The one AI belief holding you back

Let go of this creativity-killing "limiting belief"

It’s one of the oldest AI tropes in the book.

Meet your new co-worker, AI. They are smart and patient.

(And never hungover.)

I originally discovered this metaphor in Ethan Mollick’s

One Useful Thing

newsletter (and have reused it in many of my AI Training Guides).

Digital skeuomorphism (aka the “folder icon” or “recycling bin” on your PC) has always played a role in helping noobs better understand emerging technology.

But the AI-as-a-tireless-intern meme is actually a dangerous self-limiting belief.

In this post I’m going to challenge you to really stretch your use cases. It’s going to feel uncomfortable.

But by the end, I promise you’ll look at AI in a completely different — and more expansive — lens.

---

Do you want to become an expert in ChatGPT in less than a week? This includes building CustomGPTs and lightweight agents. Upgrade to get access to ChatGPT Mastery for Professionals, available exclusively for our founding members.

---

Try this question instead

There are two ways to use AI.

One is to take existing workflows and extract every drop of efficiency out of them.

It’s the proverbial squeezing water out of a stone.

Remember when AI images were bad at doing hands?

And it works.

But it also leaves a lot of money on the table.

I recently hosted a training session on Agentic Workflows with Jacob Bank (the founder of Relay.app) and he proposed a new analogy.

Instead of asking “what can I optimize?” he recommended the following question:

If I had unlimited resources, who would I hire?¹

(The CEO of Shopify codified this mantra into their hiring policies.)

I was floored by this crazy talk — but ran with it.

Assemble your dream team

I immediately asked myself this question for my new business: AI Training and Strategy for Investment Managers.

Instead of thinking of traditional employees (i.e. Operations analyst or Comptroller) I narrowed the scope to specialists.

Here’s where I landed after a 10 minute brainstorm:

Let’s pick one that is generally relatable: Slide Deck Creator.

Can I automate away Slide Deck Creation?

I create a lot of slides these days. Especially since I’m doing a lot of prospecting.

A sample slide… ain’t it pretty?!?!

Here’s how I typically come up with new slide ideas:

• In the shower

• Reading a tweet, blog post or article

• After a prospecting call

Once the aforementioned “event” happens, AI usually steps in to flush out the idea (either via a CustomGPT, a mega-prompt or a Claude project).

Then, I take the Title, Bullet Points and Image (typically a MidJourney prompt) and manually create the slide in Google Slides.

Here’s what it looks like visually:

The “first mile” and “last mile” problem

As I continued the exercise, I came across a natural stumbling block.

The First Mile challenge

The first struggle happened immediately:

How do I get the “input” into AI?

This isn’t particularly obvious, especially for someone who doesn’t know how to code — I call this the First Mile Challenge.

Said differently:

How do I get my random shower idea into an AI system?

This is where the creative muscles start flexing.

What if, as soon as I got out of the shower (and still dripping wet), I did the following:

1. Recorded a voice note

2. It gets automatically transcribed (by Otter.ai)

3. The output then gets dropped into a pre-populated ChatGPT project/custom GPT (via Zapier, Relay or an api)

Let’s look at the first mile challenge on a specific tweet that inspires a slide deck:

1. Notice the tweet

2. I bookmark it

3. The text (including the link) gets scraped

4. This scraped text gets passed to an LLM (as described above)

The Last Mile challenge

A similar challenge occurs on the other side of the AI step.

Getting the output into its final state is also challenging.

In this case, I picked a particularly bad example since PowerPoints are quite visual and (currently) don’t integrate well with LLMs.

But imagine automatically getting a task in your Asana or ToDoist that says:

So the workflow would go:

Shower thought → Voice Note → AI Processing → Task to Create Slide

The magic is around the corner (and here for some)

As you think through your dream team I encourage you to list out how you will address the first and last mile challenges.

This exercise will really force you to truly understand your core tasks and how you add value.

Second, it will push you to understand the bottlenecks in your AI workflows.

(There are many!)

And finally, you’ll realize that many of the first and last mile challenges can already be solved with:

• Automated Workflow Tools (Zapier, Make, Relay, n8n)

• Vibe coding (Cursor, Replit, Lovable and Bolt)

• Coding with APIs

If you’re reading this, you probably will never write an API call.

But the technology on the first two is moving so fast — if you haven’t tried either, you will be blown away.

Good luck!

---

Do you want to become an expert in ChatGPT in less than a week? This includes building CustomGPTs and lightweight agents. Upgrade to get access to ChatGPT Mastery for Professionals, available exclusively for our founding members.

1. He actually said “Which coaches would I hire” — but I internalized it as “which employees…”