Operating across companies with AI agent teams

Speaking to experienced CTOs and engineers

I was the lead speaker at the Ho Chi Minh City meetup and helped the team bring the event together. Most attendees were experienced CTOs, founders, engineers, and infrastructure specialists. My session, Taking my time back, drew on the agent systems I use across companies and projects, with Grok, Grok Bot, Cursor, and local and cloud agents.

The scope went well beyond coding. I covered engineering, growth systems, consulting, strategy, product, research, and operations: how specialist agents can carry distinct responsibilities while I stay accountable for decisions and results.

For that audience, the value was in the operating detail: how responsibilities are divided, how company context stays separate, and where a person needs to review the work. I connected those engineering choices to the wider challenge of managing several businesses and functions at once.

How the agents divide and carry the work

I explained the setup through Neo, my coordination point, and specialists for engineering, research, brand and content, travel, and operations. A chief-of-staff layer helps route the work, keep track of what is waiting, and bring decisions back to me.

A research task can inform product strategy, a product change can become an engineering brief, and completed work can feed a source-backed update or content draft. I showed how the Grok Bot–Cursor handoff fits into that wider operating system, rather than treating coding as the whole story.

Working across companies adds another constraint: the right context must travel with the task without mixing repositories, accounts, or company information. The talk covered local and cloud execution, memory, backups, handoffs, and the human checkpoints that keep this manageable.

The boundaries that make the workflow useful

The technical discussion was about reliability as much as capability. After the talk, attendees asked about architecture, autonomy, context, and why I designed the handoffs the way I did.

The examples ranged from meeting preparation and follow-ups to research, content, and software delivery. Each needed a clear owner and an output I could inspect. This was a look at my own operating system, not a claim that companies can be left to run themselves.

• Divide work by responsibility, not just by which model or chatbot is available.
• Keep company context separate while making handoffs between research, engineering, growth, and operations explicit.
• Verify sources, software behavior, and completed work; keep publishing and other consequential actions behind a human decision.

The people behind the meetup

Thank you to hosts David Zhang, Ivan Paulo, and Sandra Vu, and to Gradion for the venue. The original listing uses the SpaceXAI name; the current events calendar uses SpaceXSI.