Outcome
You stop defaulting to whichever AI tool you opened first, and start matching the tool to the job — including the jobs where the right answer is "don't use AI at all."
Concept
There's no single "best" AI assistant — there's a best assistant for what you're doing right now. Here's a neutral, no-hype map of how the major players tend to differ. (Free-tier limits and specific features change often, so treat the details below as general shape, not gospel — check each provider's current pricing page before you commit to one.)
- Claude (Anthropic): Strong at longer, more careful writing and analysis, and tends to be more cautious about making things up. A solid default for drafting, editing, and working through anything with real nuance.
- ChatGPT (OpenAI): The most widely used, with the biggest ecosystem of plugins and integrations. Strong all-rounder, especially if you want one tool that also does images, voice, and code in the same place.
- Gemini (Google): Deep integration with Google's own apps (Docs, Gmail, Drive), which matters a lot if that's already where your work lives.
- Copilot (Microsoft): Built into Windows and Microsoft 365 — the natural choice if your day-to-day is Outlook, Word, and Excel.
None of these is "cheating" or "the wrong one" — they're built by different companies with different priorities, and all of them are capable. Vendor frameworks and naming conventions belong to their vendors — if you hear about a specific company's official method or framework, that's their idea, not a universal law of AI.
The real decision usually isn't "which AI is smartest" — it's:
- Where does this task already live? (If it's in Gmail, Gemini has an edge. If it's in Word, Copilot does.)
- How sensitive is the information? Read the tool's privacy policy before you paste in anything personal, financial, or confidential — free tiers in particular sometimes use your data to improve their models unless you opt out. When in doubt, don't paste it in.
- Do I even need AI for this? Some tasks are faster done by hand, or need a level of certainty AI can't reliably give (see Lesson 1). Knowing when to not reach for AI is part of using it well.
A worked example: three tasks, three decisions
Walk three real tasks through the questions above.
Task 1: "Reply to a client email thread in my inbox." The task lives in your email, so an assistant built into your email app saves the copy-paste. The content is a normal business exchange, so sensitivity is moderate: skim for anything confidential before sending it anywhere. Decision: use the assistant closest to your inbox, and review before sending.
Task 2: "Help me understand my medical test results." This one is sensitive and high-stakes. Decision: don't paste identifying details into a general assistant. If you use AI at all, ask it for general background on terms and a list of questions to bring to your clinician, and treat the clinician as the source of truth. (General education only, not medical guidance.)
Task 3: "What is 18.5% of $4,260?" A calculator is faster and exact. Decision: no AI. If you want AI involved, ask it to explain how to set up the formula, then run the numbers yourself.
Same person, same day, three different answers, and one of them is "none."
Copy-paste prompts
A prompt for stripping sensitive details before you paste:
I want to ask for help with the text below, but I don't want to share personal
details. List every name, number, address, and identifying detail in it, and
suggest neutral placeholders for each. Text: [PASTE TEXT]
A side-by-side comparison prompt (use the identical text in two assistants):
Task: [YOUR REAL TASK]. Context: [2-3 SENTENCES]. Give me one version, under
[LENGTH]. Then tell me what you assumed.
A "should I even use AI?" check:
Here's a task I'm considering giving to AI: [TASK]. Tell me honestly whether
this is a good fit, what could go wrong, and what I'd need to verify by hand.
Build
Pick one real task you have coming up and run it through two different assistants — your current default plus one you don't normally use. Compare:
- Which result needed less editing?
- Which felt like it understood the context faster?
- Did either one surface something the other missed?
You're not looking for a permanent winner — you're building a feel for where each tool has an edge.
Try it now (5 minutes)
Use the "should I even use AI?" prompt on three tasks from your own week. Then send one of them, using the comparison prompt, to two assistants.
Done when: you have a yes, no, or "yes with checks" for all three tasks, and a one-line note on which assistant you'd pick for that kind of task and why.
Run
A simple decision rule you can reuse:
- Sensitive data involved? → Check the privacy policy first, or don't use AI for this one.
- Task already lives inside a specific ecosystem (Google, Microsoft)? → Use that ecosystem's assistant for the integration advantage.
- Need careful, nuanced writing? → Reach for the tool you trust most for quality over speed.
- Quick, low-stakes task? → Use whichever is fastest to open. The tool matters less here.
- Task needs certainty AI can't give (exact facts, high-stakes numbers)? → Don't use AI as the source of truth — use it to draft, then verify by hand.
You don't need to be loyal to one tool. The people who get the most out of AI are the ones who match the tool to the job — and aren't afraid to skip it entirely when it isn't the right fit.
Common mistakes
- Pasting first, thinking later. Decide what's sensitive before you paste, not after.
- Judging a tool on one answer. Compare on a real task, ideally twice.
- Assuming a feature exists. Features and limits change often; check the provider's current page instead of relying on a summary, including this one.
- Using AI because it's there. Some tasks are faster, safer, or more accurate by hand.
Checkpoint
- I've tried the same real task in at least two assistants.
- I know where my most common tasks live and which assistant sits closest.
- I have a personal rule for what I won't paste into a general assistant.
- I can name one task where I'd choose no AI at all.