The web shop that didn't want to hear it
I sat down with a business owner last week whose company builds websites as part of what they offer. A few of her staff were on the call too, the folks who actually do the building, and most of their work is in WordPress. I told her I thought they needed to migrate off it.
The staff pushed back right away, which was fair given they know the tool and build in it every day. "What CMS would you use instead?" one of them asked.
I said one of the modern ones, or just build it yourself. That last part got a look I've gotten before.
The old way is already over
A website today isn't something you build once and hand over. It's something that should be watched, updated, and improved continuously. In an agent-driven world you can stand up a service that scans your client's site every day, looks at competitors, tracks news in their field, drafts content, and suggests improvements it can make itself. You review and approve, and it ships. That's a different product than a WordPress theme with a support retainer, and it's a different business model.
WordPress isn't going away this quarter. It still powers around 42% of the web. But the tools that matter for agent workflows have already picked sides. Sanity shipped an MCP server that plugs their CMS directly into Claude Code and Cursor, and Cloudflare released EmDash this year and positioned it as the spiritual successor to WordPress for AI-native work. Industry analysts have pointed out that AI agents, left to their own devices, essentially never pick WordPress. They pick the modern options because those are the ones they can actually work with.
None of this is secret, but the knowledge just isn't evenly distributed yet. That's the window. The web shops that see it now have a year, maybe two, to reposition. The ones that don't will end up as cheap labor for clients whose agents do the work instead.
Run this on yourself
Take any job description in your company and paste it into Claude. Ask how much of that job could be done by Claude or by an agent that Claude builds for you. I've run this exercise on a lot of roles, and my bet is 75% plus on most knowledge work. Deloitte puts the current number at 40% of tasks across knowledge workers, and McKinsey says 60 to 70% of time is affected. Whatever number you get, it's going to be bigger than you expected, and bigger than your people expect.
Before anyone panics, I'm not telling you to fire your team. Some jobs will get eliminated and that's real. But the pattern I keep seeing in my own companies and my cohorts is that AI intensifies work. Jobs get bigger and wider and more demanding. Harvard and MIT researchers have landed in the same place: workers with agentic tools are taking on a wider slate of tasks and working longer hours. When writing gets faster, teams write more. When analysis gets faster, more analysis gets requested. The ceiling rises.
Once you're 10x more productive, you do 10x more work. Companies with the right setup are going to explode in capability, and that's what's about to happen across every industry over the next couple of years.
What becomes more valuable
There are still bottlenecks where AI can't help and a human has to be in the seat. Relationships are one, and distribution is the other. Trust with a customer, a partner, a referral network, a community. Getting your product in front of the right people and being known for something. These are the things agents cannot compress on your behalf, and they are the things that will go up in value as everything around them gets cheaper.
That's where I'd invest the hours the agents give back. Don't spend them worrying about what an agent can take from you. Spend them on the work only a human can do.