Software Without Human-in-the-Loop: The SaaS Industry's Existential Crisis

Something clicked for me today. I’ve been writing about the personal software revolution — how AI coding tools are putting software creation in everyone’s hands. But I’ve been thinking about it from the wrong angle.

The bigger shift isn’t about who builds the software. It’s about who uses it.

The Human Buyer Is Going Extinct

Jerry Murdock, co-founder of Insight Partners (one of the largest growth equity firms in tech, with over $90B in assets), just laid it out on the 20VC podcast: humans will not buy software. Agents will.

Think about that for a second. Every SaaS company on the planet has spent the last two decades optimizing for a human buyer. Beautiful dashboards. Intuitive UIs. Onboarding flows. Feature comparison pages. Pricing tiers designed around human decision-making patterns.

All of it — built for a customer that’s about to disappear.

Dustin captured this perfectly:

He’s right. The fatal architectural flaw isn’t in the code — it’s in the assumption. Every SaaS product assumes a human is sitting at a screen, clicking buttons, reading dashboards, and making decisions. But when an AI agent is the one executing tasks, it doesn’t need your beautiful UI. It doesn’t need your onboarding wizard. It needs an API and clear documentation.

Why I Think This Changes Everything

I’ve been open source contributor and built (ie: BoxBilling) self-hosting software for years. I’ve always optimized for my workflow — finding the best tools, customizing dashboards, setting up monitoring I can glance at over morning coffee. But increasingly, I’m realizing that I’m the bottleneck.

My OpenClaw don’t need me to check a dashboard. My monitoring stack doesn’t need me to read graphs — it needs to take action when something goes wrong. My deployment pipeline doesn’t need me to click “deploy” — it needs to ship when the tests pass.

The human-in-the-loop was never a feature. It was a limitation we designed around.

Pontus Abrahamsson, who’s been building Midday (an open-source financial platform), shared his thoughts on this shift:

This is the direction everything is heading. Software that operates autonomously, with humans setting goals and constraints rather than clicking buttons and reading reports.

The Two Worlds of Software

I see the software landscape splitting into two distinct categories:

Software for humans — Creative tools, communication platforms, entertainment, anything where the human experience is the product. Figma isn’t going anywhere because design is a human activity. Slack persists because human communication has inherent value.

Software for agents — Everything else. CRMs, ERPs, monitoring, analytics, invoicing, deployment, infrastructure management. All of this becomes API-first, agent-operated plumbing. No dashboards needed. No pretty charts. Just clean interfaces that agents can call.

The second category is massive. It’s the majority of enterprise software. And it’s about to get rebuilt from the ground up.

What This Means for Self-Hosters

If you’re running your own infrastructure — and if you’re reading this site, you probably are — this shift is actually great news.

Agent-operated software is lighter. No frontend (or very minimal) to render. No session management. No cookies. Just APIs. This means your VPS resources go further. Where a full SaaS dashboard might need 2GB of RAM, the agent-operated equivalent might need 512MB.

Agent-operated software is more composable. When everything communicates through APIs, you can wire things together in ways the original developers never imagined. Your AI agent becomes the integration layer that replaces a dozen SaaS subscriptions.

Here’s my current stack evolution:

Before (Human-in-the-Loop)	After (Agent-Operated)
Check if can find open source project that meets my needs	Look for robust API that has 100% tests coverage
Grafana dashboards I check daily	Alerting agent that takes action autonomously
Manual deployment approvals	CI/CD agent that ships when confidence is high
CRM I update manually	Agent that enriches and manages contacts
Invoice review and approval	Financial agent with rules and thresholds
Server monitoring dashboards	Self-healing infrastructure agent

I’m not there yet on all of these. But the direction is clear. Each month, I’m removing myself from another loop.

The Infrastructure Play

This is why I keep coming back to the importance of owning your infrastructure. When agents are running your software 24/7, you need:

• Reliable uptime — agents don’t sleep, so your server can’t either (see our best VPS for developers)
• Fast API responses — agents work at machine speed, not human speed
• Sufficient compute — running local AI models for decision-making takes resources
• Data sovereignty — your agents are processing everything about your business

We’re Still Early

I want to be clear: we’re at the very beginning of this shift. Most software still requires a human clicking around. Most “AI agents” are glorified scripts with an LLM call bolted on.

But the trajectory is unmistakable. Murdock called it a tsunami, and I think he’s right. The companies building agent-first software today will be the Salesforces and Workdays of the next decade. And the companies still optimizing their onboarding funnels for human buyers will be the ones wondering what happened.

The smartest thing you can do right now? Start removing yourself from loops. Automate one workflow end-to-end with tools like n8n. Let an agent handle something you currently do manually. See how it feels.

I bet it feels like the future.

---

Running agent-operated software on your own infrastructure? You’ll need a VPS that stays up when you don’t. Check our VPS comparison or get started with Hostinger VPS — reliable, affordable, and perfect for always-on agent workloads.