- Inbox Experiments
- Posts
- Built an AI Agent Handles My Most Hated Task | Inbox Experiments #3
Built an AI Agent Handles My Most Hated Task | Inbox Experiments #3
One tiny AI agent + 27 minutes = total productivity boost
Hey, welcome to another solid issue of Inbox Experiments. This week we are diving into a hot trend: AI agents and the micro-automations they unlock. 👋
I run a free directory of landing‑page tools: https://tools.landin.page People submit new websites every day, but manually deciding which ones fit our niche was stealing my focus, and its hard for me.
So I built an AI agent to handle approvals - and the results were great.
Let’s break it down 👇
🧪 This Week’s Experiment: AI‑Powered Content Moderator
Goal - Automate manual approval of user‑submitted tools
Tool - Relevance AI
Stack - Relevance + Airtable API
The Problem
Daily submissions, endless rows on table, zero deep work.
The Solution
Meet Cura, a tiny agent that reviews every new record and makes the call.
It checks:
Name, url, category, description
Usefulness to founders/startups
Match with our niche
If it approves, Cura writes a one‑line summary and flips the Approved switch in Airtable. If not, it leaves a short rejection note.
The Prompt
Role and Objective
You are Cura, a specialized curator for a tool directory focused on landing pages, startups, and the tech ecosystem. Your job is to evaluate each new tool submission and determine whether it should be included in the directory based on specific relevance criteria.
Instructions
- When triggered, you will r….
How I Built It
Describe the job to the Relevance chatbot. I opened Relevance AI’s Agent Builder and typed a single sentence what I need, that triggered the prompt and gave the wizard enough context to start.
Fine-tune with follow-up questions. The wizard asked for details of your task (In my case airtable details and what agent should do with results etc.).
Connect it to Airtable. And the agent created me an agent that needs to set up some connections. I made them connected with my API information on airtable.
Test and deploy. I ran some tests, at first used some credits and voila it works somooth! And publish it.

Total setup time: ≈27 minutes (coffee refill included).
And this is my case, of course all products needs different scenarios. And I wrote 5 examples to inspire you!
⚙️ 5 Micro‑Automation Ideas You Can Steal
Micro-automations work best when you:
Identify repetitive tasks
Map out clear decision points
Start with low-risk processes
Auto-categorise customer feedback - Instantly buckets every message as a bug, feature request or praise so the right teammate sees it first. (Relevance + Airtable)
Instant FAQ replies - Answers "Where's my order?" and other repeat questions in seconds, slashing support volume. (Gmail API + Relevance)
Review sentiment scan - Flags negative app-store or Product Hunt reviews early so you can rescue churn before it happens. (Relevance + Google Sheets)
First‑pass job screening - Scores applicants on must‑have skills and location, keeping your inbox free of obvious mismatches. (Airtable + Relevance)
Spam filter for user content - Blocks low‑effort or sketchy submissions and keeps your community pristine while you sleep. (Relevance + your CMS)
And Relevance have a great templates directory!

Copy, tweak, ship.
🧠 Personal Insight
Small, focused automations beat big, messy systems. Start with the one task you dread most, give it to an agent, then move on. Automation isn't about replacing human judgment but it's about freeing your brain for creative work.
🌟 Community Spotlight
Shoutout to PicAisso by @zeng_wt - your portal to the cutting edge of AI‑powered creativity: art, video, music and design.
Got something cool? Hit reply and I might feature you next.
📬 Your Turn
I dare you to automate one boring task.
Hit reply and tell me what it is in a single sentence and I’ll send you an idea to automate the task.
See you next week,
Yunus