What I Automate vs What I Never Automate

Everyone wants a list of “AI use cases.” But lists without principles are dangerous — you’ll automate the wrong things and wonder why your work feels hollow.

So instead of just giving you my list, I’ll give you the list and the principle behind it. The principle matters more, because your situation is different from mine. Learn the rule, then draw your own line.

The short answer: automate the repeatable, never automate the relationship. Machines handle scheduling, formatting, and first drafts; I personally handle judgement, teaching, replies, and anything that carries my name’s trust. The principle matters more than any tool list.

The Principle

Here’s the entire framework in one sentence:

Automate the mechanical. Protect the meaningful.

Mechanical = tasks where the process matters more than who does it. Formatting, transcribing, summarizing, proofreading, generating variations. If a competent stranger could do it as well as you, it’s mechanical.

Meaningful = tasks where you are the product. Your judgment, your stories, your relationships, your reputation. If only you can do it — because it requires your experience, your values, or your accountability — it’s meaningful.

Simple. But applying it honestly requires more discipline than you’d think, because the temptation is always to automate the meaningful parts too. They’re often the hardest parts, and hard is exactly what we want to escape.

What I Happily Automate

Research synthesis. “Summarize these five articles.” “What are the main arguments for and against X?” “Find the gaps in this topic.” AI does in minutes what took me hours — and I verify the important claims myself. The synthesis is mechanical; the verification is meaningful.

First drafts from my raw material. I dump messy thoughts — voice notes, bullet points, half-sentences in mixed Bangla-English — and AI shapes them into a readable first draft. The thinking is mine; the structuring is delegated. Then I rewrite heavily. (See How AI Changed My Content Workflow for the full process.)

Repurposing. One article into a script outline, social posts, email draft, FAQ set. This is pure format-shifting — the ideas already exist, they just need new containers. Highly mechanical, enormously time-consuming for humans, trivial for AI.

Mechanical editing. Grammar, spelling, repetition detection, consistency checks. AI is a tireless proofreader. It catches things my tired eyes miss at midnight. I still do the voice edit myself — that’s meaningful — but the mechanical layer is fully delegated.

Idea expansion. “Give me ten angles on this topic.” “What questions would a beginner ask about this?” AI is a decent brainstorming partner — not because its ideas are brilliant (they’re usually mediocre), but because reacting to mediocre ideas helps me find my good ones faster. The selection is mine; the generation is shared.

Scheduling and logistics. Reminders, calendar management, email drafts for routine matters, organizing files. The unglamorous infrastructure of a working life. Automate all of it without guilt.

What I Never Automate

My opinions. When I take a stance — on a strategy, a tool, a trend — that’s mine. AI can help me articulate it, but it can never form it. An opinion I didn’t arrive at myself isn’t an opinion; it’s a costume. And audiences can smell the difference.

My stories. My experiences at Grameenphone, Pathao, FrontStory. My failures, my specific observations, the moments that shaped me. These are the most valuable content I own, and they’re non-delegatable by definition. No AI can have my memories.

Final publish decisions. “Is this good enough? Is this true? Should this go out under my name?” These questions have no mechanical answer. They require judgment, taste, and the willingness to be accountable. I am the bottleneck here by design.

Replies to real people. When someone comments, messages, or emails me — a human answers. I wrote about this commitment in Why I Reply to Every Comment. Could AI draft replies? Sure. Would it be faster? Absolutely. Would it destroy the trust I’ve spent years building? Instantly. Some things are inefficient on purpose — the inefficiency is the signal that you care.

Relationships. Introductions, collaborations, mentorship, community building. AI can help me manage relationships (reminders, notes, follow-ups) but never have them. The moment a relationship becomes automated, it stops being a relationship.

Anything where I’m accountable. This is the ultimate test. If something goes wrong and someone asks “who decided this?” — the answer must be me, a person, not “the AI suggested it.” Accountability can’t be delegated. If I can’t stand behind it personally, it doesn’t ship.

The Gray Zone (Where Honest People Disagree)

Some tasks sit on the line, and reasonable people draw it differently:

Outlining. I co-create outlines with AI but always finalize them myself. Some creators let AI outline fully. I think structure is a judgment call — it determines what the piece says — so I keep my hands on it.

Headlines and titles. AI generates options; I choose. The choice is the meaningful part. But I’ve seen creators who write their own headlines from scratch and consider AI suggestions “cheating.” That’s a stricter line than mine, and I respect it.

Data analysis. AI can crunch numbers and spot patterns. But interpreting what the patterns mean for strategy? That’s judgment. I let AI do the math; I do the meaning.

The gray zone is where your personal line lives. Draw it consciously. Revisit it as tools improve. But always be able to explain why you drew it where you did.

How to Draw Your Own Line

Ask three questions about any task:

Does this require my experience? If yes → protect it.
Does this require my accountability? If yes → protect it.
Would the output be meaningfully worse without me? If yes → protect it.

If the answer to all three is no, automate it freely and without guilt. That’s not laziness — that’s leverage. The goal was never to do everything yourself. The goal was to spend yourself on what only you can do.

Automate the mechanical. Protect the meaningful. Everything else is commentary.

*I’m building my version of success in public — and helping Bangladeshi job holders build theirs. Join the journey in Bangla: @superpreneurbd*