Why AI is making you feel more overworked than ever

I'm going to try very hard to keep this as little about coding as possible.

That's because I think the people who really need to learn this are non-programmers.

Programmers are painfully aware of the limitations and growth in LLM capability in the past years.

Words like context, memory, and MCP servers are things we've been dealing with for years with huge implications on our daily work lives.

Non-programmers probably hear "context" and think:

"How large of a PDF can I fit into this chat?"
"How long can I take this chat before the LLM stops making any sense?"

The “Holy Grail” (Then vs Now)

When I first started using AI to code — basically the day it came out — I told people the holy grail would be giving it access to my entire codebase.

Back then, if you wanted ChatGPT to work on one file that had implications on other files, you had to paste everything into the chat so it could understand the full picture.

Today, we don’t have to do that anymore.

Tools like Claude Code, Codex, and Cursor now give AI full access to your codebase by default.

With integrations like GitHub, you can even ask ChatGPT questions about your business, and it can fetch the relevant code before answering.

For programmers, this has been incredible.

But it also leads to a strange new problem:
When the AI can do almost anything… what do you actually do?

This Isn’t Just a Coding Problem

I said I’d avoid coding, but it’s hard — because AI has impacted that field more than any other.

Because of that, most tooling and features are built for programmers.

But what about everyone else?

What about white-collar workers using AI every day?

They’re dealing with the same problems — just in a different form.

Why Everyone Feels Like They’re Drowning

There’s a common theme I keep seeing:

People are drowning in work.

AI was supposed to:

Reduce workload
Automate tasks
Give us more free time

Instead… it’s often doing the opposite.

If AI is so powerful, why:

Can’t you finish that investor prospectus twice as fast?
Can’t you write 10 blog posts instead of 2?

Here’s my (completely unscientific) take:

AI has 10x’d our ability to feel productive — without actually doing more.

I’ve felt this myself.

In product work, I can spend three weeks iterating on features, going in circles, and never actually shipping anything useful.

So What’s Actually Going Wrong?

Simple:
AI isn’t quite there yet.

It might be very, very, very close.

But right now, you still have to:

Verify outputs
Double-check reasoning
Oversee every step

And in high-stakes industries (medicine, legal, finance, engineering), even a 1% error rate is unacceptable.

The Real Bottleneck: Context

The core issue is this:

Your AI doesn’t truly understand your business.

It doesn’t know:

Your priorities
How you think
How you communicate
What actually matters

The Current “Solutions” (That Don’t Fully Work)

1. Memory

AI tools store things about you, like:

Your preferences
Your thinking style
Personal background

Sometimes this helps.

Sometimes… it’s unbearable.

You end up with things like:

“Let’s break this down to first principles…”
“Marcus Aurelius would say…”
“Given your childhood experiences…”

And you’re thinking:

“I don’t want to talk about my childhood right now.”

At scale, this gets even worse.

AI lacks:

Timing
Tact
Real empathy

2. Knowledge Bases

Tools now let you upload files or create “projects” the AI can reference.

In theory, this should fix everything.

In practice… it doesn’t.

Because having knowledge ≠ using it correctly.

The Real Fix (In My Opinion)

The solution isn’t new tech — it’s better use of what already exists.

1. Better Knowledge Retrieval

It’s not enough to store information.

You need:

Smarter retrieval
Better guidance on when to retrieve
Better filtering on what to retrieve

Most AI providers won’t fully solve this.

Why?

Because it’s expensive.

Running deep retrieval (RAG) on every message would massively increase their costs.

2. Smarter Memory Usage

Memory shouldn’t just exist — it should be:

Context-aware
Selectively applied
Organized by relevance

Right now, AI just sprays memory into conversations without nuance.

3. Strict, Enforced Workflows

This is the missing piece almost nobody talks about.

AI needs:

Clear steps
Checklists
Non-negotiable processes

Otherwise, it will:

Skip steps
Cut corners
Optimize for speed (not quality)

Even inside tools like ChatGPT or Claude, you have to force the workflow.

You literally have to say:

“Continue following the workflow.”

And it will.

Why This Problem Exists

AI companies are in a tough spot.

They:

Lose money on subscriptions
Limit usage behind the scenes
Optimize for cost efficiency

That’s why you see things like:

“We’ve done a lot for today, ready to call it quits?”

The model isn’t just thinking — it’s budgeting.

What You Should Actually Do

If you’re a non-programmer using AI daily:

1. Build a Knowledge Base

Give your AI access to everything it needs:

Docs
Content
Internal knowledge

2. Keep Memory Tight

Avoid clutter
Remove contradictions
Keep it minimal

3. Enforce Workflows

Don’t trust AI to “figure it out.”

Give it:

Steps
Structure
Clear completion criteria

The Direction I’m Betting On

For writing-heavy workflows, I think the winning combination is:

Centralized knowledge
Personalized writing voice
Strict execution workflows

What I Built

If you do a lot of writing, I built something specifically to solve this.

Inside Wordflows AI, you can:

Build a Content Brain

Pull in your entire website
Keep it updated automatically

Define Your Writing DNA

Create a consistent voice
Make AI sound like you

Run Structured Workflows

Produce content that’s:
- Authentic
- Emotional
- Factually accurate
- Ready to publish

All without babysitting the AI.

If you’re tired of drowning in AI work, try it out today.