The Big Idea
Anthropic dropped Routines in Claude Code yesterday. If you run an education business — course creator, coach, consultant, teacher — this is the single biggest change to agent automation since scheduled tasks shipped last year.
Here is the shift in one sentence:
Scheduled tasks are alarm clocks. Routines are employees who pay attention.
A scheduled task runs at a set time and does the exact same thing every time. A routine waits for an event — a booking, a homework submission, a community question, a course consumption spike — and then reads multiple systems before deciding what to do.
That is the difference between rigid automation and a real AI employee.
Why This Matters for Educators
You have been told for years that you can’t scale personalization. One-on-one teaching was capped at a handful of students. Past that, you hit cohorts, then courses, then static libraries — and every step lost the personal touch that made people pay you in the first place.
Routines change the math. The same trigger — a student books a coaching call — can now produce a completely different brief for every student based on their course progress, community posts, past sessions, and support tickets. The AI employee does the prep work before you walk into the room.
Three outcomes this unlocks:
- Higher completion rates. Routines watch consumption in real time. When a student stalls on lesson four, the routine reaches out while they’re stuck — not three days later in the analytics report.
- Personalized coaching at scale. Every session starts with a custom brief. Works for 1-on-1 and cohorts alike.
- Context-aware community. A member asks a question, a routine fires, reads their history, and sends a personalized answer instead of a generic link.
Breaking Down the Concept
Here is the side-by-side breakdown so you can feel the shift:
| Dimension | Scheduled Task (Old) | Routine (New) |
|---|---|---|
| Trigger | Clock (runs at a set time) | Event (runs when something happens) |
| Response | Pre-written, same every time | Generated, unique per person |
| Context | Runs blind | Reads CRM, LMS, community, support tickets first |
| Judgment | If this, then that | Decides based on what it reads |
| Output | Structured, pre-determined | Analysis before output |
| Metaphor | Alarm clock | Employee who pays attention |
“A scheduled task employee was dead. It couldn’t think other than what it was taught. A routine is a go-getter that can even think outside the box to get stuff done.”
Local vs Remote Routines
Two flavors inside Claude Code:
- Local routines — only run while your computer is awake. Good for testing or tasks you want to supervise.
- Remote routines — run 24/7 in the cloud. No need to leave your Mac or PC on. This is where real AI employees live.
One more note on where they live. Inside Claude Code they are called “Routines.” Inside the Cowork tab they are still called “Scheduled.” Same feature family, different naming.
Apply This Framework: 9 Use Cases for Educators
Here are nine places a routine can quietly out-perform a scheduled task in your business. Each one follows the same pattern: event happens → routine reads context → personalized action.
1. Coaching Call Bookings
Before: Everyone who books gets the same calendar invite and the same prep PDF. Zero personalization.
After: The routine pulls their LMS progress, community activity, past session notes, and open support tickets. You walk in with a brief written just for them.
2. Homework Submissions
Before: You open the session with “show me what you worked on.”
After: The routine reads the submission, summarizes it, and hands you an analysis before the call starts. Works for 1-on-1 and cohorts.
3. CRM Events
Before: FluentCRM captures logins, course starts, email opens, page visits — but the data just sits there.
After: Every contact record becomes agent input. The routine can act on a single event or a pattern of events (“three logins this week but no lessons completed”).
4. Zoom Session Endings
Before: Manual upload, manual transcript, manual follow-up.
After: Recording finishes → routine triggers → transcript processed → notes drafted → follow-up queued. All automated.
5. Course Consumption Watch
Before: You see “only spent 30 seconds on lesson 4” in the analytics report next week. Too late.
After: The routine sees it in real time and reaches out with a DM, email, or pop-up: “Hey, are you stuck on lesson 4?”
“Most people get stuck trying to fix stuff after it’s happened. Routines fix stuff while it’s happening — personally, for each person.”
6. Community Questions
Before: “Great question, here’s a link to the FAQ.”
After: Routine reads the question, reads who asked it, and sends a specific answer plus three next steps tailored to where they are in the program.
7. Downloads
Before: You have no idea if they did anything with the PDF you sent.
After: Routine tracks the download, waits a few days, checks for follow-through, and nudges them if they went cold.
8. Payments & Subscriptions
Before: Scheduled reminders, dunning sequences, generic receipts.
After: Routines handle each event — new sub, failed payment, cancellation — with context-aware responses.
9. YouTube Workflows
Before: Publish a video, hope people find it, hope they act on it.
After: Routine takes the transcript, generates repurposed content, and then watches how each viewer engages so you can follow up personally.
The Bigger Shift: Don’t Sell Courses. Sell Context.
This is the part most people are going to miss.
The old model — build a course library, charge for access, hope the content sells itself — is commoditized. Static courses are everywhere. AI can generate a decent one in an hour.
What can’t be commoditized is context. Your LMS knows where your students are stuck. Your CRM knows what they’ve opened. Your community knows what they’re asking. Your support tickets know where they’re failing.
Routines let you sell the personalized experience that context enables — not the static lesson library. That is what your audience can still pay premium prices for in 2026.
“You don’t worry about the calendar. You worry about the individual person.”
Your Next Steps
- Open Claude Code and find the Routines tab on the left sidebar. Create one local routine just to feel how triggers work.
- Pick one event in your business that currently produces a generic response (a booking, a homework submission, a new member joining). That is your first routine candidate.
- Map the context that routine should read before it acts. What does the CRM know? What does the LMS know? What does the community know?
- Build it as a remote routine so it runs 24/7 and your computer can sleep.
- Watch your completion and engagement metrics for 14 days. Routines are a completion-rate lever — the numbers will tell you which one to build next.
Join the Campus
If you want to build an agent-powered education business — not just read about it — come talk to the people who are actively building theirs.
Free community: trainingsites.io/join — hang out with other educators, coaches, and course creators figuring out what to build first.
Agent Builders membership ($197/mo): trainingsites.io/pass/ — live biweekly agent builds, every course in the library, and a front-row seat every time a new routine demo drops.
Expect the best. — James
