AI Adoption Accelerates, But Enterprise Readiness Still Lags
Industry estimates suggest that over 75% of enterprise developers are now using or planning to use AI coding tools, with a growing percentage of new code being AI-generated. Yet despite this rapid adoption, most organizations are still struggling to move from experimentation to production-ready AI workflows.
In this episode of AppDevANGLE, I spoke with Zvonimir “Z” Sabljic, Founder and CEO at Pythagora, about how AI is reshaping software development and why the real challenge is no longer generating code, but managing, governing, and trusting it at scale.
Our conversation explored the shift from prototype-driven experimentation to structured enterprise delivery, the rising importance of security and observability, and why internal applications are emerging as the most practical starting point for AI adoption.
AI Development Is Moving Beyond Code Generation
The market is quickly evolving past simple AI-assisted coding toward full lifecycle application generation. Instead of just helping developers write functions or snippets, platforms are beginning to generate complete applications, including front-end, back-end, and deployment workflows.
“Pythagora is basically a tool for building internal applications with AI… it will build the specification, the front end, the back end, and deploy the whole application,” said Sabljic.
This shift signals a broader transition from developer assistance to system-level automation. But while the technology is advancing quickly, enterprise adoption is still catching up.
“We’re in a transitioning period… from prototyping to actually AI coding something that goes into production,” Sabljic explained.
Internal Applications Are the First Scalable AI Use Case
Rather than attempting large-scale transformations, organizations are finding success by focusing on smaller, high-impact internal applications. These use cases provide a controlled environment where AI can deliver measurable value without the complexity of deeply integrated production systems.
“We’re focusing on smaller applications that have big utilization… like internal applications,” Sabljic said.
This approach allows teams to solve real business problems, such as onboarding workflows or internal automation, while building confidence in AI systems. It also reflects a broader market pattern: AI adoption is succeeding where scope is manageable and outcomes are clearly defined.
Security, Observability, and Guardrails Define Enterprise Adoption
As AI-generated code moves closer to production, trust becomes the central issue. Enterprises are no longer evaluating AI based on what it can generate, but on whether it can be safely deployed and managed.
“Security… is the number one priority for enterprises,” Sabljic emphasized.
However, security alone is not sufficient. Organizations also need visibility into how AI-generated systems behave in real-world environments. This includes understanding how applications interact with data, systems, and APIs.
“You really want to understand… what system it is touching, how it goes to the database, what queries it is doing,” he explained.
This introduces a new requirement: AI systems must be observable, explainable, and governed, not just fast.
Experimentation Is High, But That’s Not a Failure
High failure rates in AI projects are often framed as a problem, but they are also a signal of active exploration. Organizations are still learning where AI delivers value and where it does not.
“Having so many projects fail… that’s a good thing. It means people are willing to experiment,” Sabljic said.
This phase of experimentation is necessary for the market to mature. What matters is not avoiding failure, but learning quickly and applying those lessons to more targeted, high-value use cases.
AI Is Reshaping the Software Development Lifecycle
AI is beginning to impact the entire software development lifecycle, though not evenly. Some areas, like code generation and code review, are already seeing meaningful efficiency gains.
“You shorten that time… from five hours to one hour,” Sabljic noted, referring to code review improvements.
Other areas, such as testing and validation, are still evolving. The overall direction, however, is clear: development is becoming more iterative, faster, and increasingly automated. This shift moves the SDLC away from linear delivery models and toward continuous optimization.
The Biggest Barrier Is Not Technology; It’s Mindset
Despite rapid advances in AI capabilities, organizational readiness remains the primary constraint. Enterprises must build confidence in AI systems before they are willing to rely on them for critical workflows.
“How do we increase our confidence… so that at one point we can lift our hands off?” Sabljic said.
This hesitation reflects broader concerns around governance, accountability, and risk. Even as AI becomes more capable, organizations must adapt their processes, policies, and culture to support it.
AI Will Change Competitive Advantage
Looking ahead, AI is expected to fundamentally reshape how software is built and how businesses operate.
“I think… maybe 70% of the economy will flip… to AI-powered workflows,” Sabljic said.
While the exact scale may vary, the trajectory is clear. AI will accelerate development cycles, reduce manual effort, and enable new forms of automation across industries.
The competitive advantage will not come from experimenting with AI but from operationalizing it effectively.
Analyst Take
AI is no longer just enhancing development workflows; it is redefining them.
The industry is moving from isolated code generation tools toward fully integrated, AI-driven application delivery. But the limiting factor is not technical capability. It is trust.
Enterprises must ensure that AI-generated systems are secure, observable, and governed before they can be deployed at scale. This shifts the focus from productivity to operational discipline.
The most important takeaway is this: AI development is not just a tooling evolution; it is a systems and governance challenge.
Organizations that embrace this reality will move beyond experimentation and into production. Those that do not will remain stuck in pilot phases, unable to translate AI potential into real business outcomes. The next phase of the market will be defined not by who builds AI but by who can run it reliably.
