What to Automate with AI (And How to Stop Wasting Time on the Wrong Tasks)

Not sure what tasks to automate with AI — or where to even start? You’re not alone. They’ve heard the hype. They know AI can save time. But when they sit down to actually do something, they hit a wall — not because the tools are hard, but because they don’t know which tasks to automate with AI in the first place. The answer isn’t to automate everything — it’s to automate the right things first. This guide gives you a simple framework to figure out exactly what works, so you stop spinning your wheels and start getting time back.

Why Most People Automate the Wrong Things First

There’s a pattern that shows up over and over when people start with AI automation. They get excited, pick a complex workflow they’ve always hated, spend hours setting it up — and either it doesn’t work the way they hoped, or it only runs twice a month and wasn’t worth the effort.

I saw this firsthand in my own work as a software engineer. After COVID shifted everything remote, nearly every task moved online — writing programs for operators, tracking down bugs, patching and handing off updated builds. The structure of each cycle was simple enough: pick up where yesterday left off, fix what broke, and pass it on. But the part that actually consumed the most time wasn’t the writing or the fixing — it was the manual review work. Reading through every output, visually confirming every change, making sure nothing slipped through before handoff.

If I had today’s AI tools back then, I’d have automated exactly that review layer — set up a step that scanned each build output against a checklist, flagged anything outside expected parameters, and only surfaced the items that actually needed my eyes. The pattern was clear, the frequency was daily, and the cost of a missed error was real. It’s a textbook automation candidate. The problem wasn’t the work itself — it was not having a framework to recognize that.

That framework is what most people are missing when they start. The most common mistakes look like this:

• Starting with the tool, not the task. You sign up for Make or Zapier because you’ve heard great things, then try to reverse-engineer what to build. This almost always leads to over-engineered automations that solve problems you don’t actually have.
• Picking tasks that happen too rarely. If you only do something once a month, the time you spend setting up the automation will exceed the time you save — at least for a long while.
• Trying to automate judgment calls. Tasks that need real context, nuance, or relationship awareness don’t automate well. The output ends up generic, and you spend more time fixing it than you would have spent just doing it yourself.
• Going too big too fast. Trying to automate an entire workflow end-to-end before you’ve tested the individual steps is a reliable path to frustration.

The fix isn’t a better tool. It’s a better selection process — and that starts with three questions.

What Tasks to Automate with AI: How to Know

Before you automate anything, run it through these three filters. If a task clears all three, it’s a strong candidate. If it fails one, it’s usually not worth your time yet.

1. Do you do it at least once a week?

Frequency is the single biggest factor in whether an automation pays off. A task you do three times a week has 150+ repetitions per year — even a small time saving adds up fast. A task you do once a month has 12 repetitions. The setup cost rarely justifies itself.

In my case, the review cycle I described above wasn’t weekly — it was every single day. That kind of daily recurrence is exactly what makes a task worth the setup time. A good rule of thumb: if you can’t picture doing this task again before next week, move on.

2. Does it take more than 15 minutes each time?

Low-frequency tasks can still be worth automating if they eat significant time each time you do them. The sweet spot is tasks that are both frequent and time-consuming — but either one can work if the other is extreme enough.

My review work cleared this bar easily — it regularly consumed more than half the workday. Not because any single check was complicated, but because there were so many of them, and each one had to be done carefully. Think about email triage, meeting follow-ups, or formatting recurring reports. Each one might not feel like a big deal in isolation, but they compound across a week into hours of drained attention.

3. Does it follow a pattern you could explain to someone else?

This is the most useful filter of all. If you could write down the steps and hand them to a new hire who’d never done the task before — and they’d get it right — it’s automatable. If the task requires reading the room, using judgment that shifts based on context, or applying real expertise, it’s not a good automation candidate yet.

My work passed this filter too — but it also highlights an important nuance. Because the output was handed off to other people to use, quality and accuracy weren’t optional. A double-check before every handoff was non-negotiable. That’s exactly where AI automation adds the most value: not replacing the check, but running it consistently and flagging anything that falls outside the expected pattern, so the human review becomes a quick confirmation rather than a full scan.

Concrete examples that pass this filter: sorting emails by type, pulling data from a form into a spreadsheet, generating a first-draft summary from meeting notes. Examples that fail it: deciding how to respond to a difficult client, writing content that needs your voice, making a judgment call on a complex situation.

Once you’ve run your current tasks through these filters, the next step is knowing where to look for the highest-return candidates.

Where to Look: 5 Task Categories Worth Automating with AI

Once you understand the three filters, it helps to know where to look. These five categories consistently produce the best automation wins for individuals and small teams.

1. Email and inbox management

The average professional spends roughly 28% of the workweek — close to two hours a day — just managing email. A significant portion of that is pure triage: figuring out what needs a response, what’s FYI, and what can be ignored. AI handles this well because the rules are learnable: flag anything from this domain, sort newsletters into a folder, draft a reply template for this type of request. According to the McKinsey Global Institute’s “The Social Economy” report, interaction workers spend 28% of their workweek managing email — see the full report for methodology. (Published July 2012 — email overhead has only grown since, but verify any related stats at source.)

Good starting points: Gmail’s built-in filter rules combined with a ChatGPT prompt for routine reply drafts — no-code and set up in under 30 minutes. Or connect Gmail to Zapier with an AI step to auto-label and route incoming messages by sender or subject keyword.

2. Meeting notes and follow-ups

Recording a call, getting a transcript, pulling out action items, and sending a summary — this is a near-perfect automation candidate. It’s repetitive, time-consuming, rule-based, and the output is easy to verify. Tools like Otter.ai, Fireflies, or Fathom handle most of it automatically once set up — join the call, and a structured summary lands in your inbox 10 minutes after you hang up.

