Seven weeks ago, I installed an open-source AI agent called OpenClaw on my MacBook. OpenClaw is an agentic AI tool built on top of large language models that can operate your computer, run commands, manage files, browse the web, and chain together complex multi-step tasks on its own. Think of it as an AI that doesn’t just answer questions but actually does the work: it can SSH into a server, fix a config, commit the change, and move on to the next task without you hovering over it.

Today it runs on a dedicated homelab server, and it has become one of the most productive members of my team. It handles my email. It manages my servers and my home router. It secures my infrastructure across both my homelab and AWS. It migrates my DNS. It recovers crashed virtual machines. It sets up VPN infrastructure from scratch. It researches business strategy and drafts frameworks I can actually use.

I’m developing three businesses: Qandaba (AI consulting), Soláyae (artisan handbags), and Derezd (stream-to-print 3D printing). Between the three of them, there are domains to manage, servers to maintain, emails to filter, integrations to wire up, and a thousand small infrastructure tasks that used to eat my evenings and weekends. On top of that, I have a team in the Philippines that needs IT support: email accounts, permissions, groups, onboarding. The kind of work that, at any mid-sized company, would require at least one full-time IT person. Probably two.

I have zero.

How we got here

The first thing OpenClaw did for me was tedious. I needed to hire Node.js developers for a Qandaba client, and that meant downloading 64 LinkedIn resumes, reaching out to candidates, and scheduling interviews through Calendly. The kind of recruiting coordination that eats an entire week if you’re doing it yourself between client calls. OpenClaw handled the outreach, managed the back-and-forth, and booked candidates on my calendar while I focused on actual consulting work.

That was February. By March, things escalated.

I moved OpenClaw off my laptop and onto my homelab server, running it as a proper service that starts automatically. That same week, one of my virtual machines crashed. Instead of spending my Saturday morning troubleshooting boot sequences and checking logs, I handed it to OpenClaw. Recovered. Then I had it set up GPU transcoding on the same server, something I’d been putting off for weeks because it meant reading documentation I didn’t want to read.

By the end of March, OpenClaw was managing real infrastructure: setting up my NAS, configuring my home router and network port forwarding, migrating my Derezd domain DNS to Cloudflare, auditing my virtual machines, managing Docker containers, configuring backups, and rebuilding my entire VPN setup from scratch with fresh certificates. That last one is the kind of task that, if you’ve ever done it manually, you know takes hours of careful steps where one mistake means starting over.

It also took over security across my entire stack. My homelab servers, my AWS cloud infrastructure: OpenClaw handles hardening, monitoring, and keeping things locked down. When you’re running production services for three businesses, security isn’t optional, but it is the first thing that slips when you’re busy. Having an agent that treats it as routine rather than a special project changed everything.

The email problem

Here’s one that every founder will recognize. I have too much email. Not the “I’m so popular” kind; the “I’m drowning in noise” kind.

In late March, OpenClaw set up direct access to my email accounts and then built me an automated filter. Not a simple rule like “move newsletters to a folder.” A scoring system that evaluates every incoming message, decides what’s worth keeping, and runs on a schedule throughout the day: 6 AM, 10 AM, 2 PM, 6 PM, 10 PM. Five times a day, my inbox gets cleaned without me touching it.

When there was a bug in the filter logic (a flag was inverted, so it was keeping what it should have been discarding), OpenClaw found and fixed that too. The whole system has been running for weeks now, and I barely think about email anymore. That alone would be worth the setup.

The real story

I could keep listing things. The Notion integration for tracking business plans. The research on AI transformation strategy that helped shape Qandaba’s pivot from outsourcing to consulting. Managing IT for my team in the Philippines: setting up email accounts, configuring permissions and groups, handling the kind of admin work that piles up when you’re onboarding people across time zones. The local AI stack it helped me build just yesterday: language models, image generation, speech-to-text, text-to-speech, all running on my own hardware.

And then there are the scheduled tasks. OpenClaw compiles my top five AI news stories every morning so I start the day informed. It tracked down the cheapest flights for my upcoming Philippines trip and kept checking as prices changed. It runs regular health reports on my entire homelab, so I know the state of every server and service without logging in to check. These aren’t one-off requests; they’re recurring jobs that run on their own, like having an assistant who never forgets and never takes a day off.

But the list isn’t the point.

The point is that I’m a solo founder running three businesses, and I have infrastructure that used to require dedicated staff. Not because I’m technical enough to do it all myself (I am, but I don’t have the time), and not because I hired someone. Because an open-source tool that didn’t exist a year ago can now handle the kind of work that companies used to budget $80,000 to $120,000 a year for.

This changes the math of starting a company. Dramatically.

What I actually learned

The interesting thing about having an AI handle your IT isn’t the efficiency. It’s what happens to your time and your thinking when the infrastructure just works.

I used to spend weekends on server maintenance. Now I spend them on strategy. I used to put off DNS migrations and VPN setups because they felt like quicksand: easy to start, hard to finish cleanly. Now they get done in an afternoon while I focus on client work or product development.

The compound effect is real. When small infrastructure tasks stop piling up, you stop carrying the mental weight of them. You stop context-switching between “CEO making decisions” and “IT person debugging a config file.” You just operate at the level you should be operating at.

I’m not saying OpenClaw is perfect. It’s not. There are times when I need to guide it, correct it, or break a problem down into smaller pieces so it can reason through them. It’s a tool, not a colleague. But it’s a tool that learns, that remembers context, and that can execute multi-step technical tasks that would take me hours.

What this means for founders

If you’re a solo founder or running a small team, the era of needing dedicated IT support to operate professionally is ending. Not “someday.” Now. The tools exist. They’re open-source. They run on hardware you can buy for the cost of one month’s salary for a junior sysadmin.

This doesn’t eliminate the need for human expertise. You still need to know enough to point the tool in the right direction, to evaluate its work, and to make judgment calls it can’t make. But the execution layer, the actual doing of infrastructure work, that’s increasingly something an AI agent can handle.

Seven weeks. That’s how long it took to go from “let me try this thing” to “I can’t imagine running my businesses without it.” Each task OpenClaw completed was a small tunnel on the drive. Not the destination, but proof that I was getting somewhere.

I’m somewhere now. And the road ahead looks a lot less lonely than it did in January.