What Happens When the Cost of Building Software Drops 95%?

Everyone's focused on AI job losses. Almost nobody is talking about the projects that couldn't exist before. When the cost of execution drops by 95%, projects that were never fundable suddenly become viable bets — and the demand for talent to build them explodes. That's Jevons Paradox in action, and it's the real AI story business owners need to understand.
In this episode, Michael Pullman breaks down why cheaper coding doesn't mean fewer developers — it means more software ships, more niche markets get served, and more roles emerge to support that growth. He walks through a real construction company use case, shares his own vibe-coded productivity tool, and makes the case that operators who move now will capture disproportionate upside from the demand expansion AI is creating.

What is The AI Operator?

AI Operator is a three-episode-per-week podcast for CEOs, founders, and senior operators who are past the curiosity phase of AI and moving into the deployment phase. Each week features three formats: AI News for Operators, framing the latest AI developments through the lens of business impact; Operator Playbook, teaching frameworks for deploying AI inside organizations; and Frontier Builders, interviews with creators building real, production-ready AI systems. AI Operator turns AI into output, not experiments.

Michael Pullman: So I had a conversation
with a friend of mine yesterday about

her losing her job because of ai.

She's a middle level type in a tech
company, and yeah, there's some risk

of her losing her job because of ai.

The risk is mitigated
by the fact that she.

Can learn to use ai, but the real
story's not about job replacement.

It's about the fact that AI has collapsed
the cost of execution dramatically.

So what that means is projects
that weren't fundable that you

would look at before and say,
Hmm, we're not gonna do that.

The cost benefit's not there.

Now they're viable bets.

Welcome to the AI Operator Podcast.

If you're a CEO founder or a business
owner trying to turn AI into real

business value, you're in the right place.

I've got this fancy new microphone here.

Check that out.

I feel like a real podcaster.

Welcome to the AI operator.

There's a lot of noise in AI right
now, so there's new tools, new

models, new benchmarks, stuff coming
out all the time and people shouting

about the future, the future.

But most business leaders like me and like
you, probably are stuck with one question.

And the question is, how do I, what do I
actually do with this inside my company?

And that's what this show's all about.

The AI operators for leaders
who want to move beyond.

Experimentation and geekery and
toy use cases into like actual

real operational leverage.

With ai, we're talking about practical
workflows, repeatable systems, and then

real world applications that can help
businesses save time, increase output, and

improve decision making, which is gonna
create a measurable return on investment.

What nobody's talking about
except this one guy on Twitter.

X is the projects that
didn't exist before.

It's only, everyone's only talking
about the jobs that have disappeared.

Think about the projects that weren't
possible before that can now be funded.

So it's this classic thing
about demand elasticity.

If you've done economics or look
up demand elasticity when the

price of something goes down 80%.

Demand doesn't just stay at the
same level, it massively expands

to fill that new price point.

So because the cost of coding is going
down to not zero, but going down a

long way, let's say it collapses by
95%, the possibility of the projects

that can be funded now goes way up.

So the hypothesis of this guy,
and, and I agree with him,

Aaron Levine on Twitter is.

That perhaps we're not going
to see massive job loss.

We're gonna see Jevons Paradox.

And Jevons Paradox is when technological
advancements increase resource efficiency,

making usage cheaper, which then leads
to a higher total consumption rather

than a decrease in total consumption.

So because we're making things
cheaper with AI now, there's gonna

be so much more demand for it.

Really good example is a piece
of software I wrote recently.

Yes I've vibe coded it
using, completely using AI.

For years and years I've been
wanting a piece of software that

would help me organize my life, like
a, getting stuff done type set up.

But, I subscribed to a particular
methodology, which is from Tony

Robbins, which is rapid planning
method, or put out your result, put

out your purpose, why you want it.

So what do I want?

Why I want it, and.

My massive action plan,
how am I gonna get it?

All the tasks I need to do to get it.

So I built a piece of software
that's customized purely for me.

So that I can put together my plans
and clarify my thinking in that.

Now, that wouldn't have
existed beforehand.

I had pads, I had notepads, I had books.

None of it was good enough.

It didn't, none of it worked.

This works for me.

It's a piece of software that lives on my
Mac Mini here, on, and then it's using a

model that's on my Mac studio and I can
pull it up on my phone and I can just

type things into the piece of software.

