Every few months, the tech industry declares software engineering dead. First it was no-code. Then low-code. Now it’s AI.
ChatGPT writes code. Copilot auto-completes entire functions. Devin supposedly runs entire sprints. The narrative is compelling especially for people outside the field who see AI produce working Python scripts in seconds and assume the rest of the job is more of the same.
I have spent four years building production financial systems at scale, working across distributed payment infrastructure serving millions of users. In that time, I have used AI tools extensively. My conclusion: the engineers who fear AI are often engineers who never fully understood what their job actually was. For those who do understand, AI is the most powerful amplifier we have ever been handed.
The Data Tells an Interesting Story
Look at that chart. 84% of developers now use AI tools in their workflows, up from 76% last year. But trust in AI accuracy dropped from 40% to just 29% in the same period. More usage, less trust. That’s not a contradiction. That’s experience talking.
This pattern might appear contradictory at first, but it reflects something familiar to experienced engineers. Developers are increasingly using AI tools because they are useful for accelerating parts of the workflow. However, the more developers rely on them, the more they realise that AI output still requires careful verification. Usage is increasing. Blind trust is not.
Coding was never really the job.
There is a persistent misconception, held even by some engineers early in their careers that software engineering is primarily about writing code. It is not. Code is the output. The job is systems thinking.
Software engineering is fundamentally about decisions, not code. What should we build? How do different components interact? How does the system behave when things go wrong?
Working on distributed financial infrastructure, the challenge is never simply ‘can I write the code?’ The challenge is: What are the failure modes under load? How does this interact with the payment settlement layer when a network partition occurs mid-transaction? What is the correct behaviour given regulatory constraints in multiple jurisdictions?
These questions require deep contextual understanding built through years of watching systems fail in unexpected ways. No AI model, as of today, can walk into a mature financial codebase, absorb four years of undocumented architectural decisions made under operational pressure, and make safe, high-stakes design choices. That remains the domain of experienced engineers.
AI doesn’t know your system. It doesn’t know the three architectural decisions you made under pressure two years ago that everything else now depends on. That context lives in the engineer and no prompt can replace it.
What AI Actually Changes in Day-to-Day Work
When I started using AI tools seriously, I expected to get faster. I did. But something I didn’t expect, I got wider.
I started exploring parts of the stack I’d normally avoid. I prototyped in an afternoon what used to take a week. Less time on boilerplate, more time on the problems that actually matter.
But I also caught AI being confidently, dangerously wrong. Suggestions that would have introduced race conditions in live financial transactions. Logic that looked clean on the surface but collapsed under edge cases our system deals with every day. The AI had no idea it was wrong. I did, because I understood the system.
That’s the line. The tool is only as powerful as the engineer holding it.
The Honest Conversation the Industry Needs
Not every engineer is safe and false reassurance helps no one.
Most developers still don’t see AI as a threat 64% in 2025. But that number has been slipping every year, down from 70% in 2023. And look at the 2025 breakdown: 21% are now unsure. That’s a fifth of the global developer workforce sitting on the fence. That uncertainty is a signal worth paying attention to.
If your value is purely writing code fast without deep understanding of why, then yes, AI competes with you. But if you understand systems, design for failure, lead teams through ambiguity, and translate complexity for non-technical stakeholders? AI doesn’t replace you. It makes you extraordinarily more powerful.