Your First Claude Code
Routine in Under 60 Minutes
Automate One Weekly Task This Afternoon
A complete walkthrough for shipping your first Claude Code Routine: an automated workflow that runs on a schedule, an API call, or a repo event. See the same Monday morning briefing Routine Symphony AI runs internally, then build it for your business in nine focused steps.
Written for operators, owner-operators, and ops leads. No developer required.
Symphony AI | Strategic AI Consulting
What We Build
Custom AI Agents
Production-ready agents built around your workflows.
Workflow Automation
Routines, MCP servers, and pipelines that run themselves.
Hands-On Training
Your team learns to ship Routines, not just use them.
No hype, no hallucinations, just systems that ship.
What Claude Code Routines Are
A Claude Code Routine is a saved instruction set that runs on a trigger. The trigger fires (Monday at 7 AM, an inbound webhook, a new commit), Claude wakes up, follows your instructions, calls whatever tools it needs, and posts the result wherever you tell it to. No babysitting, no chat window, no clicking Run.
The Dumb Timer
Cron Job
- Same script every run
- Breaks when data shifts
- No language understanding
- Fails silently
The Rule Mover
Zapier Workflow
- Moves data between apps
- One trigger, one fixed path
- Pricey at any scale
- Breaks on API changes
The Operator
Claude Code Routine
- Reads, decides, reports
- Adapts to messy inputs
- Connects to anything via MCP
- Explains what it did
What a Routine Actually Looks Like
A Routine is a single file with two parts. The top is configuration: what triggers it, what tools it can use, where to send output. The bottom is plain English instructions Claude follows. That is it. No drag-and-drop builder, no node graph, no premium tier to unlock features.
Where Routines Beat Everything Else
- Tasks where the inputs are messy, varied, or written by humans.
- Workflows that need a real summary, not a copy-paste of source data.
- Anything where the cost of getting it wrong is small and the cost of skipping it is large.
- Recurring work that takes 30+ minutes and you keep meaning to delegate.
See a Routine in Action
Symphony AI runs on its own Routines. Before the first one shipped, Monday mornings started with the founder reading 47 messages across four Slack channels and a KPI sheet to figure out what mattered that week. Here is what one Monday looks like now that the briefing Routine is live.
The Timeline
7:00:00 AM Mon
The schedule fires; the Routine wakes up.
7:00:14 AM Mon
Claude pulls 47 messages from #client-ops and #eng-alerts since last Monday.
7:00:38 AM Mon
It reads the Q2 Pipeline Google Sheet for MRR, deal value, and churn deltas.
7:00:52 AM Mon
A formatted briefing posts to #monday-briefing tagged @founder.
7:14 AM Mon
The founder reads it over coffee, walks into the week with full context.
Before the Routine vs. After the Routine
Before
- Time gathering data
- 90 minutes
- Briefing arrives
- Wednesday
- Coverage
- 3 of 8 channels
- Cost per run
- 1.5 hrs founder time
After
- Time gathering data
- 14 seconds
- Briefing arrives
- Monday 7:01 AM
- Coverage
- All channels + KPI sheet
- Cost per run
- $0.34
The Result
In 8 weeks, Symphony AI’s founder has reclaimed 12 Monday mornings and never walked into a week blind again. You are about to build the same Routine for your business.
Find Your First Automation
Symphony AI’s first Routine was the Monday briefing because that was the meeting prep that ate our Sunday nights. Yours will be different. Spend 15 minutes with the audit below before you write a single line of config.
“What recurring task takes 30+ minutes and we keep meaning to delegate?”
The 5-Question Pick-Your-Routine Audit
How often does this task happen? Weekly is a great starting cadence.
How many minutes does it take each time? Multiply by frequency.
What sources does it pull from? Slack, Sheets, CRM, and email all qualify.
What does the finished work look like? A message, a draft, a list, or a summary.
If the Routine gets it 90% right, is that good enough on the first pass?
Top 5 Starter Routines (Ranked by Time Saved)
Pulls Slack activity and KPIs into a Monday briefing. Saves ~90 minutes a week.
Reads QuickBooks aging, drafts Gmail follow-ups for founder review.
Checks Notion calendar, pings owners for posts due within 48 hours.
Compares CRM stage changes, posts wins, losses, and stalls to the sales channel.
