AI agents won't save you (this will)

Last week, I tied my brain in knots trying to work out an AI Agent solution for a ‘problem’ I was having.

Turns out I didn’t need an AI Agent at all…

(and here’s why you probably don’t need one either)

Rabbit Hole

I wouldn’t say I ‘drown’ in e-mails, but they definitely take up a lot of brain space.

Combine this with the other 397 tools I used on a daily basis…

Gmail. Slack. HubSpot. Asana. Notion. Readwise.

… and it adds up to a lot of systems fatigue.

“There must be a way to work smarter” I thought to myself.

“I’m an AI dude, why aren’t I taking an AI first mentality to this?” was the voice that ended up prodding me for several days.

So I did what I normally do.

I went down the rabbit hole.

I started mapping out the logic - An agent that would read my emails, categorize them, create tasks in the right tools, and then maybe notify me in Slack for urgent stuff.

I even started researching apps that already had this stuff part built.

Then I stopped.

And asked myself a question that saved me days of wasted effort:

"Is this actually a problem?"

Turns out it wasn’t.

The real problem wasn't with the emails, or the apps.

It was with me.

Here's the thing nobody tells you about AI agents:

You can't deploy a slick AI agent if the human deploying it is buggy as hell.

The Agent Trap

Here’s what you’re probably finding right now.

You're overwhelmed, with an overflowing inbox, and tasks scattered across 5 different apps. Not to mention the migraine inducing brain fog.

And then you see it.

Another LinkedIn post about AI agents. Another founder showing off their "autonomous workflow" that "saves 20 hours a week."

The FOMO kicks in.

"I need that."

So you start researching. Zapier. Make. n8n. Relay. You watch YouTube tutorials. You join Discord servers. You bookmark 47 articles about "agentic workflows."

For me, I’d seen the new Agent capability in Make.com (the platform I use for most of my workflows) and couldn’t figure out why I hadn’t started using it yet.

And this is EXACTLY the trap.

This feels productive.

You're learning. You're planning. You're "investing in systems."

But you're not doing the work that makes money.

You're just procrastinating.

I see this all the time.

Solo operators spending weeks building elaborate AI workflows to "automate their content" or “organise their e-mails” but they’ve not taken any REAL action towards building their business.

And the worst part?

The AI marketing machine knows this. They're selling you the dream of “10xing your productivity” when the best thing for you is to work out what (actually) matters for your business.

And here’s what most people don’t mention:

Most "AI Agents" aren't even agents. They're just AI-enabled workflows.

AI Workflows

• You specify the steps.

• AI performs an action.

• This is just Zapier + ChatGPT.

AI Agent

• You set the goal

• AI works out the steps, and performs the actions

• This is ChatGPT Agent Mode and Claude for Computer Use (still not that great yet)

But even if you had a real AI agent...

If your thinking isn’t clear, your process is broken, and you can't follow your own rules, the agent won't save you.

It'll just take up more brain space. Waste more of your time. And stop you from taking action.

The Human Debug

So what's the solution?

It's not sexy. But it works.

I call it “The Human Debug”.

Here's the concept:

Before you build any AI agent, workflow, or automation, you debug the human running the system.

You.

After the 3rd day of me planning my ultimate AI Agent I realised I’d become to close to it.

I’d convinced myself that there was a smart way to build a solution to a problem that I was sure existed.

But you know what they say:

“The smarter you are, the more likely you are to believe your own bullsh**”

So, I gave myself some time to detach.

And when I came back, I asked myself a different question:

"What does good actually look like here?"

Not

"What steps do I need to automate?"

But

"What outcome do I actually want?"

And I came to 2 conclusions:

1. I needed to follow my own rules (I’d become lazy with my own process) and was trying to find a lazy approach to a micro-problem.

2. If I was going to use an AI Agent (or an AI enabled workflow) it would be better served solving a BIG problem.

One of my bigger headaches?

Bookkeeping.

Right now, I've got an automated workflow that grabs invoices from emails and dumps them into SharePoint. And it works OK.

But for the invoices I don't get via email, I still have to log into websites, find the billing section, manually download PDFs etc.

It doesn’t take a massive amount of time.

But I usually do it last thing once I’ve completed my other work. Which:

• Encroaches on family time

• Fatigues my brain when I’m already in a tired state

• Increases chances of error, and additional work for me when the accountants come back saying “You’ve missed something”…

This is a much better use case (and one I’ll help you solve in another newsletter).

