Image generated by Deeptech Times using Google Gemini
How will enterprises deploy AI at scale without rebuilding decades of technology infrastructure?
According to Simon Ritter, deputy CTO at Azul, the AI era is not rendering enterprise software obsolete. Rather, it is creating new opportunities for organisations to unlock value from the vast repositories of business logic, applications and data they have accumulated over decades.
“The reality is that enterprises have huge amounts of data that they’ve built up over decades,” Ritter told Deeptech Times during a recent interview in Singapore. “What AI allows them to do is get a different view of that data and extract more value from it.”
AI’s hidden challenge: The enterprise data problem
Much of the public conversation around AI focuses on foundation models, copilots and autonomous agents. Yet for enterprises, the real challenge is considerably less glamorous.
Most organisations already possess enormous quantities of proprietary information spanning customer transactions, operational workflows, supply chains, manufacturing processes and institutional knowledge. The question is not whether they have enough data, but how to make that data accessible and useful to AI systems.
This is where Ritter sees the next phase of enterprise AI unfolding.
“Essentially, what you’re going to do is build a large language model around enterprise data,” he said.
Unlike consumer AI applications trained on public internet content, enterprise AI increasingly depends on proprietary organisational knowledge. For organisations whose mission-critical applications already run on Java, this presents a strategic advantage. Rather than constructing entirely new technology stacks, enterprises can leverage existing infrastructure to support AI-driven capabilities.
“How do we then access that and make it searchable and interrogatable using AI?” Ritter said. “That’s one of the key things we’re going to see.”
Why AI could drive a new wave of enterprise infrastructure demand
While AI promises greater automation and productivity, it also introduces significant infrastructure pressures.
Every AI-powered interaction requires computational resources to retrieve, process and analyse data. As enterprises deploy more intelligent applications and autonomous agents, demand for computing power will inevitably rise.
“We’re already seeing more and more demand for Java compute power,” Ritter observed.
This creates a challenge for CIOs already grappling with rising cloud expenditure. AI initiatives may generate business value, but they can also rapidly inflate infrastructure costs if not carefully managed.
Ritter argues that performance optimisation will become increasingly important as organisations seek to balance innovation with operational efficiency.
“What we’re really looking at is high-performance JVMs doing more with less,” he said.
In practice, this means extracting greater performance from existing infrastructure rather than simply adding more hardware. Higher throughput and lower latency can reduce the number and size of cloud instances required to support enterprise workloads.
“If we can have better performance for each node, it means you can provision a smaller node and a smaller number of nodes in the cluster and still deliver the same load-carrying capacity,” Ritter explained. “That reduces your cost because you’re reducing the cloud bill immediately.”
As AI workloads proliferate, infrastructure efficiency may become as strategically important as AI capability itself.
Trust, governance and security cannot be an afterthought
The enthusiasm surrounding AI has also heightened concerns around governance, observability and cybersecurity. Many enterprises remain cautious about deploying AI into production environments where reliability, compliance and security are non-negotiable.
For Ritter, one of Java’s enduring strengths lies in the maturity of its governance model. Because Java operates through the open-source OpenJDK ecosystem, security vulnerabilities are addressed through a collaborative framework involving major industry contributors including IBM, SAP, Red Hat and Azul.
“We need to have a very secure system because it is used in so many mission-critical enterprise applications,” Ritter said.
This collaborative approach has led to the establishment of the OpenJDK Vulnerability Group, which coordinates vulnerability disclosures, patch development and security updates across the ecosystem.
The result is a predictable and transparent process for maintaining enterprise systems: a characteristic that becomes increasingly valuable as AI applications move from experimentation to production.
For organisations evaluating AI adoption, governance may ultimately prove to be as important as innovation.
Software engineers will think more, type less
Perhaps one of the most contentious debates surrounding AI concerns the future of software development itself. With GenAI tools now capable of writing code, generating tests and debugging applications, questions inevitably arise about whether software engineers will remain essential. Ritter believes the role will evolve but not disappear.
He points to the growing popularity of vibe coding: the practice of describing an application in natural language and allowing AI to generate the code automatically. While useful for small projects, he remains sceptical about its suitability for complex enterprise systems.
“I don’t think that’s a realistic situation for enterprise applications,” he said.
Instead, developers will increasingly move up the value chain. Rather than spending most of their time writing individual lines of code, they will focus on architecture, logic design and problem solving while AI handles much of the implementation work.
“They’ll take a higher-level view of application development,” Ritter explained. “They won’t be doing so much of the low-level typing in lines of code. That will be done by the AI agent.”
The most valuable skill may therefore become the ability to define problems clearly and design systems intelligently rather than simply write code efficiently.
The limits of AI
Despite the excitement surrounding agentic AI, Ritter cautions against unrealistic expectations.
He acknowledges that AI agents will increasingly interact with enterprise systems, automate workflows and perform tasks previously handled by humans. However, he rejects the notion that organisations are approaching a future where human expertise becomes irrelevant.
“I think there’s an overestimation of what AI is capable of,” he said.
While AI excels at pattern recognition and automation, Ritter argues that human creativity, imagination and contextual reasoning remain fundamentally different.
“AI is really just about taking existing information,” he said. “Humans still have a lot to offer in terms of what they can do with AI tools.”
That distinction may ultimately define the next phase of enterprise transformation, where enterprises are successfully integrating human ingenuity with intelligent automation while leveraging the technology foundations they already possess.