Surfaces clients silent 60+ days, drafts tailored re-engagement emails per account.
The Three Triggers, Translated
Every Routine fires on one of three triggers. Most operators only need the first one for years. Here is what each trigger does and when to reach for it.
Schedule
Cron expression. Best for recurring digests, weekly reports, daily check-ins.
Monday 7 AM = 0 7 * * 1
API
A webhook URL. Best for "do this when X happens in another tool."
POST https://routines.claude.com/r/<id>
GitHub Event
Repo push, PR opened, issue labeled. Best for teams that ship code.
on: pull_request.opened, label = ready-review
Pro vs Max: Which Tier for a 10-Person Team
Three daily Routines across 22 work days = 66 runs per month. Five daily Routines = 110 runs. Both fit comfortably inside the Pro tier. The math only flips to Max when you start running hourly cadences or wire Routines into customer-facing API endpoints.
Claude Pro | $20/mo
- 5 Routine runs per day
- ~150 runs per month
- Right for: 3-5 daily or weekly Routines
- Right for: most 10-person teams
Claude Max | $200/mo
- 50 Routine runs per day
- ~1,500 runs per month
- Right for: hourly cadences
- Right for: customer-facing API trigger
Lay the Foundation
Before You Start
- Claude Code already installed and logged in (one-time setup, ~10 minutes)
- Active Claude Pro or Max subscription (covered on page 5)
- Admin access to your Slack workspace and KPI Google Sheet
- A folder on your computer where Routines will live
- 60 minutes of focused time
You are going to build the same Monday Morning Business Digest Symphony AI uses internally. Every step is a prompt you paste into Claude Code; the only typing required is filling in your channel and sheet names.
Open Claude Code in Your Routines Folder
Use Finder or File Explorer to create a new folder anywhere on your computer. Symphony AI calls ours "symphony-routines." Open Claude Code and point it at this folder when it asks where to start. No commands required.
Ask Claude to Write Your Routine
Paste the prompt below into Claude Code. Edit the bracketed parts to match your business. Claude creates the Routine file with the right structure and a starter draft for you to refine.
Prompt
Create a Routine called monday-briefing. Trigger: schedule, Monday 7 AM America/New_York. Pull last 7 days from Slack [#client-ops, #eng-alerts] and KPIs from Google Sheet [Q2 Pipeline]. Compose a briefing with three sections: Wins, Blockers, and KPI Movement. Post to [#monday-briefing], tag [@founder]. Write the file and show it to me when done.
Read Claude's Draft and Refine the Prompt
Claude shows you the draft. Read it carefully. Anything vague or missing? Tell Claude exactly what to tighten in plain English. Examples: "Limit Wins to 5 bullets" or "Always lead with the KPI that moved the most." Two or three rounds is normal.
Connect Your Data Sources
A Routine is only as useful as the data it can reach. Steps 4, 5, and 6 connect your Routine to Slack and Google Sheets, then nail down how the briefing should look. Total time: about 25 minutes.
Connect Slack as a Source
Tell Claude to add Slack. A browser tab opens for Slack's approval screen; pick the workspace and grant access only to the channels the Routine actually needs.
Prompt
Add the Slack connector to this Routine. Grant read access only to [#client-ops, #eng-alerts]. Walk me through the approval step by step.
Connect Your KPI Sheet
Same pattern as Slack. Claude installs the Google Sheets connector and walks you through Google's approval flow.
Prompt
Add the Google Sheets connector. Grant read-only access to [Q2 Pipeline] only. Walk me through the approval.
Format the Output the Way You Want
The default briefing works, but yours should match how you actually consume info. Tell Claude exactly what the final briefing should look like.
Prompt
Update the briefing format. Three sections: 1. Wins (max 5 bullets, lead with revenue) 2. Blockers (only items with no clear path forward) 3. KPI Movement (lead with the biggest delta) Keep total under 200 words.
Schedule, Test, and Deploy
The last three steps take about 13 minutes plus one Monday of patience. By the end of this page, your Routine is live and you know it works because you have already watched it run.
Verify the Schedule
Ask Claude to confirm the trigger. Specifically check the timezone: a default of UTC fires at the wrong time in your local zone, and "Monday 7 AM" by mistake means 2 AM ET.