So it's massively helping me stay
organized in my life and achieve.

All of this, which is pretty cool.

So the cool thing is that even if
the, even if we do lose some jobs

because of ai, the demand for all
of the people that can do things

with AI is gonna massively increase.

So if you're a person who's thinking about
your future and what it looks like in the

age of ai, just learn to do stuff with ai.

Just get in there, use the
models, do what you can, and

you'll position yourself above.

All of the people who are scared of ai.

That was my advice to my friend.

Jevons paradox has happened with
every major technology, labor

saving technology in history.

The AI version that we're
seeing now isn't new.

It's the same paradox that we always
see, but it's happening heaps faster.

It's happening a lot faster.

So in action, we're seeing it as cheaper
coding means more software is shipping

means more support, more project manager,
more integration roles are needed.

And yes, AI is gonna help people
be more productive in those roles.

But that's a good thing.

That means that the support,
the project management, the

integration, it all gets cheaper.

Therefore, we can use more of it.

So a lot more software is going to ship
and it'll be software for niche use cases.

Alex Finn gave a really good
example on x, be the CRM for

Korean grocery stores, right?

Who's gonna write a CRM?

Salesforce is never gonna write
a CRM for Korean grocery stores.

Zoho is never gonna write a CRM for
Korean grocery stores, but you can,

and then you can just completely
dominate that niche, that market.

So that's as business owners.

So as business owners, that's what we
need to be thinking about is going into

a niche rather than trying to serve
everybody with our product, serving a

small niche, and maybe we have a series of
small niches and we customize the product

slightly for those different niches.

It's a huge opportunity.

So the example that was given to
support this hypothesis is that.

The most under-reported cases is the
team that went from zero engineers to 10.

So a project that never existed
that is now getting built because

of the ROI because of the value.

So if you had, let's say a great
example that they gave on Twitter on x,

If you had a feature that
takes 50 engineers to build.

But you couldn't justify
building it because it would,

it's a such a huge investment.

Maybe it's only serves a small slice of
your product, of your of your market.

It can become a 50 person team, can
become a two person project team.

Once you get agents and AI involved,
and then you can ship that feature.

And then guess what?

You can ship another feature like
that and another feature like that.

So what we're seeing, what we're going
to see, and what it probably means

for us is we're going from, we can't
afford this to let's hire and build it.

So in the old model, we wouldn't
have done that 50 person project.

It just would not be viable.

It's too much cost for
too small a benefit.

Now, if it only takes five people to do
that same job, maybe we fund that project.

The AI agents come in and help with
the heavy lifting on the code, and

the humans are just the orchestrators,
so we still need developers.

But what happens is it's not just
the one project that gets funded,

it's all 10 of the projects and net
net, we end up with the same number

of engineers employed because they
can produce code faster and cheaper.

So that means that we actually see.

Net.

Net the same employment.

And my hypothesis is because
of Jevons Paradox, we're

gonna see massive employment.

So everybody's scared about
losing their jobs with ai.

I think we're gonna see
so many more jobs with ai.

It's just gonna be a rough ride.

The people who are the laggards, the
late majority on the technology adoption

curve, they're the people who will.

Struggle to find a job if
they're not embracing ai.

They're not the ones
watching this podcast.

We can't help them yet.

Maybe there's a ton of
stuff they could do online.

There's a lot of courses to
come up to speed with it.

But it's something that we need to
be thinking about as business owners.

There is that technology adoption
curve, and we need to keep in mind

that some people are gonna adopt
this technology straight away.

A lot of people are gonna be a bit
slower to adopt the technology.

And just like with any technology,
new roles are going to emerge.

So think about when, the classic example
is the Luddites that came together

and smashed the automatic loom or
the loom when it came into operation,

they weren't happy, they were weavers.

The best weaver in their town
was making the most money.

And then they came together
and said, no, we're not gonna

put up with this technology.

We're gonna go and smash it now.

That was the beginning of
the industrial revolution.

And look what came out of it.

We can buy clothes for.

T-shirts for $11, and
there's so many more roles.

There's so many people, so
many maintenance engineers

who look after the looms.

The same thing is happening with ai.

We have.

A convergence of three
really exciting technologies.

One is ai.

