This article walks through how to find your 80/20 workflows: the two or three areas where AI can genuinely change how your week feels, not just add another layer of complexity.
The trap of doing everything
The easiest way to fail with AI is to approach it as a general productivity upgrade. You try to use it for email, content, research, strategy, planning, customer support, and internal documentation all at once. You spread your attention across a dozen use cases, none of them deep enough to produce a reliable result. After a few weeks, you have ten partially configured workflows, none of which work consistently, and you start to feel like AI is more trouble than it is worth.
This is not a personal failing. It is the natural result of an industry that tells you to “use AI everywhere.” The advice sounds helpful but it is actually counterproductive, because it pushes you toward breadth instead of depth. Depth is what produces time savings. A workflow you have used and refined three times is faster than a workflow you try for the first time every time.
The fix is simple in concept but uncomfortable in practice: pick fewer things to do with AI, not more. The discipline is in deciding what not to use AI for.
How to find your 80/20 workflows
Set aside 20 minutes. Go through your week task by task and ask the following three questions about each one. Write down the answers. The patterns will emerge quickly.
What do I dread most each week?
This is the most reliable signal. The tasks you procrastinate on, the ones you put off until the last minute, the ones that drain disproportionate mental energy. Dread is a better indicator than time because it captures both the clock cost and the emotional cost. A 15-minute task you hate feels worse than a 30-minute task you enjoy.
For most knowledge workers, the tasks that top this list involve either writing from scratch, processing large amounts of information, or making decisions with incomplete data. All three are areas where AI genuinely helps when applied correctly.
What do I do that is repetitive but still needs my oversight?
These are the tasks that feel like they should be automatable but are not quite rote enough to hand off entirely. Drafting routine responses to common client questions. Creating weekly status updates from project notes. Summarising meeting recordings. Preparing agenda outlines for recurring calls.
AI excels here because the task follows a pattern but still requires human judgement on the final output. The AI handles the pattern part (the first 80% of the work), and you handle the judgement part (the final 20%). This split is where the biggest time savings live.
Where do I lose time to context switching?
Context switching is the hidden productivity killer that most people underestimate. Every time you stop one task to look up information for another, you lose not just the lookup time but the mental momentum of the original task. If you spend 10 minutes hunting for a file or re-reading a thread before you can reply to an email, that 10 minutes cost you another 10 minutes of lost focus on whatever you were doing before.
AI tools that reduce context switching by bringing information to you, rather than requiring you to go find it, are disproportionately valuable. A tool that surfaces a relevant past project or drafts a reply based on your existing notes saves more time than the clock suggests, because it protects your focus.
“I picked one workflow (competitor research) and spent a month getting good at it before I tried anything else. That single workflow saves me about four hours a month. If I had tried to do everything at once, I would have got nothing.”
From the Insight Division Labs free course
Three common 80/20 patterns
These three patterns show up consistently across the knowledge workers and business owners we work with. They are not the only possible workflows, but they are a useful starting point if your self-audit does not immediately reveal your own.
Content preparation
Research into outline. Rough notes into first draft. Interview transcript into blog post. The common thread is taking something unstructured or semi-structured and turning it into something more polished. AI is extremely good at this, and the time savings compound because each piece of content can be repurposed into multiple formats.
The trap to avoid here is treating AI output as finished. The workflow is: generate a draft, edit it significantly, then publish. The AI saves you the blank-page struggle and the structural thinking. You still do the refinement and the fact-checking.
Communication compression
Long email thread into summary. Meeting recording into action items. Client call into follow-up notes. This pattern takes something that takes a long time to consume and produces something that takes a short time to act on. The value is not just the time saved reading. It is that you actually read the summary, whereas you might have ignored the original thread.
Decision preparation
Competitor data into comparison table. Spreadsheet of options into ranked shortlist. Customer feedback into priority list. AI cannot make the decision for you, but it can do the tedious organisational work that precedes good decisions. The time saving here comes from reducing the activation energy required to make a well-informed choice.
How to test a workflow before you commit to it
Once you have identified a candidate workflow, test it before you build anything elaborate. Run it manually three times. Same input, same process, same output format. If it does not work reliably after three attempts, do not invest in automation or tooling. Either the workflow is not well-defined enough, or AI is not the right solution for this particular task.
If it does work, run it for two weeks. Keep a simple log: did this save me time, or did it just feel like progress? The feeling of progress is deceptive. If you spend 20 minutes generating something and 15 minutes editing it, but the task used to take you 25 minutes from scratch, you have not saved anything. You have added a step. A real 80/20 workflow should save you at least 30% of the original time, or it is not worth maintaining.
Start with one
The single best piece of advice for anyone trying to get value from AI: pick one workflow and make it work before you start another. The temptation to expand will come quickly. Resist it. A single reliable workflow that saves you two hours a week is infinitely better than five half-working workflows that save you nothing.
A good place to start is the free Insight Division Labs course. It covers three practical AI workflows that map directly to the patterns above. Not a survey of options. Just a starting point you can actually use.