Stop treating AI like a senior exec: it's just a brilliant intern

Five years ago developers were completely spellbound. Sitting in front of their code editors, watching GitHub Copilot suggest lines of code that we would have spent hours writing. It was like magic. We were mesmerised by the sheer possibility.

Then it got better: AI pair programmers were able to produce fully functioning websites in seconds. It was the dream. Fast forward to today, and that awe has largely curdled into a collective sigh. Let’s be real: we are exhausted by the AI noise. The endless product launches, the overlapping tools and the relentless hype have left us fatigued.

Yet, as the technology continues to accelerate at a breakneck pace, a massive gap is widening. Not in AI capability — AI can do more today than we ever anticipated — but instead there’s a gap in clarity.

We are building, deploying and restructuring on shifting ground, often without answering the foundational questions: What are we actually optimising for? How do we measure progress when the baseline moves every single week? How do our teams operate when workflows compress overnight?

Before we can successfully transform our engineering organisations, we need to stop reacting to the noise. We need to remember a fundamental truth that many companies are completely forgetting: at the end of the day, AI is simply a tool. It is software — damn good software, but just software.

It’s just a tool

Tools in the wrong hands always lead to mistakes. A tool requires people who possess deep understanding, domain context and clear goals to guide it. When you strip those people away the system stalls.

Imagine laying off all your experienced tradies, buying up the latest and greatest high-tech trade tools — all-electric, with all the fancy digital bells and whistles — and then hiring a bunch of random people off the street to use them. Would you honestly expect the resulting houses and buildings to be safe or high quality? Of course not. The tool is only one part of the equation.

Not every problem needs AI

To navigate this transition without breaking our culture or our codebases, we need a shared rubric for how to think about — and lead through — this change. I look at the challenge through five core lenses or SCALE:

• S – Stability: Striking the delicate balance between adopting bleeding-edge models and maintaining architectural stability, security and developer trust.
• C – Culture: Guiding humans through the natural anxiety of automation. It means rewriting the management playbook in real time, focusing heavily on psychological safety and clear direction.
• A – Architecture: Rethinking the developer experience from the ground up. We can’t just paste AI onto old, bureaucratic processes; we have to design entirely new ways of working.
• L – Leverage: Shifting the conversation from “What can we automate?” to “What actually creates leverage?” True value is about solving higher-order customer problems that were previously out of reach.
• E – Evolution: Moving past the chaotic ‘sandbox phase’ of ad-hoc prompt engineering and fragmented tools into predictable, governance-backed and scalable AI workflows.
   

Stop treating AI like a senior executive, when really, AI is a brilliant intern. It knows everything about the industry, but absolutely nothing about your company. Without the proper human guardrails, context and understanding of the problems you are trying to solve, AI will fail.

Redefining what scaling means

This framework brings us face-to-face with a radical reality: the efficient size of a development team is shrinking.

This isn’t happening because software developers matter less. It’s happening because each individual developer now drives exponentially more impact. When writing boilerplate, debugging legacy code and spinning up environments are compressed from days into minutes, the entire mathematics of engineering output changes.

The old leadership playbook was obsessed with scaling headcount to scale output: “We have a massive roadmap, so we need to hire 50 more engineers.” That playbook is officially obsolete. Today, the defining question for tech leaders is no longer “How do we scale teams?”. It has now become “How do we scale judgement and orchestration?”

When code generation becomes a commodity, the premium shifts entirely to critical thinking. We don’t just need people who can write syntax; we need leaders and engineers who can curate, validate and architect complex systems. We need developers who can see the big picture when the micro-details are being handled by algorithms. We need to go back to the core. Stop throwing AI at everything. Consider what you are trying to achieve, and then decide if AI is the right course of action. The magic didn’t disappear — it just became infrastructure. Now, it’s up to us to lead the humans who run it. *Michelle Duke is a Senior Developer Evangelist at Twilio.

Top image credit: iStock.com/ultramansk