An ad appeared on X this week that stopped me in my tracks. Not a job listing for a human — for an AI agent. The listing was offering $10,000 a month for an agent (or the human who built one) to handle content and product assessment tasks, presumably running 24/7. It’s a fascinating concept, and honestly, the more I think about it, the more it makes sense — especially given what we’ve been building in our own business lately.
Garbage In, Garbage Out — Sorry, Jan
If you’ve been following along, you’ll know I’ve been working with my own AI assistant, Jan. This week she stuffed up the monthly reports. Not happy, Jan! But if I’m being honest, it was my fault. I didn’t train her properly.
That old adage — garbage in, garbage out — applies just as much to AI agents as it does to any other system. Jan picked up stale tasks and pulled in information that had no business being in a monthly report. That’s not a model failure; it’s a training failure.
Here’s the thing: with a staff member, if they get things wrong two or three times, you start to question what’s going on. But with an AI agent, it’s almost always the user who is the problem. There is always a workaround — you just have to find it.
That’s why I spend time every single week going back through Jan’s skills: how she accesses information, how she responds, how she writes, and which models are powering her at any given time. It’s a long-term investment, and I’m treating it like one.
Agentic Systems Are the New Business Asset
I’m seeing incredible agentic systems popping up everywhere right now, and it’s making me more bullish than ever on an AI-first future for commerce and e-commerce. We’re not heading toward a world where people sit at a screen browsing for products — agents will do that for us. We’re already seeing it in the traffic data.
Here’s the mindset shift I’d encourage every business owner to make: your CRM, your client base, your recurring revenue — those are your traditional assets. But what are the agentic processes and systems that take your business to the next level? That’s the new asset class.
Sure, developing these systems takes time. I’m working late after dinner training Jan, investing in the future of the business. But I know that the time I put in now means more capacity for the things I’d normally never get to — like going deep on client projects and building tools I genuinely couldn’t have built before.
Open Source Is Closing the Gap Fast
Here’s something worth paying attention to: I’ve got a model running locally on my own hard drive, for free, that is the equivalent of an OpenAI model from about a year ago. The open-source movement has genuinely caught up, and when you combine that with falling hardware costs, running powerful AI in your business is more accessible than ever.
Everything is converging toward what people are loosely calling AGI. Personally, I think we’ve already passed it in some respects. The next conversation is about artificial superintelligence, and people are taking that seriously. But regardless of where the frontier goes, the tools that exist right now can have a real impact on your business today.
The Jevons Paradox: Saved Time Gets Filled
Harvard Business published a piece last week confirming what I’ve been saying for about 18 months: the time we save with AI gets filled with extra tasks. The Jevons Paradox is alive and well in the AI era. We are busier than ever.
But here’s the upside: that freed-up time is letting me go deeper for clients, into work I previously didn’t have the bandwidth to take on. And on that note — big news — our app has just been approved by Shopify. More on that coming soon.
It’s the same shift we saw when typists disappeared and everyone started typing themselves. These tools don’t remove work, they change the nature of it.
Claude vs ChatGPT — Where Things Stand
There’s a growing number of people stepping away from OpenAI and ChatGPT right now, and Sam Altman is copping a lot of heat. In a timely move, Anthropic — the team behind Claude — set up a way to import your full memory from ChatGPT directly into Claude. For a lot of people, that memory was the main reason to stay with ChatGPT.
For me, having spent so long working in Claude Code, I genuinely haven’t missed it. I’ve barely used ChatGPT once or twice this year, yet I’m still paying for a Plus subscription. That’s getting cancelled soon — nothing to do with any boycott, it just doesn’t make sense to keep paying for something I’m not using.
My current setup is open-source models running locally for the everyday, high-frequency tasks (no token costs, no latency issues), and frontier models like Claude Code and Gemini for the heavier work when I need them. If you haven’t looked at Claude Code yet, I’d encourage you to. And if you’re not particularly technical, Claude Cowork is worth a look too.
Over to You
Would you hire an AI agent for your business? What would you pay $10,000 a month for if it worked around the clock? Drop a comment below — I’d love to hear your thoughts. You can reach out to me at jim@stewartmedia.biz.
As always, if you found this useful, please like, share, subscribe, and tell your friends. See you next week.
The post Would You Hire an AI Agent for $10k/Month? appeared first on StewArt Media.