
The Shape of Software Is Changing
I keep coming back to the same thought: AI is not just changing how we build software. It is changing what software is for.
For a long time, making an app usually pointed toward the same destination. You had an idea, built a product, polished it, and hoped it could become a company. Even small tools were often treated like early versions of a startup. That path still exists, but it no longer feels like the default.
Now the cost of making software is dropping fast. You can sketch a workflow, prompt your way into a prototype, wire up a few APIs, and suddenly something real exists. Not perfect, not always durable, but useful enough to solve a problem. Once that becomes normal, behavior changes. People make more software for themselves. Teams spin up more throwaway tools. Businesses still buy software, but they get more selective about what deserves a real platform versus what can be generated, assembled, or replaced.
I do not think the future of apps is one big category anymore. It feels like it is splitting into at least three, and that split says a lot about where software is going next.
Personal Apps
The first category is personal software, the kind of tool you make for yourself: a planner that matches how your brain works, a little dashboard for a side project, a writing assistant tuned to your tone, a research tool for one weird obsession, a personal CRM, or a tiny app that helps you track routines, organize references, or clean up some repeating annoyance in your week.
This kind of software used to be rare because the effort was too high. Most people just bent themselves around generic tools instead. You picked the least-wrong app and lived with the mismatch.
AI changes that because it is becoming realistic for people to make software that fits their own habits instead of adapting their habits to software somebody else designed for the average user.
That feels important to me because so much software has always been a compromise. The future may be more personal, more oddly shaped, and more specific. Not every app needs a thousand users, and some of the most meaningful software in the next few years may have exactly one. The benefit is not scale so much as fit.
I feel this pretty directly in my own life. Part of why I keep thinking about personal software is because I am making my own productivity tool, something that pulls together notes, time, tasks, files, and the other pieces of how I actually work. I have tried enough existing products to know that none of them quite fit the shape I want. They each get part of it right, but not the combination, not the flow, and not the feeling of having one place that really matches how I think. So instead of waiting for that product to appear, I want to craft it as my own software.
That impulse seems increasingly common now. More people are reaching the point where they do not just want better software in the abstract. They want software that feels like theirs. Not branded for everyone. Not flattened into the average use case. Just something that actually fits.
Personal apps can be messy. They can be overbuilt for tiny problems, they can break, and they can be abandoned. But they can also give people a level of agency we have not really had before. You do not have to wait for a company to decide your problem matters enough to put on a roadmap. You can just make the thing, use it, and see if it earns a place in your life.
Disposable and Internal Tools
The second category is what I think of as disposable software, though I do not mean that in a dismissive way. These are internal tools, one-off workflows, team dashboards, event apps, migration helpers, support copilots, temporary research tools, and all the small systems a company needs but never wants to treat as a polished product. In a lot of cases, this is software for a quarter, software for a project, or software for a process that may not even exist next year.
There has always been a huge amount of this kind of work inside companies, but it was often too expensive to justify doing properly. Teams either lived in spreadsheets and duct tape, or they waited in line for engineering help that never came.
In a way, we have already seen early versions of this. Tools like Airtable and Retool gave teams a way to assemble workflows, dashboards, and internal systems without treating every need like a full product build. They were a sign that companies wanted more flexible, in-between software for operational problems that were real but not glamorous.
This is where AI may have one of its biggest practical effects. A lot of internal software does not need a beautiful brand, a multi-year roadmap, or perfect architecture on day one. It needs to work well enough, connect to the right systems, and save people time. If a team can generate a useful internal tool in days instead of months, they will. If they can replace a brittle manual workflow with a narrow app that only ten people ever touch, they will, and they should.
What feels different now is how low the threshold has become. You no longer need to stay inside the boundaries of a no-code product or wait for somebody else to package the right building blocks. It is getting easier to just spin up the thing you need, shape it around the workflow in front of you, and move on when the need changes.
I think we will see more software that is intentionally temporary. Built fast, used hard, then retired. That is not failure. That is software behaving more like the specific tool it really is.
There is something healthy about that. Not every solution needs to become a permanent platform. Some problems deserve a screwdriver, not a company, and AI makes it much easier to build the screwdriver and move on.
Commercial Software Gets Harder
The third category is still commercial software, and this is the hardest one. This is the software built to serve many customers, justify a price, support a business, and keep improving over time. It is the category most of us have historically centered when we think about building products, because that is where the startup story lives. It is also where the pressure is highest now, because AI makes this lane much more demanding.