The time savings here often surprise people. It’s not just the 20 minutes of note-taking — it’s the context-switching cost of going back to your work afterward.

3. Content repurposing and formatting

If you create any kind of content — reports, blog posts, social updates, newsletters — there’s almost always a repurposing step that eats time. Taking one piece and adapting it for three platforms used to mean starting from scratch each time. With AI, you write the core piece once and let it handle the formatting variations.

This works especially well when the output format is consistent: a LinkedIn post always looks a certain way, an email newsletter has a template, a weekly report follows the same structure. A simple ChatGPT prompt — “reformat this blog intro as a 3-sentence LinkedIn post” — runs automatically via Zapier or Make every time you publish.

4. Research and information gathering

Regular research tasks — checking for news on a topic, compiling competitor updates, pulling stats for a report — are excellent automation targets. They’re repetitive, time-consuming, and the pattern is clear: find X, summarize it, put it here.

A practical setup: use Make to trigger a GPT step on a daily schedule — the prompt specifies the topic, and the output gets sent to your inbox or dropped into a Notion page automatically. You’re reviewing a digest instead of hunting for information. A starting prompt looks like this: “Summarize the top 3 developments about [topic] from the last 24 hours in 3 bullet points, one sentence each.” The basic workflow takes under an hour to configure and runs without touching it afterward.

5. Data entry and file organization

Moving data from one place to another, renaming files according to a convention, logging information from forms into spreadsheets — these are the original automation use cases, and AI makes them significantly easier to set up without coding. If you find yourself copy-pasting the same type of information between tools on a regular basis, that’s a strong signal.

A practical example: form submission arrives → Zapier extracts the key fields → AI step formats the data → row added to Google Sheets automatically. No coding needed, set this up in under 30 minutes. Another common setup: new invoice attachment arrives by email → AI step extracts the vendor name and date → file renamed and moved to the correct folder in Dropbox or Drive automatically. If you want to go deeper on connecting tools for these kinds of workflows, How to Connect Your Apps with AI Automation Tools (No Coding Needed) walks through the practical setup.

Now that you know which categories deliver the best results, the next question is which one to start with — because picking the right first automation matters more than most people expect.

How to Pick Your First One

You’ve run your tasks through the three questions, and you’ve probably identified more than one candidate. Now you need to pick the first one to actually build — and the right choice here matters more than most people realize.

Your first automation should be:

• Something you do at least 3x per week. The feedback loop is tighter — you’ll notice the time savings immediately and spot problems faster.
• Simple enough to set up in under an hour. Your goal right now isn’t to build the perfect system — it’s to build confidence and momentum. A working simple automation beats a stalled complex one every time.
• Low stakes if it goes wrong. Email draft templates, meeting summaries, file naming — these are forgiving. If the AI gets it 80% right, you edit the other 20%. Don’t start with something where an error has real consequences.

If two tasks both look like strong candidates, use this tiebreaker: pick the one you most dread doing manually. The emotional relief of not having to do that task anymore is what turns a working automation into a habit. That motivation keeps you from reverting to old patterns when the setup feels fiddly.

Here’s what this looks like in practice. Say you’re a freelancer who has three client check-in calls a week and manually writes follow-up emails after each one. That’s a daily task, it takes 20 minutes each time, and the structure is always the same: recap what was discussed, list the next steps, confirm the deadline. That’s a perfect first automation — connect your call recorder to a summary tool, then feed the summary into a GPT prompt that generates the follow-up email draft. You review and send. The whole setup takes about 45 minutes and runs automatically after that.

For most people, email or meeting notes is the right first choice. Both are high-frequency, meaningfully time-consuming, and easy to verify. You’ll know within a week whether it’s working. Once that’s running smoothly, you can expand. The goal is to build a habit of noticing automatable tasks — not to overhaul your entire workflow at once. For a step-by-step look at the broader process, How to Automate Your Workday with AI covers the full arc from identification to implementation.

Just as important as knowing what to automate is knowing what to leave alone — at least for now.

What’s Not Worth Automating (Yet)

Tasks you haven’t done manually enough to understand. If you can’t explain the pattern clearly, you can’t automate it reliably. The better approach is to do the task manually a few more times, notice what stays consistent, and then build the automation. Skipping this step is one of the most common reasons early automations fail.

Tasks that require real judgment. Responding to a difficult client email, making a hiring decision, giving feedback on someone’s work — these rely on context that changes every time. AI can assist here, but it can’t replace the judgment, and trying to fully automate it usually creates more cleanup work than it saves.

One-off tasks. If you’re setting up an automation for something you’ve never done before and might never do again, stop. The setup time will never be recovered. Do it manually and move on.

Anything that needs your voice or relationship. A personalized message to a longtime client, a nuanced negotiation, creative work that reflects your actual perspective — automation makes these worse, not better. automate these and you risk making the interaction feel generic, and that’s exactly what kills the relationship.

Knowing what not to automate is what keeps your time-saving efforts from backfiring. For a broader look at how automation fits into a daily workflow, How to Automate Your Daily Schedule with AI (5 Tools That Actually Work) is a useful companion read.

The framework here is simple and it works: filter your tasks, start small, and build from there. The people who get the most out of AI automation aren’t the ones with the most sophisticated setups — they’re the ones who picked the right first task, saw it work, and kept going.

Frequently Asked Questions

✍️ We test and use AI tools and automation workflows in our own day-to-day work — no jargon, just honest guidance based on real experience. About DailyTechEdge →