Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More
52 contributions to AI Automation Society
I've been messing around with a few things, researching some ideas
and somehow ended up with more tabs open and more work than I started with lmao. One thing I've been trying to understand is where automation agencies actually lose time or money after a client project goes live. Not the building part, but the day-to-day headaches. Is it: - Clients constantly asking for little extras that never make it into the bill? - Automations that run successfully but still produce the wrong outcome? - AI agents making mistakes that nobody notices until the client does? I'm trying to figure out which of these problems is genuinely painful, and what else I'm missing. If you run an automation agency, what's the one recurring headache you wish you could get rid of? Would rather hear the messy real-world stuff than guess from the outside 😂
0 likes • 3h
@Chetan Mishra True agreed with this one , how often do you face this issue
1 like • 2h
@Holger Morlok That is a very honest reaction from your side , but if you actually had to push for a price what would it look like and would you actually mind paying it (give it for eithe rof the scenarios)
One thing that can quietly wreck an agency project:
The phrase “while you're at it…” You agree to build an automation that qualifies incoming leads. Then the client asks if it can also update the CRM. Fair enough. Then they want a Slack notification, a follow-up email, and a way to handle leads that don't respond. None of those requests sounds unreasonable on its own. That's what makes this tricky. But suddenly, the original workflow has doubled in complexity, and you're still treating it like the same project. One thing I'd start doing differently: When a new request comes in, separate it into 3 buckets: - Bug: Something we agreed to build isn't working. - Adjustment: A small change that doesn't alter the original scope. - New requirement: Something the original agreement didn't cover. The important part isn't saying no to every extra request. It's making the trade-off visible before agreeing to it. Otherwise, you can finish a project with a happy client and still lose money on it.
0 likes • 8h
@Alexandre Seki Your welcome and hope it helps you
0 likes • 8h
@Wasilis Paliakoudis Yes tbh we need to kinda delay the free requests and deliver the actual work scope first , because every free request that you add on top is just pure pressure and scope creep , which ofc is not very good in the industry
I spent way too long to find why one of my reports felt weird.
Nothing was technically wrong. The numbers were right. The calculations were right. The layout was fine. But when I looked at it as a client, it just felt… dead. Then I noticed the problem. It was showing what happened, but not really why I should care. So I changed one line at the top of the report. From: “Automation Performance — August” to: “This automation handled 1,842 interactions this month.” Same data. Completely different feeling. Funny how sometimes the problem isn't the product. It's the sentence sitting above it.
0 likes • Oct 3
@Yogesh Mohite Thanks , and pretty sure u aren't a beginner as the clients are mainly non technical
0 likes • Oct 3
@Lorenz Wieseke I feel it would be really solid , and hence i am planning to give it a shot
I showed report to someone today who immediately ignored
the part I thought was the most interesting. I was explaining the reporting side. How the numbers get pulled in. How the monthly report gets generated. How the client sees it. He listened for maybe 30 seconds and then asked: “Can it tell me when an automation isn't worth running anymore?” ...nope 😂 That wasn't even on my list. We ended up talking about that for way longer than the actual report. Now I'm staring at my notes wondering if I accidentally found a better feature than the one I was building.
0 likes • Oct 3
@Duncan Rogoff Gotta think about it ;)
0 likes • Oct 3
@Michal Myszkowski Ye that builds the credibility required , and also shows to the client that you know what you are doing
One thing I'd do before automating almost any client process:
Run the process manually 10 times first. Not because you need to become an expert at it. Because the first few runs usually expose things the client forgot to mention. A lead comes in without a phone number. Someone replies with a weird format. A request that sounds simple actually needs human judgment. Two people handle the same case differently. Those are exactly the things that turn into “bugs” once you automate the process. And honestly, I'd rather discover those problems on run #7 with a human watching it than on run #700 after the automation is live. The automation isn't usually the hard part. Figuring out what should actually be automated is.
0 likes • Oct 3
@Gal Miller We need to test the extreme cases just to be sure nothing breaks , which is very different from people testing the same normal path again
0 likes • Oct 3
@U. H. Farzana Haque Yup we need to have that human element sometimes just to ensure nothing gets into a fuss due to automation that you have set up
1-10 of 52
Elias Chaldean
6
1,253 points to level up
@pavan-zerkaa-7883
AI freak , OS and systems > Manual Work

Active 3h ago
Joined Apr 25, 2026
Powered by