If people can build more for themselves, and if companies can generate more internal tools on their own, then a commercial product has to clear a much higher bar. It cannot just be a bundle of basic features anymore. It cannot win only because the old way was hard to build.
That is where the SaaS conversation starts to get uncomfortable. A lot of products were valuable not just because they were great, but because they were available. They packaged up a workflow, put a clean interface on top, and sold convenience into a world where most customers could not realistically make something for themselves. That was real value, but it was also value tied to scarcity.
AI starts to chip away at that scarcity. If the customer can get eighty percent of the way there with a generated tool, a lean internal app, or a custom workflow stitched together in a week, then a lot of middling SaaS products are going to feel exposed. Not dead, but exposed. The parts that were once enough, clean CRUD, a few automations, some dashboards, a layer of collaboration, may stop being enough on their own.
That does not mean commercial software goes away. It means the reasons to pay for it become more specific, more defensible, and more earned. It also means some companies are going to have to face a hard truth: they were not really building deep product moats, they were renting access to a problem that was temporarily too annoying for customers to solve themselves.
The moat is probably less about having features at all and more about having the right combination of trust, distribution, deep workflow understanding, reliability, data, ecosystem, service, and continued refinement. In some categories, the moat may be compliance. In others, it may be network effects, brand, proprietary data, or simply years of compound product quality around a mission-critical workflow.
The hard truth is that AI may strip away the illusion that every app idea deserves to become a venture-backed company. Some ideas are better as personal tools. Some are better as internal software. Some are real businesses, but only when they solve a painful enough problem at a level generic generation cannot touch, or when they do the kind of ongoing heavy lifting a customer really does not want to own.
That is not bad news. It is just a more honest market.
Some of the sharper writing around this has landed in a similar place. Joe Reis put it bluntly: software is becoming easier to build and harder to justify renting. That feels directionally right to me, especially for SaaS products that were mostly packaging convenience. And investors like Justin Kaufenberg are making a related point from the other side: the winning commercial products may not be the ones with the flashiest model, but the ones with the deepest context and strongest hold on real workflows.
Why We Need All Three
I do not think one of these categories replaces the others. If anything, they make each other more necessary. Personal apps give individuals more freedom and let people shape tools around the way they actually live and think. Disposable and internal tools give teams more speed and reduce the pressure to overformalize every operational need into a polished software product. Commercial software still matters because businesses need stable systems they can trust, buy, support, secure, and build processes around. Most companies do not want to generate their own payroll system, healthcare platform, accounting stack, or design tool from scratch every other month, because at some point durability matters more than flexibility and trust matters more than novelty.
The future is not cleaner because of this split. It is probably messier. But it is also more realistic.
We should want a world where people can make their own small tools. We should want teams to build narrow internal apps without turning each one into a product ceremony. And we should still want strong commercial software where the stakes are high, the workflows are deep, and the trust requirements are real. The categories may blur at the edges, but the underlying needs are different.
Those are different jobs. It is okay that they produce different kinds of software.
The Bigger Shift
What feels new is not just that software is getting easier to make. It is that software is becoming more disposable in some places and more demanding in others.
That changes the posture of building. A personal app only really needs to fit one person well enough to matter. An internal tool needs to solve a narrow problem cheaply and quickly enough to justify making it. But commercial software has to answer a much harder question, which is the one I keep coming back to most: why should this exist as a company at all, and why should someone keep paying for it once building gets cheaper?
In the age of AI, I do not think the best commercial products will win because they merely exist. They will win because they handle complexity people do not want to own themselves, save time at a scale that matters, and are trusted in moments where failure is expensive. They will win because they understand a workflow so deeply, and support it so reliably, that a quick generated substitute still feels thin by comparison.
AI lowers the cost of making software, but it also raises the burden of being worth paying for.
Closing Reflection
Maybe that is the picture to hold onto. The future of apps is not one lane but a spectrum. Some software will be deeply personal, some will be temporary and tactical, and some will still become lasting companies. The mistake is assuming they should all be judged by the same standard, or built with the same ambition.
If I make a small app for myself that helps me think better, that can be enough. If a team spins up an internal tool that saves twenty hours a week for six months, that can be enough too. And if a company builds commercial software that earns trust, handles complexity, and becomes part of how a business runs, that can still be more than enough.
We do not need fewer kinds of apps. We need a better understanding of which kind we are building, why it deserves to exist, and what success actually looks like in that category. AI is not flattening software into one thing. If anything, it is finally letting software become what it probably always wanted to be: sometimes personal, sometimes disposable, and sometimes durable enough to build a business on top of.