Use of AI

How we can leverage the tools responsibly

Of Course We Use AI

As an effort which is trying to outline standards for genetic AI and ITSM – managed organizations, one would expect that we would allow unlimited AI use. That cannot be further from the truth.

Using AI Responsibly

Responsible AI use

AI tools should be used thoughtfully and deliberately, to enhance human capability rather than obscure responsibility or bypass critical thinking.

Human accountability

The person using AI remains solely responsible for all work submitted to the project, regardless of AI involvement. You are the steward of quality, accuracy, and alignment with project goals.

AI as force multiplier

AI excels at automating straightforward, repetitive work, freeing people to focus on complex problem-solving, creative thinking, and strategic decisions. It is not a replacement for human expertise.

Experience matters

Humans bring context that AI can’t: years of learned experience, tacit knowledge, and an intuitive feel for nuance. Those strengths are foundational to high-quality work.

Permitted Uses

You’re encouraged to use AI to:

Investigate new ideas: brainstorm, explore possibilities, and stress-test concepts before committing time and resources.
Draft thoughts and approaches: work out preliminary thinking, organize arguments, and refine proposals.
Generate initial code, text, and documentation: produce boilerplate, scaffolding, and drafts of docs, tests, and implementations. Just remember that what it writes, you must read.
Accelerate routine tasks: speed up formatting, refactoring, testing, and other standardized work.
Submit pull requests: you can absolutely submit PRs that include AI-assisted contributions. Disclose AI involvement where it’s useful, for example in commit messages or PR descriptions.

Your Responsibilities

When you use AI in your work:

Verify and validate: review AI output critically, tesonfirm it meets project standards and requirements.
Exercise judgment: use your domain expertise to catch errors, inconsistencies, or mismatches AI may miss.
Maintain context: make sure AI-generated work fits thcture, and design patterns.
Own the outcome: you’re responsible for everything you submit (including AI-assisted work) exactly as if you’d written it yourself.
Disclose when appropriate: mention AI involvement in PRs or discussions when it adds useful context, though it isn’t required for routine AI assistance.

Your AI tools should not be…

A replacement for expertise: the tools can’t replicate the depth of understanding that comes from experience.
A shortcut to accountability: using it doesn’t reduce your responsibility for the quality of what you submit.
Sufficient for high-stakes decisions: human review and approval remain essential for critical architecture, security, and strategy calls.
A substitute for learning: solving problems with AI without understanding them hurts your own growth and the team’s collective knowledge.
An excuse for corner-cutting: AI should enhance your work, not let you skip rigor.

Case Study: Why Human Expertise Remains Essential

In my grad school days, an account in Shoshana Zuboff’s In the Age of the Smart Machine illustrated why human context can’t be automated away. It has stayed with me all these years as the tools tranformed, some say transmogrified, adding more and more capability. However, given the evidence we see in the market from early layoffs and subsequent rehirings of skilled professionals, the lesson learned remains to this day.

CAM Automation of Sheet Metal Cutting

Situation

In early sheet metal nesting and Computer-Aided Manufacturing (CAM) software, engineers built algorithms to optimize how 2D shapes are laid out on metal sheets to minimize scrap. On paper, the algal yield. But when those cuts went live on the factoryfloor, parts warped or jammed the machines.

The problem: software engineers treated metal as pure flat geometry. They missed physical realities that master sheet-metal workers understood intuitively:

Thermal expansion: high-heat plasma and laser cutters don’t allow heat to dissipate.
Grain direction and springback: bending sheet metal across its grain causes cracking, which requires specific part orientations.
Micro-tabbing and skeleton strength: leaving small tabping into the slag bed and wrecking $500,000 laser heads.

Solution

The fix wasn’t better algorithms. It was empowering the experts. When tooling vendors moved from rigid automated software to intuitive programming toolkits that let experienced fabricators steer the software, the outcomes changed: material utilization went up, scrap fell by double digits, and machine downtime dropped.

Lesson

Tools work best when they amplify human expertise rather than replacing people. It was by empowering skilled operators, not sidelining them, that the tools produced the most signfiicant investment results.

Best Practices

Start with clarity: before using AI, know what you’re trying to accomplish and what success looks like.
Review with fresh eyes: ask whether the output matches what you need, whether you’ve missed edge cases, and what could go wrong.
Layer in expertise: use your experience to refine, redirect, or reject AI suggestions.
Document reasoning: when AI output shapes a decision, note why you trusted it or why you changed it.
Keep learning: don’t outsource understanding. If you don’t see why AI suggested something, dig deeper before using it.
Collaborate openly: share how you’re using AI in code reviews and discussions, and help the team learn what works.

Conclusion

AI is a powerful tool for the Open Agentic ITSM project, faster on routine work and explore ideas more broadly. It only reaches its potential when paired with human judgment, accountability, and expertise.

Use AI boldly. Verify thoroughly. Own the result. This policy reflects the principle that technology serves human capability, not the reverse.