Every week brings a new AI tool that promises to change how you work. Most of them do change something: they add a place where you now have to go, prompt, review and correct.
That is not a reason to avoid them. It is a reason to be clear about what you are actually buying.
The distinction that matters
Tools that remove a task. You stop doing something. Transcription is the clearest example — you no longer take notes during a call, and you no longer type them up afterwards. The time saved is real and it is permanent.
Tools that change how you do a task. You still do the work, but differently. Writing with an AI assistant does not remove the writing — it changes it into prompting, reading, judging and correcting. Sometimes faster, sometimes not, and almost always worse if you stop reading what comes out.
Tools that add a task. This is the category nobody admits to. A tool that generates content you then have to check, edit and fit into a process it does not understand has added work, not removed it.
Before you subscribe to anything, decide which of the three you are looking at.
Where AI genuinely wins today
Anything with a transcript. Meetings, calls, interviews, videos. The input is audio, the output is text, and the machine is better at it than you. This is the least glamorous and most reliable category.
First drafts you were going to rewrite anyway. Not the final text — the blank page. If your process already involves drafting and heavy editing, the draft is the cheap part to automate.
Structured extraction. Pulling fields out of documents, classifying messages, summarising long threads. Repetitive, tolerant of small errors, and verifiable at a glance.
Translation and reformulation. Mature, fast, and good enough that the remaining work is judgement rather than labour.
Where it still costs more than it saves
Anything requiring your specific context. An AI that does not know your clients, your history or your constraints will produce something plausible and generic. Correcting plausible-but-wrong takes longer than writing it yourself.
Work where being wrong is expensive. Legal, financial, medical, contractual. Verification is the whole job, and the AI does not reduce it.
Anything you will publish unedited. This is not a technology problem. Content nobody checked reads as content nobody checked, and the cost lands on your credibility rather than your calendar.
The question that decides it
Not "is this tool good" but: after using it, what exactly do I no longer do?
If you cannot name the task that disappears, the tool has probably not saved you anything — it has moved the work somewhere you have not measured yet.
Who this does not suit
If you are looking for a tool to produce volume without attention, the approach here will frustrate you. It exists, it works, and the output is visible as such.
If your work is entirely relational — negotiation, management, care — the gains are marginal and sit around the edges: preparation, notes, follow-up. Not the work itself.
If you are a solo operator with limited time, the transcription and extraction categories are worth more to you than anything generative. Start there.
What we do not know yet
We do not publish prices or feature comparisons. This category changes monthly, tools merge and disappear, and a table written today would be wrong by the time you read it.
What does not change is the distinction above: remove a task, change a task, or add one. Apply it to whatever launches next week.