Automating the Wrong Thing Faster?

[#1]

AI agents can now turn tasks into code.

That’s super powerful.

But here’s the real question Agile teams should be asking:

Who’s making sure the tasks are actually good to begin with?

Because if the requirements are vague, misaligned, or incomplete…
you’re just automating the wrong thing, faster.

AI can accelerate execution.
But clarity, alignment, and thoughtful tradeoffs still begin with humans.

Before we automate delivery, we need to improve definition.

That’s where many teams get stuck — and where many AI tools fall short.

Curious to hear from this community:

How is your organization investing in better requirements and task definitions before layering on automation?

[Pálvölgyi János]

I think that's a very interesting question.
One possible direction is to try to define the requirements as precisely as possible, but very quickly we find ourselves in a waterfall type project.
I think that a properly functioning agile team, with two-week sprints and a high level of client representation at review meetings, should not be a big problem.
A problem can be there if the task has not been well sliced and the customer only meets more or less functionality at the end of development (although again, this is quite waterfall-like).
So if there is sufficient control from the client and a good agile team, I think the risk can be reduced to a manageable level.