Reading Time: 6 minutes
AI Fatigue: When Productivity Makes It Hard to Stop
I have a backlog.
Actually, calling it a backlog might be generous. Like most business leaders, I have accumulated years of ideas, improvements, processes, and projects that I know could make our business better.
The problem has never been a shortage of ideas.
It has been the capacity to execute them.
Every idea eventually runs into the realities of business. People have competing priorities. Development takes time. Requirements need to be communicated. Meetings need to happen. Work needs to be reviewed. Something urgent inevitably gets in the way.
That is simply how organizations work.
Over the past several months, however, I started experimenting with something that has fundamentally changed my perception of that constraint.
I went hands-on with AI.
I Wanted to Test It Myself
My background is in technology and architecture, so I tend to see systems as interconnected processes rather than isolated applications.
There were things I could envision in our products that were difficult to translate at the speed I wanted.
That wasn’t necessarily a problem with the people involved. There is simply a distance between seeing something in your mind and transferring that vision through an organization.
- Explain it.
- Document it.
- Build it.
- Review it.
- Correct it.
- Test it.
- Repeat.
I began wondering what would happen if AI could shorten that distance.
Rather than simply encouraging my team to use more AI before I fully understood how it would fit into our work, I wanted to understand its capabilities firsthand.
So I started working with it directly.
And the results surprised me.
The Feedback Loop Collapsed
Because I already understood the architecture, business processes, databases, dependencies, and desired outcome, I could give AI context and continuously steer it.
- Build this.
- Look at this existing logic.
- That's not quite right. Here's why.
- Change this.
- Now account for this condition.
- Test it again.
Something that traditionally involved multiple handoffs could suddenly become a continuous conversation.
The feedback loop began collapsing from days to hours and, sometimes, minutes.
Work that I would previously have measured in weeks started being measured in days. Things that had been sitting in my backlog suddenly looked achievable.
It was incredibly empowering.
And that’s when I noticed a problem I wasn’t expecting.
Just Another 30 Minutes
I started having difficulty stopping.
Not because AI was frustrating.
Quite the opposite.
I didn’t want to stop because it was working so well.
Something that had been sitting in the backlog for months could suddenly be completed tonight.
So why not spend another 30 minutes?
Then another problem looked solvable.
Maybe another 20 minutes.
Then another.
Before AI, there were natural stopping points.
A developer would continue tomorrow. Someone needed to respond to an email. A meeting had to be scheduled. An analyst needed time to prepare something.
Those pauses created boundaries whether we wanted them or not.
AI doesn’t create many of those boundaries.
- At 7:00 PM, it's available.
- At 10:00 PM, it's available.
- At midnight, it's available.
And every time you ask, “Can we do one more thing?”
The answer is effectively:
“Yes.”
I Started Calling It AI Fatigue
I don’t mean being tired of hearing about artificial intelligence.
I don’t mean struggling to learn another technology.
And I don’t mean fear of AI replacing people.
I mean something almost opposite.
AI Fatigue isn’t fatigue from using AI. It’s the fatigue that can come from not knowing when to stop using it.
When AI dramatically increases our productive capacity, the temptation is to continuously ask ourselves to do more.
If something that once required eight hours now takes one, logically you have gained seven hours.
But that isn’t necessarily how ambitious people behave.
We don’t think:
“Great. I just got seven hours of my life back.”
We think:
“What else can I get done?”
That distinction has been bothering me.
This Isn't a Developer Problem
Development is simply where I experienced it first.
The more I thought about it, the more I realized that this could affect almost any knowledge worker—and perhaps leaders in particular.
A CEO finishes reviewing a strategy and immediately asks AI to evaluate three alternatives.
A CFO finishes analyzing financial performance and starts modeling another scenario.
A sales leader finishes a proposal and immediately begins researching the next opportunity.
A marketing executive finishes a campaign and starts developing the next one.
An operations leader documents a process and immediately begins redesigning it.
A founder has an idea at 10:30 at night and, instead of writing it down for tomorrow, can research it, challenge it, model it, and turn it into something tangible before going to sleep.
Different jobs.
Same phenomenon.
The friction that once separated thinking from execution is disappearing.
And some of that friction used to tell us when to stop.
The Backlog May Be Infinite
This is the part I am still learning.
I originally looked at AI and saw a way to attack my backlog.
Now I’m beginning to wonder whether that is the wrong objective.
Because the backlog never actually ends.
Complete one improvement and you see three more.
Automate one process and you discover another.
Build something that previously seemed impossible and suddenly five other possibilities become realistic.
Ironically, the technology that allows us to accomplish more may also continuously reveal more things worth accomplishing.
So perhaps the real constraint hasn’t disappeared.
It has simply moved.
The bottleneck used to be execution capacity.
Increasingly, the bottleneck may be our own attention, judgment, energy, and ability to decide what deserves to be done.
Learning to Stop
For decades, we have been taught productivity disciplines designed around scarcity.
How do I manage my time?
How do I prioritize?
How do I delegate?
How do I accomplish more during the hours available to me?
AI may introduce a very different problem.
What happens when those same hours suddenly produce dramatically more?
Maybe AI-era productivity isn’t only about learning how to accomplish more.
Maybe it is also about learning when enough has been accomplished.
I am still figuring this out myself.
I remain incredibly optimistic about AI. What I have experienced firsthand has strengthened my belief that AI, when combined with people who deeply understand their business and domain, can fundamentally change what individuals and organizations are capable of accomplishing.
But increased capability may require increased discipline.
AI doesn’t get tired.
It doesn’t need to go home.
It doesn’t have dinner waiting.
It doesn’t look at the clock and tell you that you’ve accomplished enough for today.
That decision still belongs to us.
The goal may not be to eliminate the backlog.
It may be to stop treating the backlog as a measure of what matters.
There will always be another improvement, another idea, another problem AI can help us solve. The harder and more important question is whether solving it now is worth extending the day, consuming our attention, or postponing the parts of life that cannot be optimized.
AI can help us do more than ever before.
But it cannot tell us what more is for.
That decision still belongs to us.
I would be interested to hear how other leaders are thinking about this.
Have you experienced the same pressure to keep going simply because AI makes one more task possible? And what boundaries, habits, or expectations are you putting in place to ensure that greater capability leads to better decisions—not just longer days?