The age-old debate of build vs buy has been redefined with the availability of AI development tools. I don’t see this necessarily as a revolution but rather an evolution. For decades developers used traditional coding languages like Java, Cobol, or C++. While new languages like Python and Ruby were added it was still the same approach. This created a high barrier to entry for software development as coding skills demand outweighed supply.
The introduction of low code tools like Quickbase, Kissflow, and PowerApps helped democratize software development to a point, but it was still mostly professional developers that were using these low-code tools. The term “citizen developer” was so popular because it promised the idea of creating software without relying on scarce and expensive development resources.
As we now know, citizen development never really caught on in the enterprise. It worked well for creating simple apps that didn’t need things like integrations, or security, but it didn’t meet the standards when it came to putting enterprise-grade applications into production.
I feel like that is where we are now with AI development tools like Claude and Codex. Much like low-code tools, there is still code being generated (lots of code) behind the scenes, but the author doesn’t need to know the intricacies of how it is being generated. This has allowed almost anyone to enter a prompt and generate a working application in minutes.
However, just like with low code tools before it the question is:
“would you put this into production?”
Based on what I’m seeing, the answer is no. Estimates vary, but a recent article I read by Daniel Balcarek does a good job of breaking it down. Most AI generated code is used for what he calls “Hobby Projects”. These could be considered personal productivity tools and dashboards for individual use. These apps typically run on the users local laptop, have no security, and would struggle to scale to multiple users.
Sometimes these applications do make it into production, often without IT oversight (i.e. shadow IT), but for the most part, you still need enterprise infrastructure for enterprise applications.
This is where Pega comes in. Just like AI coding tools, Pega generates code behind the scenes. The big difference is how that code is optimized for enterprise applications.
Enterprise Ready Out-of-the-Box – Pega is built for applications that can’t afford to fail. These applications are secure, compliant, and can scale to millions of users from day one.
Explainable - Pega’s model-driven approach isn’t a black box. You’re working with a model of how the application behaves, as users experience it, not buried within millions of lines of AI generated code.
Predictable - Pega delivers predictability by embedding auditability and orchestration directly into the workflow—so every action, decision, and outcome is visible by design. Instead of observing AI behavior after the fact through bolted-on control towers, Pega defines and governs that behavior within the workflow itself.”
AI development tools are changing how software is built, but what hasn’t changed is the demands of enterprise software. Generating code is easy, but maintaining it and running it at scale is the real challenge. That’s why the conversation has shifted from build vs buy to build vs run. That is why Pega is built for applications that have to run predictably, reliably, and with full transparency.