Recently I’ve noticed a shift in how people think about development time.

AI tools have become part of our daily workflow. I use them myself, often! They’re great at generating boilerplate, helping me think through problems and speeding up small bits of work.

However, along with that, something else has started to show its head: an expectation that everything should now be faster. Much faster.

And honestly, from my perspective as a developer, that hasn’t quite matched reality.

Where the expectation starts

I’ve had moments where I generate a piece of code in seconds and think, “That would’ve taken me a lot longer before.”

And that couldn’t be more true.

But that moment, that quick win, is also the most visible part of the development process. It’s tempting to assume the whole task has been compressed into those few seconds.

In practice, that is rarely the case.

What the workflow looks like

Using AI hasn’t removed steps from the workflow. If anything, it’s shifted where the effort goes.

Understanding what needs to be solved.

Before even hitting “enter” on a prompt, the problem needs to be clear. If it isn’t, the output won’t be useful anyway.

Interpreting and adjusting the result

AI rarely gives something that can just be dropped in and shipped. There’s still a process of reading through the generated code, making tweaks and sometimes even rewriting parts completely.

Making it fit

Most development work isn’t isolated. It needs to fit within an existing system, follow established conventions and not break other features. That integration step still takes time.

Testing and edge cases

Generated code usually handles the straightforward scenarios. The tricky parts that cause problems… are still the developers responsibility to solve (with or without AI).

Taking responsibility for it

At the end of the day, the developer’s name is on the work. AI output can’t be treated any differently than code that was written by the developer themselves.

Where AI Does Help?

Just to be clear, I wouldn’t want to go back to a time without AI.

AI has made me faster in specific ways:

• Spinning up quick initial solutions
• Exploring different approaches
• Reducing repetitive coding

There are tasks that genuinely take half the time they used to.

The Part That Feels Off

What I’ve started to notice is a bit of a mismatch.

The faster parts of development are getting all the attention, while the slower, less visible parts are still very much there.

Sometimes, it feels like estimates are starting to reflect the best-case scenario rather than the real one.

That’s where pressure comes in.

The New Bottleneck (From My Perspective)

If I think about where I actually spend my time now, it’s not typing code.

It’s:

• Figuring out what the right solution is.
• Making sure it works in various environments.
• Double-checking that issues won’t be caused somewhere down the line.

AI helps with all of these, but doesn’t remove any of them.

So how should we think about it?

To me, AI feels less like a shortcut and more like a multiplier.

It makes me more efficient, but it still depends heavily on how I use it. It doesn’t replace the parts of development that require context, good judgment and accountability.

Final Thought

I’m really excited about where AI tools are headed.

But I think its important that our expectations evolve alongside them.

Faster code generation doesn’t automatically mean faster delivery. At least, not in the way it might seem at first glance.

From my experience, the real work is still all there – it’s just shifted slightly out of view.