HN Explorer

Ask HN: How to roll out Agents for knowledge workers?

1 points by luplex ↗ HN
I work in a medium-size government agency in Germany, mostly comprised of middle management. We have already built and rolled out a web-based "internal GPT" with LangGraph and a custom frontend, based on Azure OpenAI, but it of course lacks far behind the commercial offerings in functionality. We also have secure on-premise infrastructure, so I'm not looking for security debate here, and instead just product features.

I'm looking for success stories to orient our future internal product strategy around. On the web, I mostly see AI slop stories about this sort of stuff, no concrete cases. Who is in town?

To anyone in enterprise IT rolling out AI, have you found benefits in giving knowledge workers more agentic AI? What do they actually end up doing?

Or do you have any pointers to use cases?

6 comments

[ 36.8 ms ] story [ 309 ms ] thread
rossng
We created a very simple platform for shipping internal-only static web apps that can talk to a document database for storage. There's a Claude skill that describes how to write and deploy apps.

This is for a robotics company, so we're pretty ops-heavy. Non-engineers quickly found all sorts of use cases for making specialised forms/databases (inventory management, scheduling etc.)

luplex
oh that's nice! are the web-apps read-only, with nicely scoped access, or can they alter each other's documents?
rossng
writes are isolated to a certain scope, reads aren't. But I think access control rapidly becomes the most complex thing if you are in a bigger org.