Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More
154 contributions to AI Automation Society
you check your automations
Be honest: how often do you check your automations after they go live? A) Every day B) Once a week C) Only when something breaks Drop a letter below
you check your automations
Make the AI step predictable: 4 rules
1. Give it one narrow job. "Classify this message as sales, support, or spam" beats "handle this message." 2. Ask for a fixed format. Request JSON with named fields, and show one example of what you expect. 3. Validate the output before you use it. Check that required fields exist and values are in the allowed list. 4. If validation fails, retry once, then send the item to a human. Why: free text is hard to route, and models can be wrong. The risky moment is sending unchecked output to a client's customer. A made-up example: the model returns a category and a draft reply. The workflow checks that the category is one of three allowed values. If it isn't, nothing is sent and a person is notified. What's your favorite way to keep AI steps in check?
Make the AI step predictable: 4 rules
0 likes • 4h
@Zubair Aqeel good catch, the retry has to change something or it's pointless. I'd add the failed check to the second prompt, like "category must be one of these three", so the model knows what it got wrong. still capped at one retry, then a person.
0 likes • 4h
@Ali Abdullah thanks Ali, the cap is what keeps it from looping forever
follow-up to leads
Most leads don't say no. They just go quiet. Someone asks a question, gets busy, forgets, and by the time you reply, they've moved on or lost interest. It's not rejection. It's life getting in the way. There's a simple reason this happens. People buy when their interest is highest, and interest fades fast. The first reply matters more than most people think. A fast, helpful response feels like care. A slow one feels like you don't need the business. This is where automation helps, and not in a robotic way. A good setup does three simple things: 1. Replies right away, so the person knows they were heard 2. Asks one easy question to keep the conversation moving 3. Follows up gently if they go quiet, and stops when they answer Notice what's missing. No pushing. No pressure. No "last chance!" messages. Good automation doesn't make you pushy. It makes sure nobody gets forgotten. And the human part still matters. The automation opens the door, but you're the one who builds the trust once they walk in. Do you follow up with every lead, or do some slip through?
follow-up to leads
1 like • 3d
@Jaskaran Saini yeah, and thanks for the push on this, the fifth follow-up point is useful for the bigger jobs
0 likes • 4h
@Sam Jak bit of both Sam. template for the structure, but I change the first line and the number or date by hand each time.
Building an AI agent is easy. Running one in production is the real test.
Most AI agents look impressive in a demo. In production, the problems show up: - They loop and burn tokens - They lose context mid-task - They call the wrong tools - Occasionally, they take actions they were never meant to take Getting an agent to work once is the easy part. Making it secure, reliable and ready for real users is a different challenge. Here is one of the lessons worth remembering: Don’t use an agent when a deterministic workflow will do - If the steps are known in advance, a workflow is cheaper, faster and more predictable. Simplicity is an architectural advantage. I learned this while building an AI lead qualification agent for WhatsApp. I gave a single agent node every job: holding the conversation with the lead, scoring them, saving the lead, checking availability and booking the appointment. It worked, but the agent was overloaded, and I started seeing unnecessary and inconsistent outputs. Most of those tasks are fixed, predictable steps. The agent’s real value was in the conversation itself. The rest is better handled by regular workflow nodes, where the behaviour is the same every time. What’s the biggest challenge you’ve faced moving an AI agent from prototype to production? Or if you haven’t built one yet, what worries you most about it? Tell me in the comments.
1 like • 3d
@Leela Krishna that’s the right setup. saving the half-finished booking with a pending status means nobody has to start over, and holding off on the calendar event avoids blocking a slot nobody confirmed
0 likes • 4h
@Leela Krishna Well said Leela. Keeping pending separate also makes it easy to see later how many leads drop off at the time slot step, which is useful data on its own.
Our first episode of Humanly Possible is live!
Hey guys, I’m super excited to share Humanly Possible with you. It’s our new show about people in this community, what you’re doing with AI, and how it’s changing life for you and the people you care about. Our first guest is @Amy Simpson. When her daughter Emma was struggling with a college class five hours from home, Amy used Claude to make a study guide around the way Emma learns. Her grade went up more than 10 points. I’d love for you guys to watch the first episode and leave Amy a comment on YouTube. It would mean a lot to us to see this community show her some love. We’re taking guests, too. If you’ve got a story you’d like to share, apply to be on Humanly Possible. - Nate
0 likes • 4h
@J.R. Mukete yup!
0 likes • 4h
@Muskan Ahlawat yeah sure
1-10 of 154
Ameeth B.
7
4,442 points to level up
@amit-bhagria-1088
Hi, I'm from India and currently working in the HR Space

Active 4h ago
Joined Nov 4, 2025
Gurgaon, Haryana, India
Powered by