Go Where the Engineers Aren't

Sales > Software.

He always knew somebody who needed something built.

He'd find the work.

I'd build the thing.

Software engineers tend to do the opposite.

We spend six months building the world's most technically elegant product that seven people know exists (1 is our mom).

Salespeople will occasionally sell something that doesn't even exist yet.

This deeply offends us.

Unfortunately, they keep making money!

I started hanging around salespeople again

Recently, I started joining communities filled with agency owners, recruiters, salespeople and business owners.

I'm not looking to become a freelancer again.

I want to see what people outside tech are actually trying to build with AI to experiment with a new business venture.

And this has been... illuminating.

A lot of them are already building software themselves.

With Claude, Cursor and enough misplaced confidence, people are deploying applications that are pretty damn good.

Until they're not.

Eventually they hit auth, integrations, bugs, architecture, security, deployments or a codebase that has become so horrifying that even Claude starts sounding a little concerned.

That's where engineers become useful.

I met a woman in one of these communities who had built sales automation software for car dealerships.

She understood the sales process extremely well and had already built a huge portion of the product herself.

She told me:

"I don't want to be an engineer."

Agreed.

She knows the customer and sales.

I know software.

We're now talking about working together.

Another connection led me to a father-and-son team building software that integrates with CRMs so salespeople can spend less time updating Salesforce and more time selling.

Notice what neither of them asked me:

"Do you know LangGraph?"

"Explain RAG to me like I'm 5..."

"Tell me how'd you debug an agent at 12AM"

These kind of ridiculous questions don't really exist outside tech.

This is what I think AI engineers should actually practice if they want to freelance

Get a free HubSpot account.

Build an agent that can inspect deals in the CRM.

Ask:

"Which deals haven't moved in two weeks?"

Then:

"Move Acme to Contract Sent and add a note that we're waiting on legal."

Now your model isn't just generating text.

It has to choose an action, call an external service, handle the response and avoid doing something incredibly stupid.

What if there are two companies named Acme?

What if HubSpot fails?

Should it be allowed to delete a deal?

Should some actions require approval?

You accidentally started doing AI engineering.

If you want to make it more interesting, give it another tool.

Have it call a customer after an appointment is booked, confirm the appointment, then update the CRM with the result.

Now you're coordinating multiple systems and handling failures.

This is the stuff businesses are trying to pay people to build.

My hottest take: don't start a business

After telling you all of this, here's what I think most software engineers should actually do:

Don't freelance. Don't start an agency. Don't quit your job.

Build the skills.

Build proof.

Then go get paid more.

If you're making $150K and these skills help you land a $190K or $200K role, that's another $40K–$50K every year.

For most people, that's life-changing money.

And it's much simpler than starting a business.

I've been making money independently in one form or another for about ten years.

It always takes longer than you think.

Three months.

Six months.

Sometimes a year before anything starts feeling consistent.

Meanwhile, if you're already a strong engineer, you don't need to reinvent your entire life.

Learn RAG, agents, evals.

Learn how to connect models to APIs, databases and business software.

Then build something and show people what it does.

A deployed HubSpot agent with a two-minute demo is infinitely more interesting than:

"Completed Advanced Generative AI Certificate."

D'oh.

And you can do the same thing inside the company you already work for.

Pay attention to what sales, customer support, QA, operations and other teams actually do all day.

Ask questions.

Listen when people complain.

Understand their processes.

Then figure out whether there's something you can build that makes one of those processes dramatically better.

That's often how great software starts.

Not with:

"I want to build an agent."

But:

"Wait... why the hell are we still doing this manually?"

I know some of you aren't going to listen

You've been warned.

If you truly want to dip your big toe into the murky pool of freelancing and business, here's where I'd start.

Join three free communities filled with business people.

Start with Facebook groups or maybe Skool.

Yes, really. Facebook aka Boomer Town Square.

A surprising number of people with actual money still hang out on Facebook.

A surprising number of people pretending to have money hang out on Instagram.

Hang around for a week.

Don't sell anything.

Don't introduce yourself with:

"Hi everyone! I'm Brian, an AI automation expert helping forward-thinking businesses leverage cutting-edge agentic solutions..."

Jesus Christ.

Just be a person.

Read what people are talking about.

Look for:

"We spend forever doing..."

"I wish there was a way to..."

"I built this with Claude, but..."

"Does anybody know someone who can..."

Then, after you've been around for a bit, post something useful you've built.

Don't pitch it.

Show it.

Maybe:

"Saw a few people here talking about how annoying keeping HubSpot updated is, so I screwed around this weekend and built something that lets me update deals through chat. Here's how it works."

There are already plenty of people trying this strategy.

Most of them suck.

They spam.

They exaggerate.

They copy and paste AI garbage.

They build demos that fall apart the second somebody asks a question that wasn't in the YouTube tutorial.

So your competitive advantage is surprisingly simple:

Don't suck.

If you're already an experienced software engineer, you're starting with an enormous advantage over most people trying to sell this stuff.

You understand databases.

APIs.

Authentication.

Deployments.

Testing.

You know what happens when software inevitably catches fire.

Now layer AI engineering on top of that and learn to understand the business problem before you start vomiting technology at people.

That's a pretty powerful combination.

And whether you use that to get a better job, become more valuable where you already work, pick up a client or eventually build a business is up to you.

But if your goal is simply making more money?

Don't automatically choose the hardest path.

Sometimes the easiest path is becoming much more valuable at the thing you're already doing.

And if you want to learn how to build the kinds of AI systems I've been talking about here, RAG, agents, evals, integrations then you're in luck. That's what we teach at Parsity ;)