Human-Centric AI Agent Teams — Reflections on the OpenClaw × AWS Event

Last Monday evening, over 400 people walked into the AWS office in Singapore for the largest OpenClaw community meetup in Southeast Asia. Engineers, founders, C-suite leaders from MNCs, students — all in the same room. The energy was electric. 🚀

The event was a collaboration between AWS, The AI Capitol, and the Singapore OpenClaw community. Here's what I shared, what I learnt from the other speakers, and what the audience got me thinking about.

How I Ended Up Running Teams of 10+ AI Agents

A year ago, if you told me I'd be managing a small army of AI agents through Telegram, I would've laughed. But here we are.

It started with one agent to help me sort through research papers. Then another for grocery lists. Then scheduling. Before I knew it, I had 10+ agents organized into Telegram groups by topic — home, work, personal development — with a couple of these teams running in parallel, plus a few experimental ones I'm still tinkering with.

They don't just answer questions. They have their own wallet system. They have guardrails and security boundaries. Over time, they get better at what they do. 💡

What Is a Human-First Approach to AI Agents?

This is the part of my talk that seemed to land the most — and honestly, it's the question I care about most.

The question I ask before building any agent isn't just "does this make me more productive." It's also: does this fit into the dimensions of my life? Do these agents work together in a way that frees me up to be more human — think deeply, be creative, stay fit, spend time where I value and with the people I care about?

I think of it as a human-first approach. Not automation for its own sake — augmentation. Not just making us more efficient working machines, but using AI to create more space for the things that actually matter.

🔹 The same logic applies to enterprises. The question isn't whether to automate in silos. It's whether your AI aligns with the company's vision and business — and elevates the organization, not just trims a few hours off a workflow.

What the Other Speakers Covered

I was one of four presenters, and I learnt from each of them.

🔹 Melanie Li (AWS) co-presented with me on OpenClaw on AWS Bedrock AgentCore — covering multi-tenant isolation, auth via API Gateway and Lambda, DynamoDB session management, and Bedrock Guardrails.

🔹 Nathaniel Ng (AWS) mapped the full deployment spectrum — from laptop to Lightsail to enterprise AgentCore — and broke down compute versus token costs with a compelling model comparison on cost-to-performance.

🔹 Boon Kgim ran live prompt injection attacks on stage — context window flooding, attention dilution — a powerful reminder: treat agents as untrusted users, enforce access controls, sandbox everything.

🔹 Douglas Sim demoed ClawBuilder — one-click OpenClaw deployment with Tailscale VPN and exportable persona snapshots. Configure once, replicate anywhere.

The Conversations That Stayed With Me

The networking sessions after the talks were just as rich as the presentations themselves. A few themes kept coming up: taking this community beyond Singapore, the use of AI in education (raised passionately by a group of ladies and moms), memory management for long-running agents, and more.

And then one audience member showed me something that changed my evening: a dashboard for managing multiple OpenClaw agents at scale. 🎯

Why I'm Upgrading to Mission Control

Running teams of agents is fun. Keeping track of all of them is a different challenge — seeing what they're all doing, what they're costing, how they relate to each other.

After seeing what that audience member had built, I decided to upgrade my setup to Mission Control for Agents — a centralized dashboard for managing multiple OpenClaw gateways. Visual agent maps, real-time activity monitoring, cost tracking by agent/model/provider, audit trails, and full session logs with timestamps and token counts.

The install is one command — and once I've finished setting it up, I'll share how it all comes together: the human-first framework at scale, enterprise dimensions, Mission Control in action, and how the architecture is evolving. More soon. 🙏

Resources from the Event

Everything from the event, all in one place:

#AlwaysDay1 #AgenticAI #HumanFirst #OpenClaw #AWS #Bedrock #AIAgents #Singapore #Community