From skepticism to strategy: Rethinking generative AI through practical application

Author's Note:

Atlassian’s Team ‘25 was held last week, and there was significant talk of AI. If you are an AI skeptic like myself, or still exploring AI’s capabilities, this article is definitely for you.

This post represents a working perspective from an AI skeptic operating in the IT Service Management and Atlassian solution space. It captures a grounded, structured approach to using generative AI meaningfully in enterprise environments.

The Origin of Skepticism

I describe myself as an AI skeptic. Not in the contrarian sense, but as a practitioner who does not buy into the sweeping claims surrounding artificial intelligence—particularly those made about generative AI models and their supposed transformative potential in enterprise IT.

In my world of IT Service Management (ITSM), precision, structure, and verifiability are non-negotiable. Promises of AI-driven productivity, intelligent automation, or creative augmentation are often vague and ungrounded when examined through the lens of enterprise service delivery.

A Use Case That Changed My Perspective

That said, something shifted recently. While evaluating automation rules in Atlassian's Jira Service Management (JSM) platform, I engaged a generative AI assistant to help analyze and compare rule configurations.

What happened surprised me.

The assistant’s performance wasn’t just adequate—it was genuinely helpful. Not because it was autonomously intelligent or inherently insightful, but because the way I framed the task—clearly, contextually, and with precise expectations—created the conditions for value to emerge.

Key Realization: 

Generative AI isn’t a solution. It’s a responsive system that becomes useful when properly directed.

This realization didn't invalidate my skepticism; it refined it. The current generation of generative AI performs best under governance, not autonomy. It's a responsive system that can be made valuable through structure, strategy, and stewardship. The AI hype cycle suggests these tools are silver bullets. My experience showed they are malleable tools—performing best when implemented under robust governance frameworks.

The Hidden Cost of Value

The hidden cost of working with generative AI is often overlooked: it demands work to produce reliable, actionable value. To get dependable results, I had to invest in prompt engineering, define operational context, constrain output expectations, and manually validate these results. This isn't "plug-and-play AI." This is structured collaboration with a system that requires domain expertise and intentional input design. 

In many ways, this mirrors the principles of knowledge-centered service (KCS), change enablement, and structured automation. Success is not rooted in AI automation alone, but in the intentional design of how knowledge flows, how context is managed, and how systems are governed by clearly defined rules. The AI, in this sense, was a force multiplier—but only because I served as the architect of its scope and boundaries.

A More Realistic Role for AI

This leads to a more grounded vision for AI’s role in ITSM, enterprise automation, and digital transformation: a Structured Assistant in ITSM and Automation Strategy. These are the core principles I now consider when leveraging an AI assistant.

Reframing AI in this way enables meaningful integration into ITSM practices and automation workflows. Practical applications include:

• Validating logic in Jira Service Management automation rules

• Drafting knowledge base articles aligned with KCS methodology

• Summarizing root cause analyses and post-incident reviews

• Translating operational metrics into stakeholder-ready narratives

But these uses must always occur under governance. Generative AI should never operate without human oversight or review.

Reframing the AI Conversation

I remain skeptical of AI—but now with greater nuance. I reject the marketing hype, but I acknowledge the real-world utility of generative AI when implemented in a structured, intentional manner.

The performance of AI in my automation analysis workflow didn’t showcase artificial general intelligence. It showcased the value of disciplined design, structured prompting, and contextual clarity.

The real question isn’t whether AI is intelligent. It’s whether we are building the scaffolding it needs to be useful.

How are you navigating AI?

If you're navigating AI integration in ITSM, I'd love to hear your perspective-how are you scaffolding your AI strategy? Please tell us all about it!