When building becomes easier than subscribing, business models shift.

I replaced Trello in an afternoon.

I replaced the system I use to manage all my tasks and projects.

For years, Trello has been my to-do list. A Kanban board where I brain dump ideas as they come to me, refine them, plan them, and then move them into what I need to do today. It holds personal work, hobbies, projects, loose thoughts, and commitments. It is how I track everything I am working on and keep my head clear.

Yesterday, I described what I wanted instead.

Claude wrote the application. It generated the code, structured the logic, and produced a working web app. It then helped me install it in a container on my personal NAS, lock it down properly, and expose it securely using a Cloudflare tunnel. I can now access it from anywhere.

It works.

I am a product person, not a professional software engineer.

That changes who gets to build, and how businesses operate.

---

Putting software into a business generally requires compromise.

Some large organisations build systems around their operating model. Most businesses adapt to the software they buy. They either shape their workflow to fit the product, or invest heavily to bend the product toward them. Both options carry cost. Financial cost. Time cost. Operational cost.

Products must generalise. Businesses operate in specifics.

Until now, the cost of building something specific was too high for a lot of businesses.

That balance has shifted.

---

There is an important nuance here.

Getting to a working version is now cheap. Keeping it robust as it grows is still engineering.

AI can generate a strong MVP quickly. But long-term maintainability, testing, security, clean architecture, and resilience still matter — especially once other people depend on the system. If I start layering features onto this app without structure, it will likely break.

Engineering does not disappear.

What changes is the sequence.

In the past, you needed engineering capacity before you could even test whether something was worth building. Now you can build, test, refine, and validate value before committing to formal engineering investment.

That shifts risk. It shifts capital allocation. It shifts speed.

And with AI evolving as quickly as it is, the boundary is already moving. Today it scaffolds version one. It will not be long before it scaffolds production patterns — testing frameworks, modular design, deployment structure — from the outset.

Engineering still matters. But the distance between operator and engineer is narrowing.

---

This is not about Trello.

Trello is a good product. It solves a clear problem. But my workflow is personal. It evolves. I think in certain patterns. I structure tasks in a particular way. I want automation to behave differently from how a general SaaS platform designs it.

In the past, that meant compromise.

Now it means describing what I want and refining it until it works.

The barrier between workflow and implementation has collapsed.

---

Because I own the application, I can now integrate AI directly into it in ways that suit me. Not as a plug-in bolted onto someone else’s roadmap, but embedded into the workflow itself.

The system is no longer static. It can evolve. It can assist. It can adapt.

For a small business, this could be enough. A focused, fit-for-purpose tool built quickly and owned entirely. That was rarely viable before.

The broader implication is economic.

If a single operator can build a working internal tool in hours:

• The cost of experimentation drops.
• The cost of validation drops.
• The dependency on subscription software weakens.
• The leverage of operators increases.

The question inside businesses shifts from:

“Should we subscribe to this?”

to

“Can we build this?”

That is a different conversation.

---

There is already plenty of noise about AI writing code. That is not what this is about.

This is about sequencing, risk, and control.

When version one is cheap, experimentation increases. When experimentation increases, business models adjust.

AI is not removing engineering.

It is compressing the path to engineered systems.

And that changes the build versus subscription equation.

---

I have built the MVP.

Now I will version it properly. I will look at how it scales, how features get added, where structure is needed, and what breaks when complexity increases.

That is the next phase.

If AI is already this capable at version one, the more interesting question is what version two looks like — and how quickly it arrives.