[skydhash]

It is micromanagement. If the job are not being done, the best way is to investigate what current practices are blocking people from doing it (the answer is probably meetings and bad communication). The worst way is to present a tool as a silver bullet for tasks you’re not doing and not accountable for.

[mh-]

Where am I presenting the tool as a silver bullet? You seem to be confusing me with someone else in this thread, or making the mistake of turning this into a polarized conversation of "AI is a panacea" vs "AI is worthless".

I engaged in the thread in good faith, and am transparent about what I'm doing and why. I also clarified that part of the job in my org is experimenting with these tools.

[skydhash]

The complaint in the thread is that management is forcing AI tooling usage. If part of your job is to experiment with these tools, then like any experiment, the correct way is to share the findings with a report detailing the methodology and findings. But no one is doing that AFAIK. It’s all superlatives.

[sarchertech]

Have you ever monitored and encouraged the use of a particular text editor or IDE? If you had an employee whose manager thought was a high performer, but you noticed they used notepad would you encourage that they regularly give vim a try?

The reason we force people to use Jira is because it only works if everyone uses it.

AI doesn’t work like that. If it does enhance productivity 50% then use will spread and the expectations of your line managers will naturally go up and the holdouts won’t be able to keep up.

Or only the exceptional ones will. And in that case why do you care how they do it?

[tetromino_]

> The reason we force people to use Jira is because it only works if everyone uses it.

In my experience, AI out of the box is at first a useless gimmick - until someone starts seriously playing with it and defines a skill file for integrating it with some internal tool. And another person starts playing with it and figures out that AI is pretty good at using another internal tool but only if the tool runs in --silent=1 mode by default, so as not to confuse AI with too much logging output. And a third person figures out that it's actively dangerous to let AI some some other internal tool - but hey, there's a safer alternative, and which happens to perform better too. And pretty soon you end up with an ecosystem of business-specific scripts and .md files and skills and MCPs that's actually helpful 85%+ of the time. But the only way to get there is to get devs and power users tinkering with it.

[sarchertech]

In my experience all the md files just pollute the context and make it less likely to do what I want it to do. I’m at a huge org with thousands of power users doing all of this and I haven’t seen anything resembling the results you’re seeing.

But even assuming this is the case, you don’t create enthusiastic power users with threats (implicit or explicit) and metric tracking. The only thing that does is force people to do the minimum to keep their job.