“Pick Two, Not Seven.” But Only One Of Them Is Your Choice
Seven tools, one job each, laid out as seven panels with the same three fields: what it helps with, what it is best for, and — the useful one — USE WHEN. Along the bottom, in small type, the instruction that makes the whole card work: pick two, not seven.
That is the right advice and most tool charts never give it. But read the seven USE WHEN lines carefully and something becomes clear about the second of those two picks.
“Your business runs on Microsoft.” “You need to understand a video or work in Google tools.” “You want notes without making the meeting awkward.”
Those are not descriptions of a task. They are descriptions of a situation you are already in — the software your employer bought, the ecosystem your files live in, and how the other people on the call feel about being recorded. Nobody picks those. They were picked before you opened the chart.
Which gives the footer a sharper reading than it first appears to have. Pick two, yes — but only one of the two is a decision. The other is your environment, telling you what it already is.
Three kinds of “use when”
| DECIDED BY THE TASK | DECIDED BY YOUR STACK | DECIDED BY OTHER PEOPLE |
|---|---|---|
| ChatGPT · Claude · Perplexity · Gamma | Microsoft Copilot · Gemini | Granola |
| What you are trying to produce today | Which ecosystem your company already pays for | Whether everyone on the call is comfortable |
| Yours. This is the real choice | Already made, usually by someone in finance | Not yours alone — and asking is part of the job |
| Change it any week you like | Changing it means changing the company | Changing it means a conversation, not a setting |
All seven, as the card gives them
Every word in the first four columns is the card’s own.
| Tool | Its role, as named | Best for | Use when | Who decides |
|---|---|---|---|---|
| ChatGPT | Your daily work assistant | Planning work · first drafts · simple processes · summarising notes | “You need help getting everyday work done faster.” | Task |
| Claude | Your builder | Building tools · workflows · simple apps · agentic processes | “You want to build something practical from a messy idea.” | Task |
| Perplexity | Your research helper | Market research · competitor checks · fact checking · source finding | “You need evidence before making a decision.” | Task |
| Gamma | Your slide deck helper | Sales decks · workshop decks · training slides · internal updates | “You need a decent deck quickly.” | Task |
| Granola | Your quiet meeting note taker | Meeting notes · action points · client calls · sales calls | “You want notes without making the meeting awkward.” | Other people |
| Microsoft Copilot | Your Microsoft work helper | Outlook · Teams · Word · Excel | “Your business runs on Microsoft.” | Your stack |
| Gemini | Your YouTube helper | Video summaries · video questions · finding videos · Google workflows | “You need to understand a video or work in Google tools.” | Your stack |
WATCH THE ORIGINAL — 20 SECONDS
All seven panels with their best-for lists and use-when lines, and the footer telling you to pick two.
The only panel sold on a social fact
Six of the seven panels justify themselves with something the tool can do. One does not.
The meeting-notes entry is described as capturing notes without adding a bot to the call, and the reason to use it is you want notes without making the meeting awkward. Nothing there is about accuracy, speed or features. It is about how a room full of people feels when a piece of software visibly joins the call.
That is the most honest line on the card, and it is worth dwelling on because it generalises. For anything used in front of other people, the binding constraint is rarely capability. It is whether the other participants are comfortable, whether the client notices, whether somebody has to explain it at the start of every call. A tool that is slightly worse and causes no friction will beat a better one that makes every meeting start with an apology.
There is a harder edge to this row too, and the post beneath the card says it plainly: keep participants informed on how notes are recorded and used. That is not etiquette. In a great many places, recording or transcribing a conversation carries real obligations about telling people and sometimes about asking them. “Quiet” describes the interface, not the ethics — you still say at the top of the call that notes are being taken and where they will go. Said once, in one sentence, it costs nothing. Discovered later, it costs a client.
