This week Anthropic introduced "Routines" for Claude Code — structured, repeatable sequences of AI actions that can be triggered on demand. Within hours it became one of the most discussed threads on Hacker News.
Why the excitement? Because the tech community is finally naming something practitioners have known for a while: a single prompt doesn't cut it.
The Problem With the One-Shot Prompt
Since ChatGPT captured mass attention in 2022, thousands of companies have followed the same arc: initial enthusiasm, early experiments, eventual disappointment. The disappointment almost always hits at the same moment — when you try to use AI not for a discrete task, but for an actual business process.
Take a concrete example: handling an inbound customer inquiry.
- Read and classify the request from the inbox
- Look up the customer record in the CRM
- Assess priority and urgency
- Determine the right owner or team
- Route the inquiry and send an acknowledgement
- Log the interaction
Six steps. Each one touches a different system. Each one depends on the output of the previous. With a chatbot or a manually-fired prompt, you might get through step two.
The rest stays manual.
What Routines Solve — And What They Don't
"Routines" in Claude Code are a smart step forward: define a sequence of actions, save it, run it repeatedly. That's genuinely more productive than one-off prompts. For developers running the same code-review workflow every day, it's a real improvement.
But in an enterprise context, the concept hits its ceiling quickly.
Routines are static. They execute the same sequence every time, regardless of what they discover along the way. A real business process needs to adapt: if the inbound inquiry is from a strategic account, the workflow changes. If a required CRM field is missing, the agent needs to handle it — not fail silently.
Routines don't know your business. What's your lead qualification threshold? Which accounts are considered strategic? What are your SLAs by customer tier? A routine has no idea. An agent trained on your business context does.
Routines have no memory across steps. Step 4 of a routine doesn't know what step 2 found — unless you engineer it explicitly. An agent runs the entire workflow as a single coherent unit, carrying context from start to finish.
Routines are macros. Agents are colleagues.
What This Means for Your Business
The conversation around Claude Code Routines is a signal that the industry is crossing an important threshold. People are starting to think in processes, not prompts. That's the right frame.
But routines are the first step on that journey — not the destination.
The destination is a system that:
- Executes a complete process from trigger to outcome, autonomously
- Adapts to unexpected inputs without breaking
- Knows and applies your specific business rules
- Documents every decision with traceable reasoning
- Runs around the clock without requiring human oversight at each step
That's not a chatbot. It's not a routine. It's an agent.
Why Nolen Was Built Exactly for This
Nolen was never designed as a prompt tool. The question at the center of every design decision has always been: what actually needs to get done?
A qualified lead in the CRM. A customer inquiry answered and routed. A pipeline status updated. A support ticket classified and assigned.
These aren't answers — they're outcomes. And Nolen is priced per outcome: not per prompt, not per token, not per session.
When the tech world celebrates "routines" as an innovation, we're genuinely glad. It means the market is moving in the right direction. And it signals just how much potential remains — for companies willing to take the next step.
Want to see a Nolen agent take over one of your actual processes? We'll show you what's possible in 30 minutes — no prior AI experience required, no IT project needed.