We Ditched Slack and Our Intranet for Arheev
Tools don't just organize work. They architect behavior. One company replaced their internal comms stack with Arheev and it did more for culture than two years of offsites combined.
6 min čitanja
When we started building Arheev, I thought the AI part would be the hard part. Train a model, feed it some data, boom, automatic leave approvals.
I was completely wrong.
The tech wasn't the problem. The problem was that every company has a different definition of "fair." What feels reasonable at a 20-person startup doesn't work at a 200-person company with five departments and different leave policies.
Companies using AI tools in HR are cutting admin time by about 40%. That's not marketing speak, that's what our clients tell us when they've been using Arheev for a few months. Less time in spreadsheets, more time actually talking to people.
But here's the thing nobody talks about: AI doesn't fix bad processes. If your leave policy is confusing, AI just approves things confusingly faster.
Manual leave tracking is miserable. I've talked to HR managers who spend half their Monday mornings just processing requests from the weekend. Multiple approvers, checking balances, making sure the team isn't left understaffed, it adds up.
Our AI system tries to handle the routine stuff:
The interesting part? The AI isn't making decisions, it's just doing the tedious comparison work that managers hate. "Didn't we already approve this type of request last month?" That kind of thing.
One client told us their approval time went from 2-3 days to same-day for most requests. Not because the AI is approving everything automatically, but because managers aren't digging through past decisions to stay consistent.
Resume screening AI is everywhere now. Scans thousands of applications, matches keywords, predicts performance, you've heard the pitch.
The good part: it's way faster than reading 500 resumes manually.
The sketchy part: these systems can bake in bias if you're not careful. If your past hires all came from the same backgrounds, the AI will just optimize for that. We decided not to build this feature yet because we haven't figured out how to do it in a way we'd actually trust.
The AI features people actually use in Arheev:
Nothing revolutionary. Just the annoying lookups that waste everyone's time.
We tried building sentiment analysis (scanning messages to detect unhappy employees). Tested it internally. Felt creepy. Scrapped it.
Annual reviews are dying. Good riddance, they were always kind of useless. You're giving feedback about something that happened nine months ago?
AI systems can track goals in real-time, measure against objectives, all that. Sounds great in theory.
In practice, we've seen companies get obsessed with metrics that don't actually matter. Time tracking down to the minute, counting commits, measuring "productivity."
If your metrics are garbage, AI just gives you garbage faster.
We're careful about this in Arheev. Track what matters (goals, deadlines, actual outcomes), ignore the surveillance theater.
The best use of AI in HR isn't automation, it's pattern recognition.
This is where HR stops being reactive and starts planning ahead. Not because AI is magic, but because it can crunch numbers faster than humans.
AI should handle repetitive tasks. Humans should handle the complicated, messy situations that require judgment.
In Arheev, we try to automate the obvious stuff:
But we keep humans in the loop for:
One of our clients put it well: "The AI handles the 80% that's straightforward. I focus on the 20% that actually needs me."
Honestly? A lot.
We're working on better team coverage predictions. Right now it warns you about gaps, but it could probably suggest better timing for approvals.
We want to add smarter policy recommendations, "Here's how other companies handle this", but that requires way more data than we have.
And we're trying to make the AI explanations clearer. "Request approved based on previous patterns" isn't helpful if you don't know what those patterns are.
AI isn't replacing HR teams. It's just removing the tedious parts that shouldn't require human time anyway.
Systems like Arheev work best when they handle the routine stuff, checking balances, comparing past decisions, flagging problems, so HR teams can focus on the actual people work.
The companies doing this well aren't trying to automate everything. They're using AI for what it's good at (pattern matching, data analysis) and keeping humans for what they're good at (judgment, empathy, dealing with exceptions).
Success isn't about having the fanciest AI. It's about figuring out which parts of your HR work actually need automation, and which parts need a human who understands context.
We're still learning where that line is.
Tools don't just organize work. They architect behavior. One company replaced their internal comms stack with Arheev and it did more for culture than two years of offsites combined.
Most managers know they should give feedback. Few do it well. Here's what actually works, and why timing matters more than most people think.
We built an Android app for Arheev because managers kept asking to approve leave requests from their phones. Took us six months. Here's what works, what doesn't, and why iOS isn't ready yet.
Napravite prvi korak ka boljem upravljanju ljudskim resursima