest path to being useful on an agentic AI deployment?

[Dustin Gannon]

There are several questions in here, so I'll take them separately.

First: absolutely, we still need separate scopes and separate apps.

If anything, there's never been a greater need for thoughtful segmentation because the volume of applications and automation we can build is becoming so much larger.

So the skills you already have in building applications are still incredibly valuable.

The first design question is whether you're configuring inside somebody else's scoped application to meet their definition of done, or whether you're building something genuinely net new that performs a net-new function.

If it's genuinely net new, I would strongly consider designing it as a separate application.

Second, you're already tremendously valuable to an agentic AI team because you know how to configure ServiceNow.

If you know how to create a scoped app, extend the Task table, build flows, add workspace functionality, create business rules, and understand how the platform works, bring that expertise.

The AI agent team you're working with may not be experts in those areas.

You also probably know tremendous amounts about the data model of the applications you're working with - the transactions, relationships, functions, edge cases, and intricacies of how that application actually behaves.

A brand-new forward deployed engineer entering your particular use case will not automatically know all of that.

Bring that expertise too.

Finally, if you're asking how to go from zero to useful with AI agents, I actually teach this as a very straightforward progression:

1. Your first app.
2. Your first skill.
3. Your first agent.
4. Your first eval.
5. Your first deployment.

Do those things in that order.

Once you've gone through that lifecycle once, you've learned the basic building blocks.

Every project after that uses essentially the same building blocks - it just does them at a bigger scale.

[Sanjay Bagri1]

Hello @ryanyong,

Please let us know why customers are not using now assist and agentic AI in real time.

Folks are continuously learning these skills but haven't got chance to implement in real world.

I think ServiceNow is not promoting the real time use cases of the above topic. there is a lack of confidence to utilities Now assist and AI agentic.

Let us know how we can utilities our knowledge of above technology.

Thank you for this open forum.