Argos, a few days later: the weak point in the system isn't the agent

A few days ago, I published the first post about Argos, the agent I’m building to drive Neodigit’s growth. Since then, a pretty funny finding: the weak link in the system isn’t the agent. It’s me.

What it does, every day

Every morning, Argos still follows the same protocol:

• it re-reads the memory (Notion) to know where we stand on each channel,
• it checks what’s actually live before proposing anything, never based on memory alone,
• it hands me 1 to 3 actions calibrated to fit in 30 minutes, with a copy-paste-ready draft whenever there’s content to produce,
• it never publishes anything itself: it writes, I approve or fix, then I send.

Nothing new since the first post: it’s still the same protocol running, without me watching over it.

What I hadn’t planned for

What I hadn’t anticipated is that it follows up with me. More than once, a draft sat ready in Notion while I just didn’t click “send.” The action was calibrated, the text was written, all that was left was to execute it, and yet.

The agent was never the problem. I was, from time to time.

In a way, that’s the most interesting result of these last few days: Argos hasn’t gotten a single call wrong on what needed to be done. The bottleneck is human execution, not strategy or content.

What’s next

I’ll keep sharing what works, what doesn’t, and mostly try to keep up with the pace it sets rather than the other way around.

Got a similar use case?

If a repetitive task in your organization could be handed off to an AI agent (reporting, sales follow-up, customer support, monitoring), let’s talk.