AI Dev 26 X SF | Barun Singh & Kennith Jackson; The Hidden Cost of AI Velocity and AI Agents

Why It Matters

Understanding AI's speed‑vs‑debt trade‑off helps leaders invest in tools and processes that boost real productivity without incurring unsustainable technical liabilities.

Key Takeaways

• AI accelerates some development steps but creates new bottlenecks.
• Rapid AI adoption risks accumulating technical debt over time.
• Newer models reduce code debt but aren’t a universal fix.
• “AI engineer” role is evolving; AI becomes a standard tool.
• Iterative AI assistance outperforms full spec‑driven autonomous agents.

Summary

The panel, hosted by Endela’s SVP of AI Solutions and its CPTO, examined the hidden costs of AI velocity and the hype around autonomous AI agents. They argued that while AI coding assistants like Copilot boost speed in isolated tasks, enterprise‑wide productivity gains remain limited because downstream processes—QA, code review, and decision‑making—still lag behind.

Barun Singh highlighted that the current surge in speed generates pockets of technical debt, warning that rushed releases in 2025 may require massive rewrites by 2026. He cited an experiment where newer models (Opus, Sonnet) solved a complex bug faster and with less debt than older versions, yet acknowledged that models can still loop endlessly on intricate codebases.

The speakers also debated the emerging "AI engineer" label. Singh likened it to early computer‑literacy titles that quickly became obsolete, suggesting AI will become a baseline skill for all engineers rather than a distinct role. They emphasized an iterative, human‑in‑the‑loop approach over speculative, fully‑automated spec‑driven development.

For businesses, the takeaway is clear: adopt AI to accelerate discrete tasks, but redesign workflows and governance to mitigate debt and preserve quality. Over‑reliance on speculative autonomous agents risks costly rework, while a balanced, iterative strategy can deliver sustainable productivity gains.

Original Description

AI coding agents have improved productivity while simultaneously putting engineers under pressure to do more, faster. But what's the cost that no one is talking about? Technical debt, 1000 lines of AI-produced code when 200 would do, and teams losing the ability to learn, to make the right call, and to reason as fast as the new systems that AI is creating.

Shipping worse code with agents is a choice. We can instead choose to ship better code. How can developers evolve to make agentic work for them? Discipline. Self-awareness. Don't ship what you don't understand, and implement processes and systems to keep a human in the loop in ways that create not just more, but the most optimal product.

In this session, Barun Singh, Chief Product and Technology Officer at Andela, home of a 150,000-strong tech talent marketplace, and Kennith Jackson, Andela's SVP, AI Solutions & Operations, will speak to challenges and opportunities facing developers and how they can become agentic engineers to deliver best-in-class products at pace.