But what’s the biggest difference with the above vs what I’d done previously?

I debugged myself first. Before thinking about what to build.

And if I’d not had that conversation with myself, I’d still be building something completely useless. Instead of writing this newsletter to you!

How to Run The Human Debug

Before you jump straight to the pricing pages of that AI Agent tool you just saw another influencer save 37 hours from and make $976,678 a month ‘whilst they sleep’. Do this:

Step 1: Pause for 24 Hours

The next time you feel overwhelmed and start hunting for an AI solution…

Stop.

Close your browser tabs and step away.

Give yourself 24 hours to detach.

You need to give yourself a pattern interrupt.

When you do this, you break the FOMO-driven reflex to "find a tool" when the real problem might be something else completely.

You detach from the noise, and remain objective about what solutions will / won’t help you move forward.

Step 2: Ask What Good Looks Like

When you come back, ask yourself this question:

"What outcome do I actually want?"

Just the outcome.

Examples:

• Bad: "I need an AI agent for content"

• Good: "Three high-quality posts per week without burnout"

• Bad: "I need to systematise my sales process"

• Good: "Five qualified referrals from my network"

You can then ask this.

"What's actually stopping me from achieving this right now?"

Nine times out of ten, it's not a missing tool.

It's because you’re:

A) Not following a process you’ve already bit

B) Scared of actually doing the work

This is why in my afternoon journal, I have a step for Actions.

These aren’t:

• I got all of my tasks done today

• I had a really good conversation with ‘x’

These are:

• I delivered that client work ahead of time (and sent the invoice)

• I reached out to 5 connections (and booked some meetings)

Colby Kultgen says it best.

We need to have a “Bias for action”

Not a “Bias for automation”

Fix the human bug first.

Step 3: Decide on Your Tool

If you really do have a problem.

If the problem is stopping you from taking action.

THEN start thinking about a tool.

Here're your options:

Option 1: Optimize Your Current Process

Can you achieve the outcome by fixing your behaviour improving your existing system?

If yes, do this first.

Just follow your own rules.

Or - look for additional capability in your existing tool that could help.

Option 2: Build an AI Workflow

If your work can follow IF this THEN that steps (e.g when I receive an e-mail with an attachment, then download the attachment to this folder)

Use a tool that’s AI enabled.

Using a workflow that’s something like:

• Gmail - Watch for e-mails with attachment

• ChatGPT / Claude - If the attachment is an invoice from supplier A, save in SharePoint folder B

• SharePoint - Save file.

Tools: Zapier, Make, Relay

Option 3: Deploy an AI Agent

Is the work complex, and requires an element of problem solving that can’t be followed by standard steps.

Use an AI Agent.

The workflow could be something like:

• Claude for Computer Use - Prompt “Using Google Chrome, go to these URL’s [Enter links] and download the latest invoices”

These are still in their infancy, so I wouldn’t spend 6 days trying to set them up now, to find that in 6 months when they’re infinitely better and they can be set up in 6 minutes.

The reality is, that the fastest way to free up time isn't building a fancy AI agent.

It's debugging the human who keeps opening their inbox at 9pm (and spends more time thinking about how to do stuff better, then doing the thing at all).

The Final Word

AI agents aren't going to save you.

Not because they don't work (even though they still kind of don’t).

It’s because:

• Most of us are trying to automate instead of eliminate.

• Most of us struggle more with doing than thinking.

• Most of us get FOMO waaaay to easily.

We're buggy operators trying to build on rocky foundations.

And it doesn't work.

So before you spend another hour researching the "perfect AI workflow," run ‘The Human Debug’:

1. Pause for 24 hours - Break the FOMO cycle and get clarity

2. Ask what good looks like - Define outcomes, not processes

3. Decide on your tool - Optimize yourself first, automate second, deploy agents last

The harsh truth is this:

Most of your "time-saving" fantasies are just procrastination with better branding.

Fix your habits. Follow your own rules. Only then should you build the machine.

Because the fastest way to free up time isn't in the next AI tool.

It's in the discipline to stop looking for one.

Until next time.

Adam

P.S - If you want to see the workflows I’ve automated to save myself 1 day a week and (actually) become more productive, you can find my 2-Hour Automator course here for free.

P.P.S - Know a friend who could really use this? Please share, and drop me a reply with who you shared it with. I love making new friends :)