SaaS solved a big problem
Software as a Service made software much easier to buy and use.
Instead of installing, maintaining and running systems ourselves, a company could simply log into a browser and pay continuously for a service.
That was a huge shift.
CRM, accounting, project management, marketing, HR, design, collaboration and hundreds of other functions became subscription services.
SaaS turned software into a service.
But there was also a downside:
Companies often started adapting themselves to the software.
You choose the system.
You choose the workflows the system supports.
You teach your employees how to use it.
And when the company's needs change, you may start looking for plugins, integrations, workarounds or an entirely new system.
That is not necessarily a problem. Standard software is excellent when what you actually need is a standard solution.
But what happens when it becomes cheap enough to build something yourself?
The old barrier wasn't the idea
Most companies have plenty of ideas for software.
The problem has traditionally been the cost and complexity of turning those ideas into reality.
You had to find developers.
Plan a project.
Design the architecture.
Set up servers.
Build authentication.
Build a database.
Handle payments.
Test.
Deploy.
Maintain.
And every single part could become a project of its own.
This is where something fundamental has changed.
Today, enormous amounts of infrastructure already exist as services.
Database.
Authentication.
Hosting.
Payments.
File storage.
Email.
Realtime data.
Analytics.
AI.
APIs.
Deployment.
You don't necessarily need to build the entire machinery yourself.
You can buy the building blocks and use them to build the product you actually need.
Buy the primitives. Build the product.
AI is also changing who can build
The next layer is AI.
I don't work as a traditional software developer. My background is in design, visual communication and creative digital projects.
Yet today I can use AI as an active part of the development process and go from an idea to working software much faster than before.
That doesn't mean I press a button and get a finished product.
It means the distance between an idea and working software has become dramatically shorter.
AI can help write code, structure databases, build components, find bugs, explain APIs and iterate on functionality.
But AI doesn't automatically know whether what is being built is actually the right product.
That still requires people.
Someone needs to understand the user.
Someone needs to understand the problem.
Someone needs to make the trade-offs.
Someone needs to recognize when a solution only looks right on the surface.
That is why I think the interesting change is not just about AI.
It is about who can now take an idea all the way from thought to working software.
I've built an example myself
A concrete example is Klokly.
Klokly is a SaaS solution for time tracking, but what interests me is not the timer itself.
It is how much of a real digital business can be built using a relatively small number of specialized building blocks.
Klokly uses technologies including React / Next.js and TypeScript, Supabase for database, authentication and realtime functionality, and Stripe for payments. It has user management, teams, live status and a real payment model.
It is no longer just a prototype.
It runs as a real product.
And this is where my SaaS provocation begins.
Because if one person or a small team can build this kind of software much faster than before, the question of buy versus build starts to look different.
Maybe the question is no longer:
Should we buy or build?
Maybe it is:
Which parts make sense to buy, and which parts make sense to build ourselves?
Software is becoming more specialized
Imagine a company with a workflow that no existing SaaS platform fits particularly well.
Previously, the choice might have been:
"Buy the next-best system and adapt the process."
Or:
"Pay a development agency a lot of money to build our own."
Today, there is a third option.
Combine the technical building blocks and build exactly the workflow the company needs.
It doesn't have to be a gigantic ERP system.
It could be a small internal tool.
An ordering platform.
A customer portal.
A product configurator.
A specialized app.
A dashboard.
A digital sales tool.
A browser extension.
An automation.
Or an entirely new type of product.
This is where Software as I Want It becomes interesting.
But here comes the important contradiction
Software has become easier to build.
It has not become equally easy to run good software.
AI can generate code quickly.
That doesn't mean the system is automatically secure.
A database can work.
That doesn't mean the data model is good.
A login can work.
That doesn't mean permissions have been designed correctly.
A payment integration can work.
That doesn't mean all edge cases, refunds, webhooks and subscription changes are handled properly.
And GDPR compliance doesn't happen automatically just because you use a particular cloud service.
Data protection needs to be considered in the design and architecture from the beginning, and companies remain responsible for both technical and organizational security measures.
That distinction matters.
AI has made software cheaper to produce. It has not removed responsibility for software.
AI doesn't make bad ideas better
There is another danger.
When it becomes easier to build, it also becomes easier to build something nobody needs.
That is almost a direct consequence.
Previously, the technical barrier was so high that many bad ideas died before they ever became products.
Now they can have a landing page, database, login and payment flow in a very short time.
That is both exciting and problematic.
