As AI makes information and digital output easier to produce, the opportunity shifts toward what remains harder to automate—real-world skills, human judgment, personal trust, and getting the job actually done.

September 2026

Every week there's another headline:

AI can write this.

AI can design that.

AI can build websites.

AI can edit videos.

AI can answer questions.

And naturally, people start asking:

But maybe that's the wrong question.

Because trying to predict which job AI will never touch is almost impossible.

A better question is:

That's where things get interesting.

A machine can generate an image.

But can it stand at someone's wedding and paint the couple?

AI can explain how to fix drywall.

But can it come to your house and repair the hole?

AI can generate a workout plan.

But does that mean people stop hiring coaches?

AI can generate thousands of videos.

But does that eliminate the value of someone who understands the audience, brand, story, and desired result?

Probably not.

So instead of building your side hustle against AI, let's look for opportunities where technology can become your assistant...

while you remain the reason somebody buys.

The source divides many of these ideas into skill-based services, care/community work, and digital opportunities with a human element.

I'd simplify that further.

There are five things I'd look for when evaluating an AI-resistant side hustle:

Does somebody physically need to be there?

Does the work require physical execution?

Does the customer care who performs the work?

Does judgment and personal interpretation matter?

Does the customer's experience depend partly on human connection?

The more boxes you check...

the less your business is simply selling information.

👉 Don't just sell what you know. Sell what happens because you're there.

Let's start with the obvious.

AI cannot currently:

paint your wall...

walk your dog...

photograph your wedding...

do your nails...

repair your cabinet...

or personalize a gift by hand at your corporate event.

Information can tell somebody how to do these things.

But information isn't execution.

That's an important distinction.

Calligraphy sounds like something technology should have destroyed years ago.

We have:

fonts...

printers...

design software...

AI image generators.

Yet people still pay for handmade personalization.

Why?

Because sometimes the imperfection is the product.

Think:

weddings...

luxury events...

corporate gifting...

personalized products...

live brand activations.

The transcript specifically describes wedding calligraphy, on-site personalization, corporate events, greeting cards, and workshops as possible monetization paths.

Notice what's happening.

You're not merely selling:

You're potentially selling:

Much harder to reduce to a prompt.

👉 When digital perfection becomes cheap, handmade imperfection can become part of the value.

The same idea applies to art.

Instead of asking:

"Can AI make a portrait?"

Obviously it can generate imagery.

Ask:

Now we're talking about something different.

A live painter.

A typewriter poet.

A sketch artist.

A caricature artist.

An interactive art station.

The transcript specifically emphasizes interactive art—including live event painting, poetry, workshops, and handmade client gifts—as a way artists can differentiate their work.

The output matters.

But so does:

the performance...

the memory...

the story...

the human interaction.

This gives us our first major strategy.

If AI can reproduce your output, ask:

Instead of:

Painting

try:

Live Wedding Painting

Instead of:

Poetry

try:

Personalized Event Poetry

Instead of:

Calligraphy

try:

Live Luxury Gift Personalization

Same underlying skill.

Completely different value proposition.

👉 AI can compete with a product more easily than it can compete with an experience.

Here's about as unsexy as side hustles get:

Handyman work.

Hole in drywall?

AI can tell you how to patch it.

Sink cabinet loose?

AI can explain the repair.

Room needs painting?

AI can generate the instructions.

Customer still has a problem:

The source specifically mentions home repair, painting, and handyman services as examples where existing practical skills can become paid work.

This category could become increasingly interesting.

The more people outsource thinking to software...

the more physical execution remains stubbornly attached to the real world.

Ask this about any side hustle:

AI creates landscaping plan.

Someone installs it.

AI suggests room layout.

Someone moves everything.

AI diagnoses simple website problem.

Someone implements the fix.

AI writes event plan.

Someone runs the event.

AI creates renovation concept.

Someone builds it.

That final gap between:

and

is the last mile.

