AI Application Development: Merrick Reveals How to Make it Enterprise-Ready

AI is dramatically accelerating application development, making it easier than ever to turn ideas into working software. But building an application and preparing it for enterprise use are two very different things. A solution that works perfectly well as a prototype may face significant challenges in a business-critical environment, where data security, regulatory compliance, infrastructure reliability, and scalability are non-negotiable.

For CTOs, this presents a critical balancing act: enabling rapid innovation while ensuring the underlying enterprise infrastructure can support the demands of real-world business operations.

In this interview, Phillip Merrick, co-founder and CEO of pgEdge, shares his perspective on these emerging challenges. He discusses why infrastructure decisions must be made early, how organizations can balance innovation with risk, and why clear standards for AI tools and database platforms are essential. He also shares his perspective on the evolving landscape of AI application development and how technology leaders can lay the foundations for sustainable innovation.

When AI-Built Apps Outgrow Enterprise Infrastructure

AI application development is democratizing software creation, enabling business users to solve problems independently of IT. But rapid creation does not guarantee enterprise readiness.

AI can help teams build applications faster than ever. What happens when those applications start outgrowing the platforms they were originally built on?

Merrick: AI is not only helping teams build applications faster. But it is also helping non-IT staff build applications to solve specific business problems that previously they needed IT to build for them. In many cases, these AI-generated applications are being developed on cloud-based platforms that are not enterprise-ready. As a result, it lacks the security, compliance, and reliability the enterprise demands. To meet these requirements requires re-platforming on compliant infrastructure, which might be on-premises. And enterprise data often can’t leave the enterprise.

Building for Scale Starts with Security and Compliance

Infrastructure decisions cannot be an afterthought. Organizations must understand their security, compliance, and data requirements before choosing where AI-built applications should run.

What infrastructure decisions should engineering teams make early instead of waiting for scale problems?

Merrick: Engineering teams need to understand the security and compliance requirements of applications built with AI. They must also consider the data these applications will access or create. Based on this understanding, a suitable choice of platform can be made. Be it hosted in a vendor-managed cloud, the enterprise’s own cloud account, or on-premises.

When AI Agents Make the Unexpected Decision

As AI agents take on more complex tasks, organizations must understand the risks of incorrect decisions and their potential business impact. The challenge is knowing when AI can act independently and when human oversight is essential.

What happens when an AI agent makes a decision that a human engineer never anticipated?

Merrick: The impact depends on what the application is used for. And how serious the consequences of a wrong decision could be. When applying AI agents to any problem domain, the probability of an agent error needs to be understood through modeling and testing. A business decision then needs to be made as to whether, say, a 2% error rate is acceptable. This may or may not compare favorably with how a human performs the task.  If only a zero error rate can be tolerated, then it may be necessary to include a human in the loop for review. Or alternatively, build a software for deterministic automation – although that itself could be built with an AI coding agent.

Enabling AI Innovation Without Creating Shadow IT

As AI application development becomes more accessible, employees may turn to unapproved tools and platforms. CTOs must give teams room to innovate while maintaining enterprise-wide standards.

How should CTOs balance the pressure to adopt AI quickly? How can they ensure infrastructure supports it?

Merrick: CTOs should provide their organizations with lists of approved AI coding and database infrastructure platforms. One that allow their organizations to move quickly while meeting requirements for security, compliance and reliability. Without this staff will likely choose tools and platforms that don’t meet these requirements, creating both a shadow AI and a shadow IT problem.

The Next Five Years: Enterprise Infrastructure Must Evolve with AI

As AI tools empower both professional developers and non-technical employees, enterprise governance will become increasingly important. The challenge is to establish standards that keep pace with innovation without stifling it.

Looking five years ahead, how do you see the relationship between AI, application development, and database infrastructure changing?

Merrick: With such unprecedented rapid change it is hard to project out one year, let alone five. However, we do see some emerging trends. We already know that for AI to deliver value it needs carefully governed access to corporate data. We also know that AI coding tools have made it possible for non-IT staff to build applications, and for engineers to be far more productive. These trends will only continue to accelerate, requiring CTOs to work with their organizations on acceptable standards and policies across each of AI, application development and database infrastructure.

Key Takeaway: The Future of AI Depends on More Than Building Faster

The rise of AI application development calls for a shift in how organizations think about software ownership, accountability, and long-term growth. As more employees build applications and AI agents take on increasingly complex tasks, CTOs will need to ensure that innovation is supported by clear policies, responsible decision-making, and infrastructure that can evolve with changing business needs.

Ultimately, the goal is not simply to keep up with the pace of AI, but to use it wisely. Organizations that combine experimentation with sound technical judgment will be better equipped to adapt, scale, and create lasting value. The future will belong not just to those who build with AI, but to those who know how to make what they build work, safely and sustainably.

Point to Remember: Treat Every AI-Built Application as a Long-Term Responsibility

The ability to generate applications quickly could leave organizations with a growing portfolio of software that nobody is fully responsible for maintaining. CTOs should therefore establish an application lifecycle strategy that covers ownership, documentation, security updates, performance monitoring, and eventual retirement.

This is essential to making AI application development sustainable beyond the initial build.