Hi friend! How’s it going?? 😊

So more people are using AI agents to automate their workflows, which is great…

But I’m also seeing a lot of people struggling because they fall into the same three mistakes.

These mistakes either stop the agent from ever working or cause major problems down the line.

So today I wanted help you by sharing them with you!

Let’s start with the first one, which is… 🥁🥁🥁

1. Over-Engineering

Honestly, as powerful as I think agents are, there are some workflows that just don’t need them.

A really common mistake anyone with the ‘engineer-brain’ tends to make is build a really complex agentic system for a task that could probably be done faster manually or with a simple no-code automation.

On to the next…

2. Bad Prompting

This is a big issue most people don’t even realise is a problem.

Bad prompting can be best explained by cooking. Let me explain..

If you order a chef to follow a bad recipe, the food will be terrible no matter how great the chef is (unless they rebel, which thankfully AI doesn’t do….yet.)

In the same way, you can have the best agentic system design in the world, but if your agent prompts aren’t done right, the whole thing will fall apart. Not good :/

Now for the final one…

3. System Design

This is the last and biggest mistake I see even AI influencers make.

You can spot it when you see these huge complex AI agent systems which look ‘cool’ but are usually inefficient, fragile and difficult to debug.

Essentially, it’s people not really understanding the core agentic workflows.

They don’t take time to figure out the right underlying architecture to use.

So instead of following sound engineering principles, they just build whatever comes to mind.

As you can imagine, this becomes problematic if the agentic system is being built for a real business managing real tasks and people.

Now on that note, I have two exciting announcements.

First, I’ll be covering the solution to some of this stuff in tomorrow’s workshop:

Building AI Agents: A Step-by-Step Guide!

It’s from 8 to 9pm EST.

Second, next Thursday I’m launching the AI Agent Bootcamp.

This is a hands-on practical program where you build 4 agent projects, starting simple and ending with advanced multi-agent systems.

There will be only 100 spots because we can only give so many people the needed attention.

But there are already 1400+ people on the waitlist, so I’ll open doors to the waitlist to give them a chance to secure a spot first.

If you’d also like a better chance at securing a spot, you can join the waitlist by clicking here:

Join the Waitlist

Thanks friend,

Tina