# AI Makes it Harder to Say No!

As AI coding agents have gotten better, I am catching myself taking on work I would previously have pushed back on. With an AI coding agent, “that’s too difficult” or “that’s too costly” feels less convincing. Cost and complexity used to be an easy gatekeeper when considering what work we “should” be doing.

For an Oracle APEX developer, that is exciting. Work that takes us beyond familiar SQL and PL/SQL feels more approachable. But we still need to ask the same question:

> Is this actually a good requirement?

Just a few weeks ago I was asked to change the standard APEX required field indicators because users were finding it difficult to distinguish the red triangle in the corner of the field/column. A year ago I would have pushed back on this (I am not the greatest CSS developer). With APEXlang and Codex, it was easy and I just made the change. Looking back, I should have asked how often users were missing required fields and what problems that caused before deciding whether the change was worth making.

Even if the code is quicker to write, we still own the testing, support, and maintenance.

Before saying yes to a new requirement, I want to know:

- What problem are we solving, and for whom?
- Does the application or business process already provide a reasonable answer?
- Does the benefit justify the ongoing responsibility?

> Of course, these are all things we should be doing anyway.

The challenge now is that our “easy out” of saying it’s too hard or costly has become harder to rely on. As software developers, we must up our game and do our part in making sure these questions are answered before we start development.

“I can build it” has become easier to say. “We should build it” still needs a reason.

## Bonus Tip

I have also found it useful to share page view counts and feature usage statistics with clients. Seeing which features actually get used gives them perspective and encourages more thoughtful conversations about future requirements.
