Why LLMs Became Part of My Toolbox

Why I Don't See Myself Working Without LLMs Anymore

Like many people, I was skeptical when Large Language Models started becoming mainstream. There was a lot of hype, plenty of exaggerated claims, and just as many people predicting they would replace entire professions overnight.

After using them every day for quite a while, my opinion has changed completely.

Not because they always give the right answer. They don't.

The real shift happened when I stopped expecting answers and started using them as a thinking partner.

Today, an LLM sits alongside PowerShell, Excel, Visual Studio Code, and my notebook. It's simply another tool. One that helps me organize ideas, challenge assumptions, review my work, and avoid spending hours staring at a blank page.

The final decision is still mine. That's the part that matters.

It Doesn't Replace My Work. It Changes How I Start It.

Very little of what I produce comes directly from an LLM.

Instead, I use it to create momentum.

I'll ask it to review a document I've written, point out inconsistencies, suggest a different architecture, summarize documentation, or play devil's advocate when I'm not fully convinced by my own design.

It's less like outsourcing work and more like having a colleague who's always available for a second opinion.

That changes the way I work.

Knowledge Makes LLMs Better

People often ask whether AI will make expertise less important.

I've found the opposite.

The more you know about a subject, the more useful an LLM becomes. You immediately spot weak reasoning, outdated recommendations, missing context, or ideas that simply don't fit your environment.

Without that critical mindset, it's easy to confuse confidence with correctness.

I sometimes describe LLMs as "pre-chewed information." They do the first pass remarkably well. They save time. They don't replace judgment.

Where It Saves Me Time

The biggest impact hasn't been writing code. It's everything around it.

On a typical week, I use an LLM to:

• Challenge migration strategies for Microsoft Intune and Microsoft 365 projects before proposing them to customers.
• Estimate the effort required for workshops, assessments, and consulting engagements.
• Review and improve PowerShell scripts that I've already written.
• Rewrite technical documentation and blog articles so they're clearer without losing the original intent.
• Brainstorm governance decisions or alternative operating models.
• Design and troubleshoot Home Assistant automations, including analyzing temperature trends and other environmental data collected in my apartment.
• Build internal tools, prototypes, and small utilities much faster than I could starting from scratch.

None of these tasks are impossible without AI.

They're simply much faster.

If I can validate an idea in five minutes instead of spending several evenings building a prototype, I've saved days or even weeks of work. If I can generate a first draft of documentation instead of starting with an empty page, I can spend my time improving it instead of creating it.

That's where the real productivity gain comes from.

It's Not for Everyone

I don't think LLMs benefit everyone equally.

They're incredibly effective if you're naturally curious, willing to verify information, and comfortable questioning what you're given.

They're much less useful if every answer is accepted without thinking, or if there's no interest in understanding the subject behind the response.

Like any tool, its value depends on the person using it, and lot of people are confirming my opnion everyday with what they us AI for.

Automation Still Needs Humans

There is one area where I remain cautious: automation.

An LLM can write surprisingly good PowerShell, Python, or Bash scripts in seconds.

That doesn't mean they're ready for production.

Business context is often missing. Edge cases are forgotten. Error handling isn't always sufficient. And unattended automation is usually far less forgiving than a chat window.

I'm happy to let an LLM write the first version of a script.

I'm not prepared to let it manage production systems without thorough review and testing.

That balance hasn't changed.

The Technology Isn't Going Away

Every major technology shift has gone through the same cycle: excitement, disappointment, then adoption.

I think LLMs have already passed that point.

Whether people like them or not, they're becoming part of everyday work. Not because they're perfect, but because they're capable of removing an enormous amount of repetitive cognitive work.

For me, the biggest benefit isn't that an LLM thinks instead of me.

It's that it lets me spend less time producing first drafts and more time making decisions.

After working this way for a while, going back would feel like voluntarily giving up one of the most useful tools I've added to my toolbox in years.