Lesson 3 of 8 · 8 min read
You now know what AI is and how it learns. The obvious next question: which one do you actually open?
There’s a whole industry built on answering that, mostly badly. Benchmark tables, leaderboard screenshots, “we tested all four” posts that were out of date the week they published. It’s noise, and for a beginner it’s worse than noise — it makes a simple decision feel risky.
Here’s the shortcut: for what you’ll be doing, they’re much more alike than different. The gap between the leading tools is far smaller than the gap between someone who prompts well and someone who doesn’t. Pick one. Learn to use it properly. That’s the whole game.
This lesson is a map, not a ranking.
The four you’ll actually meet
ChatGPT (OpenAI) is the one that started the noise in 2022 and still the one most people mean when they say “AI.” Broadest feature set, most third-party integrations, and the one your colleagues have probably used personally. Its main advantage for a beginner is gravity — most guides, tutorials, and templates you’ll find were written with it in mind.
Gemini (Google) lives where your Google life already is. Gmail, Docs, Drive, Sheets, Android, Chrome. If your organisation runs on Google Workspace, it’s sitting inside tools you have open anyway. Also strong on very long documents and on audio and video, if that’s your work.
Microsoft Copilot is AI inside Word, Excel, PowerPoint, Outlook, and Teams. Its value isn’t the model underneath — it’s proximity. It can see your actual documents and emails, so you’re not copying and pasting context in. If your company is a Microsoft shop, this may already be switched on and you haven’t noticed.
Claude (Anthropic) tends to be preferred for longer writing and for following detailed instructions closely. Less consumer reach than ChatGPT, more common among people whose work is mostly words or code.
There are others worth knowing exist — Perplexity for research with citations, Grok, various open-source models — but the four above cover the overwhelming majority of workplace use, and you shouldn’t be shopping yet.
Two kinds of tool, not four
This is the distinction the comparison posts usually bury, and it’s the one that actually decides things for you.
Standalone assistants — ChatGPT, Claude. You go to them. Blank box, you bring the context, you do whatever you want. Maximum flexibility, but you have to paste in your material.
Embedded assistants — Copilot, Gemini. They come to you. They sit inside the suite you already work in and can already see your files, your inbox, your calendar. Less flexible, far less friction.
So they’re not four competitors in one race. They’re two different answers to two different questions: what tool do I want to think in versus how do I get AI into the software I already live in.
Most people end up with one of each, and that’s fine.
Which should you start with?
Genuinely, whichever is nearest.
If your workplace already pays for one — start there. Not because it’s best, but because it’s approved, it’s connected to your actual work, and you won’t be pasting company information into an unapproved tool. That last point matters more than any benchmark. We’ll come back to it in Lesson 8.
If nothing’s provided, start with ChatGPT. Not because it wins on any particular measure, but because it has the most tutorials, the most community, the most people you can ask for help. For learning, ecosystem beats capability.
If you live in Google Docs all day, use Gemini. If you live in Outlook and Excel, check whether Copilot is already in your ribbon. The one you don’t have to switch tabs for is the one you’ll actually use.
That’s it. That’s the decision. It should take you two minutes, and it’s reversible.
The free tiers are enough
You do not need to pay anything to complete this course, or to become genuinely competent at this.
Every major tool has a free tier, and they’re generous enough for real work. The paid tiers mostly buy you higher usage limits, faster responses, and access to the newest model — none of which matter while you’re learning what a good prompt looks like.
The advice you’ll see to “just pay the twenty quid, it’s worth it” isn’t wrong for heavy users. It’s premature for you. Get to the point where you’re hitting the free limits regularly. That’s the signal you’re ready to pay, and by then you’ll know which one to pay for — because you’ll have used two or three.
The version numbers don’t matter
You’ll see a lot of noise about which model is currently ahead. A new one launches, it tops a leaderboard, and there’s a wave of posts declaring a new winner. Six weeks later it happens again with a different name.
Ignore all of it. Two reasons.
First, the lead changes constantly and the differences at the top are marginal for ordinary tasks. Drafting an email, summarising a report, cleaning up some notes — every leading tool does these well. Benchmark gaps show up on hard reasoning problems, not on your inbox.
Second, and more usefully: the model isn’t your bottleneck. You are. A well-constructed prompt on last year’s model beats a lazy one on the newest release, every time and without exception. Every hour you spend comparing tools is an hour not spent getting better at the thing that actually moves the needle.
Which is the next lesson, and the one after it.
Don’t tool-hop
The most common beginner mistake isn’t picking the wrong tool. It’s picking four.
It looks like diligence — trying everything, staying current, keeping options open. It isn’t. Every tool has its own quirks: how it responds to instructions, what it’s fussy about, when it goes off. That knowledge is real, it’s earned by repetition, and it doesn’t transfer cleanly between tools.
Spread thin across four, you never build it with any of them. You stay a beginner four times over.
Pick one. Use it for everything for a month, including things you’re not sure it can do. Then, if you’re curious, try another — from a position of actually knowing what good looks like.
Key takeaways
- The leading tools are far more alike than the comparison posts suggest — the gap between them is smaller than the gap between good and bad prompting
- Two categories, not four: standalone tools you go to (ChatGPT, Claude) and embedded ones that come to you (Copilot, Gemini)
- Start with whatever your workplace provides; if nothing, start with ChatGPT for the ecosystem
- Free tiers are enough to become competent — pay only when you’re regularly hitting the limits
- Ignore model version news; the model isn’t your bottleneck
- Use one tool for a month before trying another
Next: Lesson 4 – Writing Your First Prompts
Tool chosen. Now the part that actually decides whether AI is useful to you: what you type into it.