AI tools & productivity: the rules I follow (so “helpful” doesn’t turn into chaos)

 # AI tools & productivity: the rules I follow (so “helpful” doesn’t turn into chaos)

**Last updated:** 10/02/2026

AI tools are everywhere now. That’s not automatically good news.

Used well, they save time and reduce mental load. Used badly, they create noise, false confidence, and a messy workflow where you’re constantly switching tools instead of finishing work.

This post is not a “top 50 AI tools” list. It’s the skeptical approach I personally follow—rules that keep AI useful, predictable, and non-intrusive.

---

## 1) First principle: AI is not a source of truth

AI is a **pattern machine**, not an authority.

Rule:

- If it’s important (money, legal, health, security, contracts, client decisions), I verify it.

- If it’s low stakes (drafting, brainstorming, formatting), I let AI move fast.

AI should accelerate your thinking—not replace your judgment.

---

## 2) Use fewer tools, not more

Most people don’t need 6 AI apps. They need **one primary tool** and one backup.

Rule:

- One main assistant for writing, planning, and problem solving.

- Optional: one “search-first” tool for quick web research.

- Everything else must justify its existence.

If a tool doesn’t save time every week, it gets removed.

---

## 3) Define the “job” before you ask

Bad prompt = vague problem.

Good prompt = clear output.

Rule:

Before I ask anything, I define:

- **what I want** (deliverable)

- **constraints** (tone, length, audience, format)

- **what I already know**

- **what I don’t want**

Example (good):

“Write a 300–400 word blog intro for Ubuntu users, skeptical tone, no hype, include 3 bullet points, end with a clear next step.”

---

## 4) AI should produce “drafts”, not “final decisions”

If you publish AI output unchanged, you will eventually embarrass yourself.

Rule:

- AI gives me a first draft.

- I tighten it, fact-check key claims, and add personal context.

Your brand is built on *your* voice and credibility.

---

## 5) Don’t let AI touch your accounts by default

Agent tools that connect to email, calendars, and cloud drives sound attractive.

They also expand your risk surface instantly.

Rule:

- AI can draft emails and plans.

- I send, approve, and execute important actions manually.

- If I ever automate actions, I start read-only and keep strict permissions.

Convenience is not worth losing control.

---

## 6) The “two-layer workflow” that keeps things clean

This is my simple system:

### Layer A: Capture (fast)

- dump raw ideas, notes, and drafts quickly

- no perfection

### Layer B: Publish/Execute (clean)

- only refined, checked, structured items move here

- one source of truth

AI belongs mostly in Layer A.

Layer B is where you protect quality.

---

## 7) Analytics without obsession

Productivity isn’t about tracking everything.

It’s about removing friction.

Rule:

- I track only what changes my decisions:

  - what content performs

  - what tasks create income

  - what actions reduce wasted time

If a metric doesn’t change behavior, it’s just stress.

---

## 8) What I’ll cover next (practical series)

In the next posts I’ll publish:

- my “minimal AI stack” (what stays, what goes)

- practical prompts for tech writing and troubleshooting

- how I use AI to plan content without burnout

- how to do research without trusting random outputs blindly

---

## Closing thought

AI is useful when it’s boring.

Predictable. Controlled. Repeatable.

The moment it becomes a shiny distraction, it stops being a tool and becomes a problem.

If you want this blog to stay practical, that’s the line we’ll keep.

— Dimitris