And customers pay to cross it.

👉 Information is becoming cheaper. Execution is still valuable.

Nails.

Lashes.

Skin treatments.

Other appropriately licensed beauty services.

AI can generate:

design inspiration...

color combinations...

marketing...

appointment reminders...

social posts.

But the actual service?

Physical.

The transcript mentions home-based nail, eyelash, and related beauty services while correctly noting that these can require more training, certification, materials, and startup expense.

That caveat matters.

Before offering any regulated personal-care service, check the applicable licensing, sanitation, insurance, zoning, and local business requirements.

But economically, the underlying principle is clear:

Use technology to acquire the customer.

Use human skill to deliver the value.

Can AI generate a gorgeous wedding photograph?

Yes.

Can it photograph your actual wedding as it happens?

Different problem.

That's why event photography and videography are interesting.

The transcript specifically points toward weddings and corporate work and recommends starting with smaller shoots to develop experience and proof before gradually increasing pricing.

And there's another lesson here.

Your first goal isn't necessarily becoming:

Your first goal is becoming:

Then improve.

Portfolio.

Customer.

Feedback.

Better portfolio.

Better customer.

Better pricing.

👉 Don't wait until you're the best. Get good enough to create value, then let paid experience make you better.

Now we move from hands...

to something even more human.

Being there.

A dog doesn't care how advanced your language model is.

It needs:

food...

water...

walks...

attention...

supervision...

and somebody physically present.

The transcript distinguishes between boarding an animal in your home and staying or visiting at the owner's home, while acknowledging that your living and work situation affects whether the model is practical.

That's important.

Not every side hustle needs to fit everybody.

Instead ask:

Remote worker?

Flexible schedule?

Love animals?

Comfortable staying elsewhere?

That combination might create an advantage.

This may be one of the most human opportunities in the entire transcript.

Community companionship / peer or respite support.

The source describes helping someone in the community with outings, companionship, and occasional respite support.

This isn't simply selling:

time.

It's selling:

reliability...

companionship...

trust...

presence.

Depending on the nature of the work and jurisdiction, some support roles may require screening, training, certification, agency involvement, or other qualifications. So this isn't something to enter casually.

But the business principle is enormous:

And care is deeply human.

👉 Presence isn't inefficient when presence is the product.

AI can explain:

piano.

French.

Math.

Cooking.

Knitting.

Calligraphy.

Kombucha.

So why would anybody still pay a teacher?

Because teaching isn't simply:

A teacher:

notices mistakes...

adjusts explanations...

provides accountability...

encourages...

answers follow-ups...

adapts to the learner.

The transcript recommends teaching something you already know—through ongoing lessons or even one-off workshops.

That's a much bigger opportunity than:

"Become a tutor."

Ask:

Excel?

Photography?

Sourdough?

Public speaking?

Makeup?

Chess?

Video editing?

Gardening?

Language?

Then package it.

Suppose you create a workshop priced at:

You don't necessarily need one person to pay you $1,000.

You need:

Or:

That's why the first $1,000 becomes easier to understand when you stop treating it as one giant number.

More on that shortly.

Now things get more interesting.

Because several opportunities in the transcript can absolutely be affected by AI.

That doesn't necessarily mean they're dead.

It means the value proposition has to change.

AI can generate:

scripts...

voices...

avatars...

images...

videos.

But it doesn't have:

your childhood...

your embarrassing first job...

your failed business...

your strange hobby...

your personality...

your relationships...

your lived experience.

That's where human content still has an advantage.

The transcript recommends creating around something you're already genuinely interested in or naturally knowledgeable about, emphasizing relatable human content rather than trying to manufacture a personality for an algorithm.

That's the play.

Don't compete with AI at:

Compete at:

👉 When generic content becomes infinite, specificity becomes scarce.

Here's one of the more unusual ideas in the transcript:

A paid physical mailing club.

