Developers promoted to architects without notice

A lone figure stands at a crossroads in a vast, half-built digital landscape — one path lined with glowing but tangled wires, the other clear and minimal. They wear a developer’s jacket but hold an ar

A developer’s role is shifting — not because of corporate restructuring, but because AI now writes most functional code. Yet few have been told they’ve effectively been promoted into architectural roles. Without guidance, teams ship slow, insecure, or overcomplicated systems, not due to incompetence, but from misaligned responsibilities.

One real-world case illustrates the stakes. A Google Forms add-on slowed to a 15-second load on its welcome screen. The cause? A developer used AI to implement logging that attempted to fetch user emails and, failing that, user IDs — each wrapped in try/catch blocks. Each attempt triggered multiple API calls to Google’s servers, all executed before the UI rendered. The fix was simple: log only the timestamp of installation and show the welcome screen immediately. But AI didn’t suggest this. It followed instructions to the letter.

Why AI can’t be the architect

AI excels at translating requirements into code, but it lacks context. It doesn’t know whether an app has 10 users or 10 million. It won’t question whether logging an email on install justifies a 15-second delay. It assumes every requirement must be fulfilled exactly as given. This sycophantic precision leads to bloated, slow implementations.

Moreover, AI prioritises security and backward compatibility by default. It avoids unauthenticated payloads, even when a temporary, low-risk async call from the browser could work. It creates redundant database tables to preserve old logic, inflating complexity. These are architectural trade-offs — decisions that should involve business context, user experience, and long-term maintainability. AI cannot make them. Only a human can.

What architects do that AI cannot

An architect doesn’t just design data models. They push back. They ask: Is this feature worth the cost? Can we defer part of it? Can we decouple the tracking from the critical path? In the add-on case, the architect would have told marketing: You won’t get emails at install time — not because we can’t, but because we won’t slow down the product.

They also enforce constraints. Instead of letting AI reinvent the schema, the architect defines the data model first — what fields exist, how they evolve, when to deprecate old versions. Then, AI operates within those boundaries. This prevents the sprawl of half-used tables and duplicate logic that plagues AI-generated backends.

The unspoken promotion

Many developers now in architectural roles don’t realise it. They miss coding — the craft they trained for — only to find AI doing it faster. But their real value is no longer in writing lines, but in shaping systems. The confusion and frustration around ‘vibe coding’ stem from this gap: people are doing architecture without the title, training, or clarity.

This shift matters most in fast-moving startups and scale-ups in the UK, UAE, and Kazakhstan, where lean teams rely on AI to ship quickly. Without deliberate architecture, they risk technical debt that scales with user growth. US enterprises face similar risks, but often have formal roles to absorb the change.

What it means

Businesses must acknowledge this promotion. Developers need training in system design, stakeholder negotiation, and constraint-based AI use. Companies should stop measuring output in lines of code and start valuing trade-off decisions. Architects — whether titled or not — must own the data model, set performance budgets, and say no.

For founders and tech leads, the message is clear: don’t let AI decide your architecture. Define it first. Use AI as the implementer, not the designer. Review its output not just for correctness, but for alignment with performance, cost, and data control.

The coding job is shrinking. The architect’s job is expanding. The promotion has happened. It’s time to accept it.

Sources

Leave a Reply

Your email address will not be published. Required fields are marked *