Non Dev Ways to Use AI in Your Job
Posted by Zealousideal_Meet482@reddit | ExperiencedDevs | View on Reddit | 14 comments
I feel like I've heard a lot lately about AI and how great it is for development and idk if it's mainly because I'm in dev circles, but I've heard less about the other more tedious tasks. Like the things that I've hated or have taken up the most space in my head for my job have always been the non dev things like meetings, remembering to follow up with people about things, setting up processes for things and actually making sure people are following said processes, etc.
So I was wondering, what are some non dev ways that you all use AI? What tools do you like to use? I've been looking into AI for meeting notes and follow ups specifically, but I'm also wondering - because devs can be crazy creative - if there are other creative ways out there to make life easier that I probably just haven't thought of.
slimracing77@reddit
I have a note taking AI I like to use to scaffold documentation and just generally get thoughts out of my head. I take a walk around my neighborhood and just talk out my thoughts, it's pretty good at taking 20 minutes of rambling and returning a nice formatted outline.
lokivog@reddit
Claude has significantly enhanced our husky commit hooks to enforce code quality and processes for things that linters can’t easily validate. Not really creative but it’s helped enforce policies and processes that we normally wouldn’t have had time to implement.
PositiveUse@reddit
My PM colleagues use AI to write overly simplified docs to throw over the fence with the question „can you implement this by end of week? I already went through all the details and possible questions are answered in the Q&A section at the end of the document“
rover_G@reddit
Research
Leading_Yoghurt_5323@reddit
i mostly use it for writing, summarizing long threads, and drafting replies… saves way more time than coding use cases tbh
mmm19284202@reddit
The Atlassian MCP lets you read write Jira and Confluence pages. I’ve found this useful for program reporting work (how are my initiatives progressing, create a release notes page showing what was delivered in March, summarise these documentation pages, etc)
zmerlynn@reddit
Yeah, like others have said: - summarizing and research: being able to scour hundreds of design docs, postmortems, etc. - brainstorming/design/planning: this is kinda like “rubber duck debugging” but for design. If you follow the previous step and give enough context it’s honestly kind of great, though you have to really watch out for confidently wrong things (I’m constantly trying to poke agents to research more). - process stuff: I’m starting to toy with a skill for iteration planning, TBD if it helps. Definitely good for research but I’m trying to play with skills to maybe drop some of the less interactive parts of iteration meetings.
ilikeaffection@reddit
I use it to write emails to leadership.
GrowthProfitGrofit@reddit
It's very good at translating technical concepts into meaningless buzzwords that make leadership feel like they understand what's going on.
GrowthProfitGrofit@reddit
I use it to write tech specs. In my experience most tech specs go almost entirely unread, they're just produced so that managers can check a box. Also they often strongly prioritize length over relevance.
I actually quite like writing but I can't stand tech specs. So I just dump a few pages of unstructured thoughts into an LLM and let it go nuts. It usually needs a few edit passes since it'll just lie and inject arbitrary unrelated nonsense. But I find it much easier to edit than to get past the initial phase of staring at a blank screen while listening to loud static noises inside my head.
Buff_Lightyear@reddit
A little unsure what qualifies as non-dev but work related, but I use it to summarize and document slack conversations pertaining to jira tickets and add comments via atlassian mcp for discoverability and transparency as to why we decided to do something some way, have it pull issues from sentry and propose causes/solutions, have it refine epics/tickets, used it to make a script that pulled all my prs in our org over the past year and summarize my contributions to use in my annual review, also used that script to recommend augmentation to my resume. Also had it build a script to pull all prs over the past two weeks and provide summaries for ritual meetings.
olddev-jobhunt@reddit
This is still dev-oriented, but: my company has some alerts that go to slack for things like API credits usage for things we refill manually (yeah it's dumb but whatever.) I copied a few months of the conversation straight from Slack and said "give me a CSV with dates and the values for only the one service" and it did. I used that to project usage in a spreadsheet.
I think there's a lot of "take this unstructured data and make it more structured" where LLMs shine pretty effectively.
Empanatacion@reddit
I don't have a visual brain, so when I write up a design doc, I have it write the mermaid for a diagram just by describing the process in English.
The huge one for me though is giving it an mcp for as many things as I can. It can talk to jira, confluence, GitHub, databricks, our databases and message queues, datadog. It can connect to the correct VPN as needed. It can fire up a headless browser, log in manually with our test accounts that must go through our SSO, then sniff the oauth token. It then uses that token to make direct rest calls to our services.
Any of those things are useful on their own, but then it can stitch them together. It looks in datadog and sees a problem with a particular entity and sees the ID and figures out from our config files what DB and table to look in and queries the data directly. Then sniff the dead letter queue to see if the message died. Then sends a test message with a format it figured out by looking at the code.
Then reports back with a diagnosis of the issue. An issue I didn't need to describe directly because I just gave it the jira ticket number.
rocketpastsix@reddit
Due to my dog needing chemo, I had to stop having a personal trainer to make sure I could easily afford her treatments. I was with the trainer for 3 years. Great accountability buddy, always answered questions, motivated me, etc. When I made the decision to stop, I went to Claude and gave it my goals, and let it create a 4 week cycle of workouts for me. Initially did it in obsidian but then moved to a tracking app.
Also exported my credit card CSV and had Claude do a detailed analysis on it. Spending trends, subscriptions I may miss and can cancel, things like that. Was incredibly helpful and eye opening.