AI Makes Apps Cheap. Platforms Are Still Hard.
I recently heard someone say on a podcast, “AI makes yesterday’s competence cheap.”
That line stuck with me.
Writing decent code, creating mockups, building internal apps, drafting strategy documents, producing analyses and reports. Tasks that once signaled expertise can now be produced in minutes with a prompt. That changes organizational dynamics fast.
I am experiencing this personally right now. People across organizations are suddenly building software. Roles that previously depended on centralized software teams can now generate working applications themselves. A prompt and a weekend can produce something that looks surprisingly close to software.
At first glance, this feels like the collapse of traditional bottlenecks. Why wait on a software team when AI can help you build the tool yourself? That question is starting to show up in a lot of organizations.
And if I am being honest, there is also a human side to this that people do not talk about enough.
Is there a twinge of jealousy and identity crisis for people who spent years developing technical skills that can now be partially replicated with a prompt? Yeah. Absolutely.
People built careers around hard-earned competence. Around being the person who knew how to build the thing others could not. AI changes that emotional equation whether we want to admit it or not.
But there is a difference between generating software artifacts and building systems, and I think many companies are underestimating how important that distinction is.
Before AI, organizations accumulated spreadsheets, access databases, scripts, macros, and dashboards scattered across desktops and shared drives. Individually useful, collectively chaotic. They solved local problems but rarely strengthened the organization as a whole because they were disconnected from each other, disconnected from shared workflows, and disconnected from long-term operational thinking.
AI risks accelerating the same pattern at a much larger scale without thoughtful systems thinking.
Now instead of spreadsheets and macros, organizations may accumulate:
- Claude apps
- Cursor apps
- copilots
- workflows
- lightweight tools
- generated interfaces
Many of these applications will absolutely provide value to the person or team creating them. That is not the issue. The issue is that disconnected applications remain disconnected applications, even when they are generated with far more sophistication than the tools that came before them.
This is the tension many organizations are experiencing right now. Domain experts can build things faster than ever before, and that is genuinely valuable. But speed of creation and organizational capability are not the same thing. A collection of disconnected apps does not automatically become a platform any more than a folder full of spreadsheets becomes an operating system.
The challenge is no longer whether people can build software. They clearly can. The challenge is whether organizations can turn all of this new creation into systems that compound instead of fragment.
That requires a very different kind of thinking than simply generating applications quickly.
When people look at mature software platforms, they often see slowness. A domain expert can generate an internal tool in days while a platform team may take months to release something similar. From the outside, this can feel inefficient or bureaucratic. The generated app works. The platform team is still discussing architecture, permissions, workflows, integrations, scalability, and operational concerns.
But those concerns are the work.
Mature platform teams are solving problems the generated application often ignores:
- identity and access management
- shared data models
- multi-tenancy
- auditability
- security
- integration patterns
- reliability
- operational support
- versioning
- scalability
- user consistency
- governance
- long-term maintainability
These are not side concerns or unnecessary process overhead. They are the reason organizational software survives beyond the first enthusiastic demo.
A generated application may work perfectly for a local use case. But working is not the same thing as compounding. A platform creates shared capability across teams, workflows, data, and operations. That takes systems thinking. It takes coordination. It takes intentional architecture.
That is the shift many organizations are struggling to understand. AI compresses the cost of execution, but it does not compress the complexity of systems thinking. In many ways, AI may actually increase the importance of architecture because organizations can now create disconnected software artifacts at an unprecedented rate.
For years, technical competence itself was the differentiator. Knowing how to code was valuable because relatively few people could do it. Now AI can produce large portions of what used to represent technical skill. Implementation alone becomes less differentiating over time.
But when competence becomes abundant, organizational judgment becomes more important.
Knowing:
- where something belongs
- whether it should exist at all
- how it connects to existing workflows
- what data model it depends on
- what breaks because of it
- whether it creates long-term capability or short-term clutter
This is why AI does not eliminate the need for platforms. It increases the need for them. Without platforms, organizations risk creating the modern equivalent of thousands of spreadsheets scattered across shared drives. Useful individually. Fragile collectively.
AI can generate unlimited software artifacts.
The hard part is building platforms where those artifacts compound instead of fragmenting the organization.