Most AI productivity advice is a list of 40 tools. Installing 40 tools is not productivity. This is the short version: the handful of uses that reliably save time, and the ones that quietly cost it.
Where the time actually comes back
Editing text you already wrote. Shortening, clarifying, changing tone. The model is working from your material, so the failure modes are small and correction is fast. Highest-confidence use there is.
Reading things you would otherwise skim. Long threads, transcripts, contracts, research. "What does this say and what do I need to decide" is a genuinely different task from writing, and models are good at it.
First drafts of structured documents. Job ads, SOPs, briefs, proposals. You will rewrite half. Half is still a large saving from a blank page.
Extraction. Pulling action items, dates, names and figures out of messy text into a table. Tedious, well-specified, reliable.
Explaining unfamiliar material. A domain you are new to, a codebase you inherited, a document written by someone else's lawyer.
Where it quietly costs time
Tasks you could do in under two minutes. Prompting, reading, correcting is often slower than just doing it.
Anything you then have to verify carefully. If checking takes as long as writing, you saved nothing — and this applies to anything with figures, dates or citations.
Tool sprawl. Every new tool has a setup cost, a learning cost and a switching cost. Three tools you use daily beat fifteen you evaluated once.
Automation built too early. Wiring something into a workflow before you know whether the manual version helps is how people end up maintaining a pipeline that produces work they then redo.
"Let me just ask the AI" as procrastination. Recognisable once you notice it.
The two-minute rule
Before reaching for AI, ask: would this take me under two minutes?
If yes, do it. The overhead of prompting, reading and correcting usually exceeds the task. AI pays off on things that take fifteen minutes and up, and on things you would otherwise avoid entirely.
A realistic stack
You need fewer tools than the lists suggest:
- One general assistant for writing, reading, extracting and explaining. This covers most of the value on this page.
- Something in your editor if you write code.
- Whatever your calendar and notes app already added, if it is any good.
That is it. The productivity gain comes from using one thing constantly rather than sampling many.
Habits that separate people who get value from people who don't
Paste your material. The most common reason for a useless answer is a question with no context attached.
Follow up rather than restarting. "Shorter." "Too formal." "Keep paragraph two." Starting a new chat discards everything it learned about your task.
Batch similar work. Ten emails in one session beats ten scattered interruptions — same reason batching anything works.
Know what it cannot do. Arithmetic, counting, current events without web access, and knowing when it is wrong. Stop trying to fix those with better prompts.
Keep one place for prompts that worked. A note file is enough. You will reuse five of them constantly.
Measuring whether it is helping
The honest measure is time saved minus time spent correcting, on one task, over two weeks.
Most people never do this and rely on the feeling of productivity, which AI is unusually good at producing. Being busy in a chat window is not the same as finishing work.
If the number is not clearly positive on a given task, stop using it for that task. It will still be positive on others.
Common questions
What are the best AI tools for productivity? One general assistant covers most of it. Add an editor integration if you code. More tools than that usually costs more than it saves.
Does AI actually make you more productive? On editing, reading and extraction, reliably. On short tasks and anything needing careful verification, often not.
How much time can AI realistically save? Enough to pay for a $20/month subscription within days if you write or read a lot. Claims beyond that are usually measuring enthusiasm.
What's the biggest mistake people make? Collecting tools instead of building a habit with one, and automating before proving the manual version helps.
Should I automate my workflow with AI? Only after using it manually for that task for a few weeks and confirming it saves time.
Which model is best for productivity work? Claude for anything a person reads, Gemini for anything current, and a small fast model for high-volume mechanical tasks.