Everybody knows about that.

Our large language models, large world
models, large quantitative models.

There's some really cool stuff coming
that I need to do an explainer on all

the different terms that we're seeing.

So large language models or AI models.

Robotics.

Robotics is going to be massive.

So we're gonna have so
many robotics engineers.

I've got a 10-year-old daughter
who's into engineering.

I said, let's buy a robot kit.

Let's get you building robots,
because the roles of the

future are in humanoid robots.

You've got figure, you've got
optimists coming out from Tesla.

You've got, there's, I was listening
to the abundance 360 live stream

and there was half a dozen robot.

Humanoid robot companies
mentioned on that live stream.

It's coming and it's gonna come fast.

So there's gotta be people to
maintenance those robots to service

those robots, to build those
robots, to program those robots.

So the three technologies are
AR models, robotics and quantum.

Quantum is in its infancy now,
but I think we're at the bottom of

the inflection point with quantum.

It's going to be massive
because what it allows us to

do is generate synthetic data.

Let's say that you were doing
materials engineering, you

wanted to material science.

You wanted to find a particular type of
concrete that was stronger and lighter.

You feed that into these large
quantitative models, these quantum

models, and it builds out it.

It takes all the different
possible combinations.

Of molecules and atoms and simulates
their interactions and generates

billions of synthetic data points.

And the AI models can then sort through
those, that synthetic data and find

the ones that are most promising.

And then it might serve up 200 to be
tested as opposed to in the old model,

we had a scientist putting pipet into.

Current chemicals into different
pipes and petting chemicals into

different Petri dishes and testing
these chemicals one by one.

Now we can generate billions
of data points and narrow it

down to the 200 most promising.

So material science is gonna
change as we know it very shortly.

Now there's a slightly uncomfortable
counterpoint to this, and that's that.

Higher wages go to the
workers who survive the cut.

The workers who are enabled with AI
and who have enabled themselves with

ai, they've learned the technology, but
not everyone's gonna survive that cut.

So the late majority that we were
talking about, they're not gonna

survive, but they do have an opportunity.

They can become creators,
they can become entrepreneurs.

They can build a podcast like this.

They can start building things
online and building in public.

That's huge on X right now.

So it's something that we need to
be coaching the people who aren't

surviving the cut and teaching them,
Hey, you'd need to get onto x onto

the internet and sign up for a course
and just learn about this stuff.

But there's a bump.

Not everybody's gonna be ready
for entrepreneurship, right?

There's a transition period that.

I don't think we really
have a clean answer for yet.

We don't know how those people
are going to survive that period.

It's some job categories, like DevOps
some monitoring or like quality

assurance roles, they're gonna face
demand that doesn't bounce back.

So those people are gonna
need to re-skill pretty quick.

So if you're listening to this podcast,
the question isn't whether to adopt the

software, whether to adopt the new tool.

It's whether you can be the architect
of a new system or you're just hoping

somebody else designs it around you.

If you're watching this podcast,
you are the orchestra conductor.

You know that you need to
do something about this.

So get out there, do the
research, find a person who can

help you implement this stuff.

And have some fun, get in there and
automate parts of your business that you

never thought could be automated before.

I had a conversation just this
morning with a business owner.

He runs a company, a small construction
company, and he said the biggest pain

for him is talking with customers.

And so not talking with customers
in the initial sales phase.

He loves that.

It's talking with customers
during projects, keeping them

apprised of progress and.

We were just in the early stages.

I had to bite my tongue and
say, let's just text them.

Let's just text them every time
something changes on the project.

And but I'm designing a system for
him rather than just interjecting

with individual bits of usefulness.

I just thought I'd hold back and
I'm designing a full system for this

construction company around how we
can take the inputs from the workers.

Who are working on these projects and feed
that back into a central system to notify

customers of progress on their build.

So that's gonna be a
really exciting project.

I'm gonna see if the owner's
open to it, if we can build

that in public on this podcast.

But if you've got a use case you've got
a problem, a business problem you want

solved, or you're just curious about
AI and you think you've got a problem

that AI might be able to help with,
reach out to me at x at Michael Pullman

or send an email [e-mail].

I'm Michael Pullman.

This is the AI Operator Podcast.

Thanks for watching.

I'll see you again soon.