AI is already in the stack. The hard part is making it work inside real systems without adding risk or slowing teams down. And few understand this better than Alex Ter-Zakhariants, Field CTO at Sphere. In a recent episode of SphereCast, Alex sat down with host Adin to unpack the challenges and opportunities AI presents for modern engineering teams.

From confronting the weight of technical debt to building scalable, AI-ready infrastructure, the conversation offers a grounded, pragmatic view into what’s hype and what’s real.

Taming the Beast: AI and Technical Debt

Technical debt isn’t just about bad code; it’s a business bottleneck. According to Alex, most organizations accumulate tech debt because they prioritize speed over long-term architectural soundness. This leads to patchwork systems that are hard to maintain, scale, or modernize.

“You have to understand why companies have tech debt in the first place,” Alex explains. “It’s not just messy code. It’s poor processes, a lack of governance, and short-term thinking.”

AI, when applied properly, can help identify and even remediate technical debt. Many modern platforms offer AI-powered tools that analyze codebases for inefficiencies and suggest modernization paths. But Alex warns: tools are only part of the solution.

“There are no shortcuts. AI helps—but without the right engineering practices, it’s just lipstick on a pig.”

From Data Swamps to Data Lakes: Making AI Work for Big Data

Organizations that deal with high volumes of user-generated content often struggle with content moderation, relevance, and governance. The problem is rooted not in the lack of AI tools, but in the disorganized nature of their data environments.

“What were once data lakes have become data swamps,” Alex notes. “Without structured, clean, and governed data, AI models simply can’t operate effectively.”

To fix this, companies must invest in data modernization — consolidating datasets, establishing strong governance, and creating abstraction layers that allow AI to process information coherently.

He offers practical guidance on creating “data-as-a-service” layers, where different personas have tailored access to datasets, enhancing both performance and security.

Scaling for the Unexpected: Avoiding the Pitfalls of Poor Planning

Scalability is another major concern for today’s tech leaders. While cloud infrastructure provides “scale-on-demand,” it’s ineffective without an application architecture designed to take advantage of it.

Alex shares a compelling example from the telco industry: the annual iPhone release. Telco websites experience massive spikes in traffic, and if not properly prepared, companies risk losing both revenue and reputation.

“If Verizon.com crashes, consumers will head straight to AT&T.com. That’s millions in lost revenue.” 

AI-powered AIOps tools like Dynatrace and New Relic can model system behavior under load, detect anomalies in real time, and automate remediations. These platforms are critical for spotting bottlenecks before they become incidents.

AI as a Dev Co-Pilot: Productivity Without Complacency

AI-enhanced developer tools—from code generation to auto-documentation—have entered the mainstream. But do they live up to the hype?

“Think of AI tools like giving someone a Ferrari,” Alex says. “It’s fast, but if you don’t have a driver’s license, it’s useless.”

The tools can accelerate workflows, but they’re not a replacement for foundational knowledge. Developers still need a strong grasp of software engineering principles. Without it, relying on AI can be dangerous and counterproductive.

“We’re far from a world where English input equals flawless machine code. Developers still need to know what they’re doing.”

Future-Proofing Enterprise Applications: Back to Basics

So how should tech leaders prepare for an AI-driven future?

Alex’s answer is both simple and profound: go back to basics. Clean, modular code. Encapsulation. Proper abstraction layers. These time-tested principles not only reduce technical debt, they also make applications easier to adapt as AI capabilities mature.

“There’s no such thing as bulletproofing your system for the future. But you can make it adaptable to change.”

When integrating AI into product roadmaps, leaders should focus on real use cases. For instance, a financial forecasting app might use predictive AI to simulate market trends based on real-time data. But these models still require context and oversight.

“No matter how good your prediction is, it won’t guarantee an outcome. Just like stock investing—past results don’t guarantee future performance.”

Final Thoughts: AI Won’t Replace You, But It Will Empower You

The episode closes with a reminder that AI isn’t a silver bullet—it’s a powerful ally. Organizations that build a solid foundation in engineering best practices, data hygiene, and scalable architecture will be the ones best positioned to harness AI’s full potential.

Alex’s advice? Invest in your teams, your processes, and your data. The rest will follow.

“AI is here to stay. But if you want it to work for you, you have to do the work first.”