Imagine somebody pays monthly and receives:

an original print...

letter...

poem...

mini illustration...

postcard...

sticker...

or another small creative piece.

Digital audience.

Physical delivery.

The transcript describes subscription-style mailing lists where followers pay monthly to receive physical mail from a creator, while also acknowledging the postage, material, and labor costs involved.

This is fascinating because it reverses the usual trend.

Everyone moves:

You move:

Scarcity returns.

This matters especially here.

Suppose:

100 subscribers × $10 = $1,000 revenue.

But you still have:

postage...

printing...

materials...

packaging...

payment fees...

your time.

So:

We'll keep that distinction throughout this issue.

UGC is often confused with influencer marketing.

They're different.

In the model described in the transcript:

Brand pays partly for access to your audience.

Brand pays you to create content it can publish through its own channels.

That's important because it separates:

from

Could AI-generated UGC compete?

Absolutely.

The transcript acknowledges that too.

So your defense isn't pretending AI doesn't exist.

It's becoming better at:

authentic demonstrations...

natural delivery...

brand interpretation...

product understanding...

creative concepts...

human credibility.

Suppose somebody wants to redesign their living room.

AI can provide:

ideas...

layouts...

shopping suggestions...

color palettes.

But the customer may still think:

That's the opportunity behind consulting.

The transcript gives an example of interior-design consulting where customers provide photos or videos and receive personalized recommendations.

This distinction matters enormously.

Information says:

Consulting says:

That's judgment.

AI is extraordinarily good at producing information.

So move upward.

"What are my options?"

AI is excellent here.

"What does this mean for my situation?"

More valuable.

"What should I do?"

More judgment.

"Can you do it for me?"

More valuable again.

"Can you make sure this actually happens?"

Now we're getting very human.

👉 The closer you move from information toward outcomes, the stronger your value proposition becomes.

AI video editing is improving rapidly.

So saying:

"AI can't edit videos"

would be silly.

It can.

The question is whether a customer wants:

or:

The transcript argues that increasing amounts of video from creators and businesses can continue creating demand for editors, and it gives podcast production as one example extending beyond simply cutting footage.

That's the opportunity.

Don't become:

person who cuts clips.

Become:

person who turns a weekly podcast into a publish-ready content package.

Now maybe you handle:

editing...

audio cleanup...

thumbnail...

upload...

chapters...

short clips...

publishing workflow.

And yes:

use AI wherever it makes you faster.

👉 Don't compete with the tool. Become the person who knows how to use the tool to deliver the outcome.

AI website builders exist.

Templates exist.

No-code builders exist.

Free tutorials exist.

And people still hire web designers.

Why?

Because:

possible to DIY

doesn't mean:

customer wants to DIY.

The transcript makes essentially this argument: AI can help create websites, but getting a specific desired result can still require time, effort, and judgment.

Again, don't sell:

Sell:

For example:

"I build simple booking websites for independent therapists."

Now you can understand:

their customers...

booking flow...

services...

credibility requirements...

FAQs...

calls to action.

AI becomes part of your production stack.

Not necessarily your replacement.

There's one final opportunity.

And AI barely matters to the fundamental economics.

Flipping.

Buy/find something.

Sell it for more.

Or begin by selling things you already own and no longer need.

The transcript recommends starting with unused furniture, technology, clothing, and similar household items if the immediate goal is generating extra cash, while noting that consistent flipping requires deeper knowledge of a particular category.

That second part is the business.

Anybody can sell an old chair.

A serious reseller understands:

which chair...

which brand...

which condition...

which market...

which price...

which buyer.

That's expertise.

The business isn't:

That's the outcome.

The actual skill is:

Maybe your niche is:

cameras.

Sneakers.

Vintage furniture.

Musical equipment.

Trading cards.

Designer bags.

Tools.

Electronics.

Specialization creates speed.

You recognize opportunities other people scroll past.

