Technology
What I Stopped Delegating to AI (After Delegating Everything)
I handed too much work to AI and watched my voice flatten. Here is what I took back, and what I still let the tools handle.

I gave AI too much.
Not in a dramatic “the future is cursed” way.
In a boring, modern, very believable way.
Client emails went to AI.
First drafts went to AI.
Research summaries went to AI.
Planning went to AI.
At first, it felt like I had found hidden time.
Then my work started getting strangely flat.
My writing sounded clean but less alive. My emails sounded like a corporate helpdesk that had read one feelings book. My planning got broader, but my own instincts got weaker.
Worst part?
I felt busier, not freer.
I Took Back First Drafts
First drafts look like writing.
They are actually thinking.
That was the first thing I had to admit.
When AI wrote the first version, I was not starting from my own messy idea. I was editing toward something.
That sounds efficient.
It also changes the work.
The blank page is not wasted time. It is where your actual position shows up.
So I took back first drafts for anything that needs my voice:
- Blog posts
- Point-of-view essays
- Sales pages
- Strategy notes
- Difficult emails
AI can help after the ugly version exists.
But it does not get to write the first sentence when the sentence is the work.
I Took Back Relationship Messages
AI can write a polite message.
That does not mean it should write every message.
The moment emotion enters the room, I write it myself.
Apologies.
Pushback.
Scope conversations.
Pricing tension.
Bad news.
Anything where the other person needs to feel that I am there, not that I found a nice template.
This is also why when not to use AI became a hard rule for me instead of a nice idea. Relationship work is not admin. It is the work.
I Took Back Strategy
This one hurt.
I was using AI to avoid sitting with decisions.
Should I reposition the offer?
Should I say no to this client?
Should this article exist?
Should I change the plan?
Instead of wrestling with the decision, I would ask the model for options.
The options were useful.
The habit was not.
At some point, I noticed I was collecting choices instead of building judgment.
That is not strategy. That is hiding inside a menu.
The move: make your own rough call first.
Then ask AI to attack it.
Not replace it. Attack it.
There is a huge difference.
What I Still Delegate
I still use AI.
A lot.
I use it for:
- Turning messy notes into a checklist
- Finding holes in an outline
- Making a draft clearer after I know what I mean
- Creating alternate headlines
- Summarizing material I will still verify
- Stress-testing an argument
That is scaffolding.
Useful.
Fast.
Not the soul of the work.
I wrote a cleaner breakdown in how I actually use AI in my daily workflow, but the whole rule is simple:
Use AI after you have a position.
Not instead of having one.
The Test I Use Now
Before I delegate something to AI, I ask:
If the output is good, will I still understand why?
If no, I do the work myself.
If the client reads this, would they care that I wrote it?
If yes, I write it myself.
If AI makes this smoother, will it also make it less mine?
If yes, I write the ugly version first.
That is the boundary.
Not anti-AI.
Anti-disappearing from your own work.
If you are using AI because you are stuck, read how to build momentum when everything feels stuck before you prompt for the thirteenth time. Sometimes the next step is not another answer.
Sometimes it is deciding.
Advertisement