Which AI tool to use, and why it matters less than you think
Why the choice matters less than you think, and what actually decides it.
If you've spent any time trying to choose the right AI tool, you've noticed two things. Everyone has strong opinions, and the recommendations contradict each other.
Here is the honest take. For almost every small and medium team, the choice matters far less than the comparison articles want you to believe. What matters is picking one, using it consistently, and getting good at it.
The four you're choosing between
As of 2026 it is realistically ChatGPT, Claude, Gemini or Copilot. ChatGPT is the best known and a strong generalist, occasionally inconsistent. Claude is stronger on writing, reasoning and long documents, and tends to produce more measured output. Gemini lives inside Google Workspace. Copilot lives inside Microsoft 365 and runs on models from OpenAI and other providers.
Others exist, such as Perplexity, Mistral and Llama, but for a small or medium team those four cover the territory. Pricing is roughly the same across all of them: around €20 to €30 per user per month for a paid tier.
Where they genuinely differ
Integration with the suite you already work in
Default tone: measured, cheerful, or somewhere between
Holding a very long document in mind at once
Image and file analysis
Admin controls and data terms on team tiers
Where they're interchangeable
First drafts of emails and messages
Summarising a meeting or a long document
Brainstorming headlines and angles
Turning rough notes into structure
Translating, reformatting, shifting tone
You can always change tool later. You can't get back the time you spent comparing.
Where they're actually similar
For the work most teams actually do, the major tools are interchangeable to a degree that surprises people. A first draft of an email, a summary of a meeting, ten headline options, rough notes turned into structure: you would not reliably tell which tool produced which.
The variable that does move the output is your prompt. A well-prompted task on the "worse" tool beats a badly prompted one on the "better" tool, every time. That is exactly why the Copy Drafter on The Bench exists: it builds a structured prompt in a few clicks, and the result pastes into any of the four.
Pick one and get on with it
Three ways people lose months here. Tool-hopping, trying every new model that appears and never building real fluency with any of them. Comparison paralysis, reading about the tools instead of using them. And trusting the reviews, most of which weigh a tool someone used for an hour against one they have used for a year, and a fair number of which are sponsored.
Three things to do this week
01
Start with whatever you're already working in
On Google, start with Gemini. On Microsoft, Copilot. Neither, pick ChatGPT or Claude. Reduce the friction to nearly nothing.
02
Get on a paid business tier
About €25 to €30 per person a month buys the data protections, and the paid quality is meaningfully better than free.
03
Give it three months before you reassess
Real fluency takes time. Until then, "is this the right tool" is mostly a way of not starting.
Conclusion
01
"Which tool is best" is the wrong question. The one you actually use, consistently, wins.
02
On everyday work the major tools are far more alike than the comparisons suggest.
03
The thing that changes your output is your prompt, not your choice of tool.
Go to next