As part of my ongoing initiative to go beyond the basics and develop a deeper, practical understanding of AI, I recently completed the “Intro to ChatGPT and Generative AI” course on the ServiceNow Now Learning platform. I highly recommend it as a starting point—whether you’re a ServiceNow architect, administrator, or simply someone looking to leverage AI more effectively in your daily work.

Even though I’ve been a heavy user of ChatGPT for over a year (thanks to a Plus subscription and plenty of hands-on experimentation), this course offered a fresh perspective and several new ideas I hadn’t explored before. Here are the highlights and insights I found most interesting and actionable from the training:

Fresh Tips for ServiceNow Architects and Technologists

Most ServiceNow architects and technical leads already use ChatGPT for everything from requirements workshops to code reviews. Yet, after completing ServiceNow’s “Intro to ChatGPT and Generative AI” course on Now Learning, I was surprised to find new actionable tips and overlooked some mechanics that do benefit seasoned professionals. Below, I’ve distilled the key insights, with some examples relevant to ServiceNow.

Prompt Engineering for Architects: The Real Differentiator

ChatGPT isn’t magic; its value is directly tied to the specificity and context of your prompts.

It always reminds me of the book “The Hitchhiker’s Guide to the Galaxy”. In the story, a supercomputer named Deep Thought spends 7.5 million years calculating the “Answer to the Ultimate Question of Life, the Universe, and Everything” and concludes that the answer is simply “42.” The joke is that, despite the computer’s intelligence and effort, the answer is meaningless because the original question was never clearly defined—highlighting that even the smartest systems give useless answers if you don’t ask the right question.

This goes double for architects who want more than generic answers. Here’s how to make your prompts work harder:

Effective Prompt Structure for ServiceNow Use Cases

WHO: “As a ServiceNow architect working on [module/project]…”

WHAT: “I need to design [process/automation/script/test case]…”

WHERE: “Context: [industry, region, regulatory framework]…”

WHY: “Because [business driver, compliance reason, customer need]…”

WHICH/HOW: “Please output as [bullets, table, JSON, script, workflow diagram]…”

Example:

As a ServiceNow architect working on Change Management, I need to design a change approval process that ensures PCI DSS requirements are embedded,
Context: financial services company operating in the US,
Because all changes impacting payment card systems must meet PCI compliance and be auditable, Please outline the process as a step-by-step workflow diagram (described in text), highlighting roles, approval stages, and documentation requirements.

It’s not so far to writing a Development Story. It’s unfortunately natural that we tend to jump on the solution and forget to state clearly the problem to Solve, probably because that’s the hard part and it doesn’t look like progress.

Iterative Prompting and Context Stacking

ServiceNow professionals rarely get it right on the first try—AI interactions are no different. Use iterative prompting to narrow solutions or refine requirements:

• Initial Prompt:
  Design a ServiceNow Change Management approval workflow for PCI compliance in a US financial services company. Provide as step-by-step text.
• Follow-up:
  Clearly identify which persona (e.g., PCI Compliance Officer) is responsible for each approval step.
• Follow-up:
  Explain how to handle rejection by some approvers or ignored approvals.

Use ChatGPT’s memory in a single conversation thread, but always summarize outcomes at the end if you’ll continue later—context does not persist across sessions. I learnt it at my expense while developping a business case.

Power Commands for ServiceNow-Related Tasks

I wasn’t really aware there was commands in ChatGPT, but the course highlighted several that are incredibly useful.

Command	What it Does	ServiceNow Example
CONTINUE	Extends incomplete responses	Long technical designs, scripts, policy lists
COMPARE/CONTRAST	Side-by-side solutioning	Compare Flow Designer vs. Workflow Editor for automation
PROS & CONS	Decision framing	“Pros & cons of Virtual Agent vs. Knowledge Article for L1 support”
DEVIL’S ADVOCATE	Forces critique	“Argue against the idea to use a selft hosted environement”
ROLEPLAY	Simulates interviews or customer objections	“Roleplay as a skeptical CISO challenging my security model”
REPHRASE/SUMMARIZE	Refines technical content for different audiences	Convert my workshop notes in a set of stories. Convert the Sales proposal in a statement of work.
TROUBLESHOOT	Diagnoses code or integration issues	“Why does my REST call return 403 in ServiceNow Scripted REST?”

To be fair, you don’t have to specifically write the command “CONTINUE” you can ask this command in natural language “Is that all???” and it will do the same.

Document and Image Handling for Solution Architects

• Document upload: Instantly summarize process documentation, SOWs, or policy PDFs. Saves hours in RFP or compliance reviews.
• File-based Q&A: “From the attached ServiceNow update set, identify risky business rules.”
• Image generation: Rapidly prototype UI mockups or architectural diagrams. For example, “Generate a workflow diagram for an ITSM major incident process.”

I really love the File-base Q&A. Imagine you have a very extensive process documentation as PDF, but you have a very specific detailed question. You can upload. The file and ask, “according to the process, what would be the applicable urgency for an incident that’s blocking a user on a non critical service?”

Custom GPTs: From Point Solutions to Strategic Tools

Most ServiceNow orgs haven’t tapped the potential of custom GPTs. Some strategic uses:

• Story Review Assistant: Validate user stories for completeness, acceptance criteria, and technical details (paste XML or description).
• Code Quality GPT: Scan for ServiceNow best practice violations, e.g., GlideRecord misuse, synchronous integrations (after loading your style guide)
• Architecture Advisor: Summarize platform implications (licensing, data residency, scalability) for solution designs.

Tip: Regularly upload updated architectural standards or documentation to keep custom GPTs current.

Using custom GPT is for me the greatest revelation in that course. It feels like you really have a partner working on your side that cares as much as your do. It’s very easy to set up. It has all knowledge of ChatGPT + anything you teach from documents. It will use the tone, and the level of details that you specify.

Prompt Engineering Table: Avoiding Common Pitfalls

Prompting Pitfall	Result	Solution
Vague: “How to integrate SAP?”	Overly generic, non-actionable	Specify: “Integrate SAP S/4HANA via MID Server for outbound IDoc, stepwise with example payload.”
One-shot prompt	Misses important context	Build context iteratively, reference uploaded files when possible
Asking for “best practice” only	Gets outdated/general advice	Specify ServiceNow version, instance size, regulatory region, and use case.

Ethics, Hallucination, and Quality Assurance

Even with ServiceNow code, ChatGPT can hallucinate—producing non-optimal scripts. Takeaways:

• Never deploy AI-generated code without review. I’ve seen GlideRecord loops with performance issues and REST examples missing error handling.
• AI excels as a brainstorming partner, not a final authority. Use it to draft, not decide.

But it will have sometihng right most of the time. It provides a great baseline for non-developers, and even human developpers are sometimes below optimal…. after days on a script.

Maximizing Collaboration and Continuity

For multi-day design sessions, architecture reviews, or iterative solutioning:

• Rename each ChatGPT conversation (e.g., “CSM Integration Architecture 2024-04-27”) for tracking.
• Summarize and download key outputs at day’s end. If continuing later, re-upload context or summary file.
• Cautiously use conversation sharing—ensure no sensitive client data is exposed in public share links.

Conclusion

GPT is fundamentally reshaping the daily reality for ServiceNow architects and developers. It’s shifting the focus from creating solution to stating properly the problem and validating + refining the solution.

Rather than replacing our expertise, these tools amplify and accelerate it.

The future of work belongs to those who can combine deep platform knowledge with the ability to leverage AI creatively and responsibly.