On a good morning, my AI workflow was impressive. Clear prompts, sharp outputs, fast iteration. I could go from idea to published article in under two hours.
On a Tuesday afternoon after a bad night’s sleep, the same workflow produced nothing. Not because the tools stopped working – but because I did. The prompts were too open. The steps required too many decisions. The whole thing assumed I’d show up with the same energy every time.
I don’t. Neither do you.
So I rebuilt the entire thing around a different question. Not “how do I get the most out of AI when I’m sharp?” but “how do I make this work when I’m running on fumes?”
That turned out to be the only version worth keeping.
Why most AI workflows break when energy drops
The tools aren’t the problem. ChatGPT, Claude, whatever you use – they work the same whether you’re rested or exhausted. The bottleneck is everything that happens before you get to the tool.
Think about what a typical AI writing workflow actually asks of you. First, decide what to write about. Then figure out the angle. Then craft the right prompt. Then choose a format. Then evaluate the output and decide what to keep.
That’s five decisions before you’ve produced a single usable sentence.
On a high-energy day, those decisions feel like nothing. You blow through them on autopilot. But decision-making is a resource, and it depletes faster than most people realize. By Tuesday afternoon – or any moment when your mental reserves are low – each of those open questions becomes a wall.
For me, the first thing that always broke was the “decide what to write about” step. I’d open my laptop, stare at a blank prompt, and feel the familiar weight of infinite possibility. Which is really just a fancy word for friction.
The blank prompt is the enemy of the tired mind. Not because you lack ideas – but because selecting one requires energy you’ve already spent elsewhere.
The principle – design for depletion, not peak
Most productivity advice assumes you’re operating at full capacity. Optimize your morning routine. Batch your tasks. Use AI to 10x your output.
That’s fine for the version of you that shows up caffeinated and clear-headed. But that version is maybe 30% of your working hours. The other 70% is the slightly foggy, somewhat distracted, definitely-not-peak version. That’s the one your systems need to serve.
The shift I made was simple in concept: instead of asking “how do I get more out of AI,” I started asking “how do I need less from myself?”
That reframe led to three design principles.
Pre-decided inputs. The topics, angles, and formats are chosen in advance – during a high-energy session. The low-energy version of me doesn’t decide what to create. He picks from a list that past-me already curated.
Minimal branching. Every fork in a workflow is a decision. Every decision costs energy. So the workflow should have as few forks as possible. Ideally one: which item from the queue do I work on today?
A single entry point. One file to open. One place to start. Not a collection of bookmarks, not a folder of prompts, not a mental model of which tool to use when. One door.
My old workflow had five decision points before writing started. The new one has one. That’s not a minor improvement – it’s the difference between a system I use and a system I used to use.
Building a low-energy AI workflow
Here’s how this looks in practice. Four layers, each one removing a category of decisions.
Layer 1: The topic queue. This is a simple list – I use a spreadsheet, but a text file works – where I capture article ideas, angles, and working titles during high-energy moments. The key is that this list is pre-filtered. Everything on it has already passed the “is this worth writing” test. When I sit down tired, I don’t generate ideas. I pick one.
The queue gets filled during planning sessions, after reading something interesting, or when an idea strikes during a walk. The point is separation: generating ideas and executing on them happen at different times, matched to different energy levels.
Layer 2: Pre-built templates. Each content type has a corresponding prompt template. An article has one. A newsletter has one. A social post has one. The templates include the structure, the tone guidance, and the context the AI needs.
I don’t write prompts from scratch anymore. I open a template, drop in the topic from the queue, and run it. The template does the thinking that past-me already did.
Layer 3: A single starting action. My entire workflow starts with opening one file. That file contains the queue, links to the templates, and the current week’s content plan. No hunting through bookmarks. No remembering which tool I was using for what.
This sounds trivially simple, and it is. That’s the point. The tired version of me can do exactly one thing reliably: open a file. So that’s what the system requires.
Layer 4: A default output format. Instead of deciding whether something becomes a blog post, an email, or a social thread, I have defaults. Articles go through the article template. The newsletter gets written on Thursdays. Social posts get extracted from finished articles, not created from scratch.
Every removed decision is energy preserved for the work that actually matters – thinking clearly about the content itself.
The whole sequence, from opening the file to having a working first draft, involves about five minutes of actual decision-making. The rest is execution. And execution is something even the tired brain can handle, as long as the path is clear.
The real measure
There’s a paradox here that took me a while to accept. The less my workflow demands of me, the more consistently I produce. Not because I’m lazy – but because consistency comes from removing barriers, not from adding willpower.
Every system you build will be tested on a bad day. The flashy, optimized, peak-performance version will sit unused while you scroll your phone because the activation energy was too high.
The boring, minimal, one-entry-point version will still be there when you’re tired. And it’ll work.
Build your AI workflow for your worst day. If it survives that, it’ll survive everything.
