試過上百個 AI 工具後，真正留在我工作流裡的只有這幾個

Doing this line of work, new tools emerge one after another, and I've almost tried every popular one. But to be honest, trying something out and sticking with it are two different things. Most tools are novel for only three days before I forget about them, and there are actually very few that I open every day and feel uncomfortable leaving behind. This article shares those "few" tools and why they've stuck around — perhaps it can help you try out fewer tools and find your own combination sooner.

The tools that stuck around all have one thing in common

First, let me share my observation: the tools that have stuck around aren't the ones with the most impressive features, but rather those that genuinely solve a problem I encounter every day and seamlessly integrate into my workflow. No matter how powerful a tool is, if it's cumbersome to use or only used once a week, it will eventually be phased out.

1. A primary conversational AI — my "universal second brain"

ChatGPT is the tool I open the most. I ask it questions, write drafts, brainstorm ideas, and debug issues — almost everything starts with it. The reason it's stuck around is simple: it can help with so many things that it's become an extension of my thinking. Everyone should have a "universal" tool like this.

2. A writing assistant — turning drafts into manuscripts

When writing, I use Claude to help me refine my drafts, identify logical flaws, and rephrase sentences. It's stuck around because its writing is more engaging and helps me turn "readable" drafts into "engaging" manuscripts. For those who work with text, the value of this type of tool is very practical.

3. An AI search engine for researching — saving me from opening ten tabs

To quickly grasp a topic, I use Perplexity because it provides answers with sources, so I don't need to open multiple tabs to compare information. It's stuck around because it genuinely saves me a lot of time researching (I've written about how to use it in this article).

4. An automation tool — outsourcing tedious tasks

Zapier helps me automate repetitive tasks (data transfer, notifications, aggregation). It's stuck around because of the satisfaction of "setting it up once and having it run automatically forever" — tasks that machines should handle shouldn't waste my time.

5. A tool for organizing notes and data

To turn a pile of data into a knowledge base that can be queried, I use NotebookLM because it only answers based on the data I provide and includes sources, without providing irrelevant information.

My advice to you: don't be a tool collector

I've seen too many people get caught up in the cycle of constantly trying new tools, installing a bunch, but not mastering any. Instead of broadly and shallowly trying out many tools, it's better to choose a few core tools and use them extensively. What truly makes you stronger isn't how many tools you know, but how deeply you use a few tools.

The standard for choosing tools is simple: does it solve a problem you encounter every day? Is it user-friendly? If so, keep it; if not, phase it out. Don't let the fear of missing out hold you back. For further reading: Common pitfalls for AI beginners.

In a nutshell: trying something out and sticking with it are two different things — the tools that truly stick around in your workflow aren't the most impressive, but those that solve your daily problems and are easy to use. Don't be a tool collector; choose a few and master them.

Sources

Compiled from the long-term, actual usage experience of our editorial team.