Artificial Intelligence

Quick Takes: AI Tools I Actually Use Daily

By · Updated

The AI tools that survive daily use are not the flashiest ones. They remove friction from work you already understand.

Minimalist desk setup with laptop and notepad, neutral tones
Photo by Unsplash

Most AI tools do not survive Monday.

They look impressive in a demo.

Then real work arrives and suddenly the tool needs setup, cleanup, verification, permissions, credits, prompts, and a small prayer.

No thanks.

The AI tools I keep are boring in a specific way:

They remove friction from work I already know how to judge.

That is the rule.

If I cannot evaluate the output, I do not hand it the job.

ChatGPT: the general workbench

The move: Use ChatGPT for mixed work that changes shape.

I use it for:

  • brainstorming
  • quick explanations
  • data cleanup
  • draft outlines
  • image or file questions
  • messy planning
  • sanity checks

ChatGPT is useful because it can move between modes without making me open five different apps.

That is the win.

Not perfection.

Range.

Where I do not trust it: final facts, final voice, final judgment.

It can be useful and still need supervision.

If that sentence feels annoying, good. That sentence saves work.

Claude: the thinking and writing pass

The move: Use Claude when the shape of the answer matters.

Claude is the one I reach for when I have:

  • a rough article
  • a messy argument
  • a client email that needs tact
  • a long document
  • a piece of writing that sounds too smooth
  • a code review where I want a careful second pass

It is not “the writer.”

I am the writer.

It is the sharp reader who helps me see where the piece got lazy.

The difference matters.

If you let AI become the author, your work starts sounding like it was assembled in a room with no windows.

For the longer version, read AI writing tools: what they’re actually good at.

Cursor or Copilot: coding without blank-page tax

The move: Use a coding assistant for the obvious parts, then stay awake for the important parts.

Cursor and GitHub Copilot are useful because they sit close to the code.

They can suggest edits, explain errors, draft tests, fill in patterns, and help you move through a codebase without pretending you have every syntax detail memorized.

That is great.

That is also not the same as architecture.

I use coding AI for:

  • small scripts
  • tests
  • refactors I understand
  • boilerplate
  • debugging clues
  • reading unfamiliar code

I do not use it to make decisions I cannot explain.

If the code matters, you still need to read it.

Sorry. The machine did not abolish responsibility.

Perplexity: research starting point, not final authority

The move: Use AI search to get oriented quickly, then verify the important bits.

Perplexity is useful when I need context fast:

  • what changed in a topic
  • which sources to read first
  • how people describe a trend
  • what terms I should search next
  • where the disagreement is

The useful part is not that it gives an answer.

The useful part is that it shows sources and gives me a map.

Maps are not territory.

Do not cite the summary if the source matters.

Open the source.

Read the source.

Be an adult for three minutes.

Otter or any solid meeting notes tool

The move: Use meeting AI to listen better, not to stop listening.

Meeting transcription and summaries are useful for:

  • client discovery calls
  • sales calls
  • long planning meetings
  • interviews
  • action-item capture

But there is a trap.

The transcript can make you lazy.

You think, “The tool has it,” and then you stop hearing the human in front of you.

Bad trade.

Let the tool catch details.

You stay present for judgment, emotion, risk, and the sentence the client almost said but softened at the end.

That is where the work is.

The filter

Before I keep any AI tool, it has to pass four tests.

Does it save time after cleanup?

Not before cleanup.

After.

That is the honest number.

Can I judge the output?

If not, I am not using it for that job.

AI is not a substitute for expertise you do not have.

It is a multiplier for judgment you already built.

Does it fit where I work?

A great tool in the wrong place becomes another tab.

Another tab becomes another excuse.

Would I pay for it if I had to cut the stack in half?

That question gets rude fast.

Good.

Subscriptions should fear questions.

The bottom line

My daily AI stack is not sacred.

It changes.

The jobs stay the same:

  • think
  • draft
  • code
  • research
  • capture meetings

Use AI where it removes friction from work you understand.

Do not use it where it gives you a confident answer you cannot inspect.

That is how tools become liabilities with nice buttons.

If you want the broader workflow, read how I actually use AI in my daily workflow. And before you hand over too much, read what I stopped delegating to AI.

The goal is not to use more AI.

The goal is to do better work with fewer stupid detours.