3 Easy, Unsexy AI Workflows that Will Save You 7 Hrs per Week

 

If you’re frustrated that you don’t know how to use agentic workflows to automate your work but you want to learn, this is for you.

It’s easy to see the viral Tweets and articles of people exclaiming how awesome their new AI system is and feel restless, stressed and left behind.

We know that this is the future of work, but how do we actually start using this?

Everyone will tell you you’re behind, but I’m here to tell you the opposite.

Claude Cowork has only been in existence for four months.

You’re still early.

And I’m going to give you 3 unsexy agentic workflows you can use to automate your work.

You don’t need to be technical. You just need to be willing to try it out.

Why unsexy automations?

Because usually the most useful AI use cases are unsexy.

Most of our time goes to a myriad of random tasks — writing follow-up emails, compiling reports, pulling numbers from dashboards. These random, unsexy and seemingly inconsequential tasks are our biggest opportunities.

Why listen to me?

I’m the head of product at a startup and 4 weeks ago, I automated 3 tasks that saved me 7 hours of busy work in a week.

I’ve been leading teams and building digital products for about 10 years, and most importantly, I have nothing to sell you.

So, with that out of the way, let’s get into it.

3 easy use cases that will save you 7 hours/week

Think of these as examples and jumping off points for you. Look for your analogous tasks.

Your imagination is most likely the limiting reagent to what’s possible, so let’s feed it some ideas.

1. Use case 1: Last week I left 4 back-to-back meetings and opened Gmail to find 6 draft emails waiting for me — each one based on what was actually discussed in the meeting. They were suspiciously accurate. I reviewed and sent all of them in about 12 minutes. That used to take me over an hour.
2. Use case 2: I used to spend forever writing weekly reports to my team. Before the automation, I had to DM people to remind them to send me their updates so I could compile them into a bigger report. Now I check Slack and see the draft report already written. I adjust a few sentences and hit send.
3. Use case 3: I need to send regular updates on KPIs and metrics. Usually this requires checking multiple analytics dashboards and cobbling together an analysis. Now my automation has created a draft that’s 90% of the way there. I update the remaining 10% and send to my team.

So how do I actually build these out? (OVERVIEW)

Here are the speedy instructions so you can get a sense for the process below, and then I’ll walk through a specific example below.

1. Download Claude desktop and get the $20/mo subscription. I recommend Claude because it can build automations for you, which ChatGPT can’t do. If you want to compare the quality of their outputs, give them both the same prompt and see for yourself.
2. Click the “Cowork” tab. Cowork can create automations for you, while Chat cannot.
3. Describe to Claude Cowork the automation you want to create and ask it the best way to do that.
4. Do what it says and use your brain matter to double check the logic.
5. KEY TIP: for emails, Slacks, or any actual output, clarify you only want a draft and that it shouldn’t send anything out. You should be the final sender.
6. Connect the MCPs (connections to your tools) you need.
7. Test the automation
8. Fix the automation

The instructions here are light because when you hit step #3, you’ll need to follow Claude’s personalized instructions for your situation.

If you get stuck on a step, show Claude where you got stuck and ask it how to get past it.

That’s going to be way more helpful to you than if I give you the step by step of how I created my own automations, because your setup will be different.

Deep dive example: how I created an automation that listens to my calls and drafts any emails I need to write

This is the most valuable automation I’ve created.

I realized that a lot of my work came out of meeting action items, and that if my AI buddy Claude could listen into those meetings, I’d have to provide a lot less context to it manually when I asked it to help me with those action items.

I use Granola to record and transcribe all my calls and my company uses Gmail, so the first thing I did was to connect the Granola and Gmail MCPs (Claude calls them “Connectors”) to Claude so that it could access both.

You can add “Connectors” or MCPs by clicking Customize < Connectors < + button.

Connecting the tools here was sooo easy, it was awesome. And the information and tasks that my automations can now access and execute are tenfold.

Write a prompt of what you want to do

I dug up my old prompt to see what I wrote. Seeing how it’s written, I probably used dictation (highly recommend).

My prompt:

Can you create a daily automation searching for any calls that I have in granola and then find any action items that I need to do and then draft up a summary of my action items and then actually complete them.

If an email needs to be written, then I want you to generate a draft and let me know that I need to review it. If a user story or linear ticket needs to be written, same thing.

Draft it and let me know what the outstanding questions are and let me know so that I can review it.

Use that microphone icon and speak your prompts to your AI rather than typing it all out.

You’ll provide it more context, and it’ll perform better as a result.

Here’s a screenshot of the actual prompt I dictated to claude

This prompt isn’t winning any Pulitzer awards, but I lean towards making the initial prompts simple and giving it context where it asks.

I only spend time on the prompt when it’s a bigger or more complex task. Otherwise, I figure it out as I go.

Claude will ask you for more detail when needed, and you’ll realize where you need to provide more detail as you build it.

(For instance, I initially didn’t specify how often this automation should run. I later realized that I needed the automation to run every 30 minutes to check for any meetings that I just had.)

Be patient setting it up.

I had a lot of back and forth with this automation and I’m going to be honest with you, it had me banging my head against my Mac after the 7th repeat error.

But don’t give up.

We’re early to the tech, and there are pros and cons to that.

Think of it like training a new employee, and don’t give up just because it takes longer to set up than it would take to do it manually the first time.

The first couple times, it takes longer.

And then you never have to do it again.

The moment your automation works is a beauty like no other. Screenshot of the emails and user stories that my automation wrote for me right after a meeting I had.

How to troubleshoot when you’re not technical

The biggest issue I ran into was that the automation didn’t run when it was supposed to. It would just… not trigger.

I told Claude “hey, the automation was supposed to run at 10:30am but it didn’t — what happened and how do we prevent this?”

We ended up making a second automation to monitor the first one and fix it when needed. Automation babysitting another automation.

Welcome to the future.

Another issue: the email drafts were pulling action items from the wrong meeting. I told Claude which email was wrong, what meeting it should have pulled from, and asked it to fix the matching logic. It adjusted the automation and I haven’t had the issue since.

The pattern is always the same: describe what went wrong, what you expected, and ask Claude to fix it and prevent it from happening again.

In the beginning, you need to double check the automation’s work several times. That’s normal. You’re training it.

The more specific your feedback, the faster it gets good.

What does life look like after these 3 automations?

First of all, it’s extremely empowering to create your own personalized automations.

When you create your own automated systems, you are literally multiplying your impact and effectiveness in a way that was never possible before.

You become way more valuable as an employee, way more effective as a business owner, and you’re starting to learn how to coexist in the AI age.

Second of all, as the title suggests, I eliminate at least 7 hours of menial, boring, repetitive tasks a week through these 3 automations.

I don’t have to follow up with teammates about weekly reporting, I’m not writing tactical emails from scratch, and I’m not cobbling together an analysis across multiple dashboards.

I’m simply reviewing and editing content that has already been created for me, which is a lot faster and easier.

These 3 automations save me at least 7 hours of menial, boring, repetitive work a week. That’s almost a full work day I get back to spend on things that actually require my brain.

Now it’s your turn.

Download the Claude desktop app, click the Cowork tab, and try this prompt:

“I want to automate [your repetitive task]. Here’s how I currently do it: [your steps]. How can we automate this?”

Start with whatever task annoys you the most.

That’s your best first automation.

Ally Mexicotte

https://tinyurl.com/3bxunavu