← Back to archive

AI Agents Will Become the Primary Users of Software

I'm increasingly convinced that AI agents will become the primary users of software.

I already see it in my own workflow.

I've built shared infrastructure for my apps: my own RevenueCat-like service, analytics exposed through MCP, and Claude skills that analyze campaigns, publish apps, patch them based on user requests, and validate the code.

I now talk to my data in plain English instead of searching through interfaces I can never remember.

I see the same shift at Binitex.

We recently started helping a factory automate its processes. They were already using Claude, so we suggested connecting it to their data, analytics, incoming orders, and internal systems through MCP.

And you don't need to replace everything first.

You can build an adapter over the existing old, dusty systems and make them accessible to agents. With AI helping write and maintain those integrations, this has become much easier.

Once everything is connected, you suddenly see how much work can be automated.

The key question for future software is simple:

Is the human here to act, or to judge?

If they are acting, give the task to the agent.

If they are judging, build the interface for approval, verification, and trust.

Dashboards are an intermediate layer between data and action.

Agents can compress that chain:

data → interpretation → decision proposal → action → verification

Screw the dashboard. We need outcomes!

A man looks surprised at a screen with eyes, while a robotic figure in a red dress holds a sign stating "ALL DONE."

Originally posted on X

PS: I'm pretty active on 𝕏 if you'd like to follow along

Archive

2026

2025

2024

2018