Start with the task, and the card tells you how to choose one
The best part of this post is not on the card at all. It is the list of what makes a task worth automating in the first place, and it is unusually precise.
| The criterion | What it means in practice | Why it matters | The trap |
|---|---|---|---|
| A clear definition of “done” | You can say in one sentence what finished looks like | Without it you cannot tell whether the tool helped | “Make it better” is not a task |
| Consistent inputs | The same kind of thing arrives every time | Varied inputs break any workflow, automated or not | One odd file a week will eat the time you saved |
| A measurable quality standard | A number or a check somebody else could apply | Otherwise you are grading your own homework | Feeling faster is not being faster |
| A designated owner | One named person who fixes it when it breaks | Unowned workflows rot silently | “The team” is not an owner |
| Measure the whole workflow | Include review, editing and handoff in the timing | Generation got faster; checking did not | Timing only the generation step is how people prove a saving that is not there |
That last row is the one that quietly decides whether any of this was worth doing. Generation got dramatically faster; reviewing, correcting and handing over did not. A workflow that produces a draft in ten seconds and then needs forty minutes of checking has not saved forty-nine minutes — and the only way to know is to time the whole thing, including the part where somebody reads it.
Four honest caveats
Two numbers are missing from the written version. The post numbers its entries 1, 3, 4, 5, 6 — two of the seven panels have no paragraph beneath the card. Nothing sinister, and the card itself has all seven, but it is worth knowing if you are reading the text and wondering why it does not add up.
Two panels are ecosystem facts dressed as recommendations. “Your business runs on Microsoft” and “work in Google tools” are true and useful, and they are also a description of lock-in. That is not an argument against either one — using the assistant that is already inside the software you pay for is usually the cheapest good decision available. It is an argument against reading those rows as a comparison. They are not competing with the other five; they are describing where you already live. We looked at how many of those in-app assistants people never open in Four Of These Ten Are Not Tools. They Are Switches.
“One job each” is the card’s claim, not a guarantee. Several of these seven overlap heavily in practice — two of them will both write you a first draft, and two will both answer a research question. The card’s discipline is the right one, and the way to hold yourself to it is to write the job as a sentence with an object in it, which is the method in It Promises No Overlap. There Are Three Overlaps In It.
Nothing here checks the output. Every one of the seven produces something confident. The research panel’s own instruction is the model to copy — access source references, check dates, and verify critical information yourself — and it applies to all seven, because the thing that made the draft cannot also be the thing that approves it: The Thing That Makes It Can Never Be The Thing That Grades It.
Questions people ask
Which two should I actually pick? One for the task that eats most of your week, and whichever assistant already lives inside the software your company pays for. That second one costs nothing extra and you are probably not using it.
What if my team already uses something different? Then your team has made the choice and the useful question is whether anyone is using it properly. An unused licence is the most expensive tool in any stack — the method for finding those is in Stop Paying Twice: The Overlap Audit.
Is it worth telling people the meeting is being transcribed? Yes, always, and the post says so itself. One sentence at the start, and where the notes will live. It is the cheapest trust you will ever buy.
How do I know the workflow actually saved time? Time it end to end, twice, on real work — including the review. If you cannot produce two timings, you have an impression rather than a saving.
One choice, one environment
Seven panels, seven jobs, and a footer that is wiser than most of the chart above it: pick two, not seven.
Just be clear about which two. One of them is a genuine decision, made by looking at the task that repeats every week and asking whether it has a clear “done”, steady inputs, a quality standard and an owner. The other was made for you — by the software your company bought, or by the people who will be sitting on the call.
Recognising which is which saves you the most expensive mistake on any tool chart: trying to choose something that was never yours to choose.
Learning a lot does not make anyone good. Doing a lot does.
If you want a deadline rather than a seventh saved chart, we run a free challenge — there is no barrier to entry, only a commitment fee that is refunded when you finish. It exists because a deadline runs the workflow once, end to end, with the review included.
Which repetitive task would you put through this first? Tell us in the comments under the original reel.
More on building a channel for a global audience at mmoyoutube.com.