$1,000 feels significant.

But it becomes much less intimidating when you convert it into transactions.

1 client × $1,000

2 clients × $500

4 clients × $250

10 customers × $100

20 customers × $50

100 customers × $10

All equal:

But these are completely different businesses.

That's why your first question shouldn't be:

Ask:

Now we can build something.

Imagine four possibilities.

These are illustrative business-model examples, not typical prices or guaranteed outcomes.

Their purpose is to make the target concrete.

Because:

isn't a strategy.

is getting closer to one.

If I were choosing a side hustle for the next several years, I'd want at least two of these three:

Trust.

Taste.

Relationship.

Judgment.

Personality.

Hands.

Location.

Presence.

Real-world execution.

You don't merely provide information.

You help something actually get done.

The strongest opportunities may sit in the middle.

For example:

Human + Physical + Outcome.

Physical + Outcome + Trust.

Human + Judgment + Outcome.

Human judgment + Outcome, increasingly assisted by AI.

Human + Physical presence + Relationship.

That's much more useful than stamping:

AI-PROOF

on a job title.

Here's the irony.

You should probably use AI.

Handyman?

Use it to help draft estimates or organize schedules.

Photographer?

Use tools to accelerate culling or administrative work.

Consultant?

Use it for research and first-pass analysis.

Video editor?

Use it for transcription, rough cuts, captions, or repetitive production steps.

UGC creator?

Use it to brainstorm concepts.

Workshop teacher?

Use it to organize lesson materials.

Web designer?

Use it to accelerate drafts and development.

The goal isn't:

It's:

Let technology handle more of the repetitive work.

Keep your attention on the parts customers actually value you for.

👉 The safest position may not be refusing AI. It may be owning the human outcome while using AI underneath it.

Don't start 15 side hustles.

Choose one.

Write down:

What can I physically do?

What do people trust me to do?

What am I naturally good at?

What do people ask me for help with?

What do I genuinely enjoy?

Find the overlap.

Not:

"I do photography."

Try:

"30-minute outdoor family mini-session."

Not:

"I edit videos."

Try:

"Weekly podcast editing + publishing package."

Not:

"I teach music."

Try:

"Four-week beginner guitar program."

Specific offers are easier to understand.

Start close.

Friends.

Family.

Former coworkers.

Neighbors.

Local groups.

Existing audience.

Professional network.

Relevant businesses.

The transcript repeatedly mentions networking, social media, referrals, or existing relationships as ways several of these services found early customers.
Don't spam everyone.

Find people who plausibly have the problem.

Customer #1 isn't only about money.

Pay attention to:

What did they actually value?

What took too long?

What could AI automate?

What required you?

What did they ask for next?

Would they refer somebody?

Would they buy again?

That information tells you whether you have:

a gig...

or the beginning of a business.

AI will keep getting better.

Trying to identify the one career technology will never affect isn't much of a strategy.

Instead, move toward what remains scarce.

Then use AI to make the rest of your operation faster.

That's the opportunity.

Because customers don't ultimately pay you because you're human.

They pay because something they care about gets:

fixed...

created...

captured...

taught...

improved...

delivered...

or handled.

AI may change how you accomplish that.

But if you continue owning the outcome...

you still have something valuable to sell.

Don't pick from all 15.

Choose the one closest to something you already:

Then write:

MY OFFER: ______

MY CUSTOMER: ______

MY PRICE: ______

CUSTOMERS NEEDED FOR $1K: ______

WHY THEY NEED A HUMAN: ______

HOW AI CAN MAKE ME FASTER: ______

Now you're no longer asking:

You're building one around a much better question:

The future probably won't belong exclusively to people who can do things AI can't.

That list may keep shrinking.

It may belong to people who know:

what technology should do...

what humans should do...

and how to combine both to produce something customers genuinely value.

👉 Don't try to become AI-proof. Become useful in ways AI makes harder to commoditize.