AI Customer Operations Agent with Human Approval and Auditing

AI Customer Operations Agent with Human Approval and AuditingAI Customer Operations Agent with Human Approval and Auditing

The network for creativity

Join 1.25M professional creatives like you

Connect with clients, get discovered, and run your business 100% commission-free

Creatives on Contra have earned over $150M and we are just getting started

Back to feedPost

Atique Ahmed

• Oct 4

I built an AI agent that handles customer operations — refunds, order lookups, and support tickets — with a human approval gate built into the workflow. The agent proposes the action, pauses, and waits for a human to approve before anything executes. The LLM never makes the final call. Role-based access and a full audit trail are enforced in code, not prompts. Deployed live on Azure Container Apps.

Completed work

AI Agent Engineer AI Automation Business Workflow Automation Azure FastAPI Python

Johnson Afolabi

• 20h

The pause for human sign off keeps refunds safe while still offloading the lookup work to the agent.

Atique Ahmed

• 12h

Exactly — and the pause isn't a soft one. The graph suspends at a LangGraph interrupt() node, persists state to the database, and waits. Even if the server restarts, the approval is still pending. The agent doesn't move forward until a human explicitly clicks approve. That's...

Samuel Lasisi

pro

• Sep 29

I cannot get my demo to be below 3 mins.. its looks impossible. So I'll drop my pitch here I built an AI Agent that books appointments directly inside chat apps like WhatsApp, Instagram, Telegram and Facebook Business through natural conversations. I hooked up a few things to make it happen: - WhatsApp Business API - Meta scopes for Instagram - Scopes for Facebook messaging - Google Calendar API - AI Agent (powered by Lovable) All of these work together to help businesses get appointments without lifting a finger, saving time and money in the process. PROBLEM: Usually, appointment-based businesses would use booking websites (pretty common these days) to take bookings, but bookings generally dont start from there. It starts with customers sending messages on WhatsApp or Instagram (or social media generally), like, “Are you free on Saturday?”, and then they wait to be directed.

That normally means someone has to stop what they’re doing, reply to the customer, check the calendar, suggest a time, follow up, create the booking (or send a booking link), send reminders, collect payments, and handle any changes afterwards or, worst, answer repeated questions for the nth time. and Because clients/customers require fast responses, businesses lose hundreds of customers a year because their competitors replied first! So I built PingBase PingBase removes that entire back-and-forth. Instead of sending customers somewhere else to book, PingBase turns the conversation they’re already having into the booking. This is usable fo any time of business for multiple use cases. To test the project: Send a WhatsApp message to the fictional business (Studio 9 - already set up on teh platform) at +447587492916 or message @samuellasisi1st on Instagram. Or give it a spin yourself on Lovable: (maybe create your business) https://pingbase.lovable.app Demo below - 3min video - Some screenshots of the product (+ WhatsApp and IG Conversations) - 10+ Min breakdown of how the product works end-to-end Things to note: Meta-required scopes for this are not available for public use yet as it's not something I can control, so you might not be able to connect your own business Instagram or WhatsApp acounts just yet. but you can try...

lovablechallenge