Because when software becomes cheaper to create, product understanding becomes more important, not less.
A bad idea doesn't become good because it was built quickly.
A poor user experience doesn't become good because it was written by AI.
And more code does not equal more value.
"Can you build it?" is no longer the only question
I think we are therefore moving toward a new sequence of questions.
First, people asked:
Can it be built?
Then came:
Can we afford to build it?
Increasingly, the question becomes:
Should we build it?
And perhaps even more importantly:
What software would actually fit the way we work better?
That is a much more interesting conversation than simply choosing between 37 SaaS tools.
SaaS isn't going away
It is important not to overstate the point.
SaaS still makes enormous sense for many things.
I have no interest in building my own accounting system, email platform, payment infrastructure or database.
Why would I?
There are companies that do those things better and at a much larger scale.
The interesting shift is therefore not that companies will suddenly build everything themselves.
It is that more of the middle layer can become custom-built.
SaaS provides the standard functions.
Cloud provides the infrastructure.
APIs provide the connections.
AI reduces development friction.
And the company can shape the layer that makes its own way of working unique.
"Software as I Want It"
That is why I think Software as I Want It is an interesting way to describe the development.
Not as a new technical standard.
Not as a replacement for SaaS.
But as a change in mindset.
From:
Here is the software. Adapt to it.
To:
Here are our needs. Let's shape the software around them.
That could change much more than the development process itself.
It could change how companies think about digital products.
And then search enters the picture
There is another interesting layer.
As it becomes cheaper to create software and content, it also becomes more important to make it understandable and discoverable.
Not just for humans.
But for search engines and AI systems.
Google has made it clear that fundamental SEO principles still apply to generative search experiences. There is no magic GEO or AEO shortcut. Google particularly emphasizes useful, original and reliable content, a technically clear structure and crawlability.
This means that a modern digital solution increasingly needs to do more than simply work.
It needs to be:
usable
understandable
discoverable
explainable
and increasingly understandable to machines.
This is also something I work with directly at KodeKid, where the site itself is built as an example of this approach.
The new question
I don't think the future will be:
SaaS or custom software.
It will be a combination.
The best solutions may very well consist of SaaS, cloud services, APIs and AI combined with software built specifically for the company.
In that world, SaaS doesn't necessarily become less important.
It simply becomes less visible.
It becomes the infrastructure underneath the software layer the company actually experiences.
And perhaps that is the real change:
We are moving from a world where software is primarily something we subscribe to, to a world where software is increasingly something we shape.
Software as a Service gave us access to software.
AI can help give us the ability to shape it.
That is quite a significant difference.
Software as I Want It
Maybe SaaS isn't dying.
Maybe SaaS is simply moving down the stack.
And perhaps the top layer becomes:
Software as I Want It.
Not software we have to fit ourselves into.
But software that fits what we are actually trying to build.
That is the development I think is worth watching.
FAQ
What does Software as I Want It mean?
Software as I Want It is a term for a development direction where companies can increasingly shape or build software around their own needs instead of simply adapting their workflows to standard software.
Is SaaS disappearing?
No. SaaS is still an effective model for standardized functions such as accounting, CRM, payments and collaboration. The point is that more specialized functions may increasingly become worth building yourself.
What has AI changed in software development?
AI can significantly reduce friction in the development process. It makes it faster to prototype, write code, test ideas and iterate. However, it does not eliminate the need for product understanding, technical judgment, testing and operations.
What does "Buy the primitives. Build the product" mean?
It means you do not necessarily need to build every technical foundation yourself. You can use existing services for things such as databases, payments, hosting and authentication as building blocks for a more specialized product.
Can a small business build its own software?
Yes. Modern cloud platforms, APIs and AI tools make this more realistic than ever before. That does not mean every company should build its own software. The key is to evaluate where a custom-built solution provides a real advantage.
Can AI build a complete SaaS product?
AI can help with large parts of the development process, but a product requires more than code. Architecture, security, user experience, data handling, testing, payments, operations, support and ongoing maintenance still need to be managed.
Are GEO and AEO the new SEO?
Not necessarily. Google describes generative AI search as part of the existing Search experience and emphasizes that fundamental SEO principles remain the foundation. GEO and AEO can be useful concepts, but they should not be treated as magic shortcuts.
What is the biggest change?
The biggest change may not be that software is becoming easier to build. It is that more people and businesses can go from an idea to working software without first having to build the entire technical infrastructure from scratch.
Author: Lars Sohl, founder of Klokly.app and KodeKid