Prompt
Read the trigger in monday-briefing. Confirm it fires Monday at 7 AM America/New_York. If anything is off, fix it and show me the change.
Test-Run the Routine
Tell Claude to do a dry test first. Read the draft briefing Claude shows you. If it looks right, give the green light to ship the real one to Slack.
Prompt
Run monday-briefing as a dry test. Show me the briefing it would post, but do not send to Slack yet. After my review, I will tell you to ship it.
Activate and Monitor the First Real Run
Tell Claude to turn on the schedule and add a fallback so silent failures never go unnoticed. The first three Mondays, read the briefing critically and refine the prompt; by week 4 it should feel like it was written by your sharpest analyst.
Prompt
Enable monday-briefing on its schedule. Add a fallback: if any run fails, post the error to [#ops-alerts].
Your Routine Is Live
Total time: roughly 60 minutes. From now on, every Monday at 7 AM, your briefing arrives without you.
Make It Smarter
Your first Routine works. Now make it better. These three upgrades are the same ones Symphony AI rolled out on its own Monday briefing in the first 90 days, plus the next two Routines we shipped after the briefing proved itself.
Upgrade 1
Add a Memory File
Point the Routine at a CLAUDE.md file that captures recurring context: people, channels, KPI definitions, and last quarter’s priorities. Briefings get sharper week over week instead of resetting every Monday.
Upgrade 2
Build the Next Two Routines
Once the briefing is stable, ship the Nightly Overdue-Invoice Chase and the Weekly Content-Calendar Nudges. Same pattern, new connectors, 30 minutes each.
Upgrade 3
Add a Human Approval Step
For Routines that send anything outside the company, like invoice chases or re-engagement emails, add a draft-only-then-post-to-Slack-for-approval step before delivery. The Routine still does the work; a human still owns the send.
Measure What Matters
Routines are easy to ship and easy to forget. Track these five metrics from week one to know whether yours is delivering value or quietly drifting.
Hours Saved Per Week
The clearest ROI metric. Multiply minutes saved by frequency. The Monday briefing alone reclaims ~90 minutes per week.
Run Success Rate
Of scheduled runs, how many completed without an error. Anything below 95% means you have a data-source or auth problem.
Time to First Value
How long between the trigger firing and the result landing. Routines should clear this in under a minute.
Cost Per Run
Total monthly Claude tier divided by Routine runs. Monday briefing costs Symphony AI $0.34 per run.
Operator Trust Score
A 1-to-5 weekly rating from the person who reads the output. If it drifts below 4, refine the prompt before the trust dies.
ROI Quick Calculation
When Routines Win, When They Do Not
Claude Code Routines, n8n, and Zapier all automate work, but they win in different lanes. Use this table to pick the right tool the first time instead of rebuilding in month three.
| Use Case | Routines | n8n | Zapier |
|---|---|---|---|
| Summarize messy text | Best | Weak | Weak |
| Move data between known apps | Fine | Best | Best |
| Draft an email a human reviews | Best | Weak | Weak |
| Strict if-this-then-that | Overkill | Best | Best |
| 8,000+ pre-built integrations | No | No | Best |
| Self-hosted control + cost | No | Best | No |
| Setup in under 60 minutes | Best | Fine | Best |
Quick Decision Guide
If your task is "read messy stuff and summarize it":
Routines. Nothing else handles unstructured text this well.
If your task is "when X in app A, then Y in app B":
Zapier or n8n. Routines work but you are paying for intelligence you do not need.
If you want one tool for both kinds of work:
Start with Routines and call n8n through an API trigger when you need it.
If cost matters more than convenience:
Self-hosted n8n. Routines and Zapier both come out cheaper at low volume; n8n wins at scale.
Your Next Three Routines
Your first Routine is live. Here is the order Symphony AI recommends shipping the next three, plus the one we should build with you when you are ready to wire Routines into your full stack.
QuickBooks aging in, Gmail drafts out. Founder approves with one click.
Notion calendar in, Slack pings out. Posts due in 48 hours get owner-tagged.
CRM stage changes in, sales-channel summary out. Wins, losses, and stalls at 8:55 AM.
The One Symphony Should Build With You
When you are ready to wire Routines into your CRM, billing, and ticketing with proper auth, audit logs, and human-approval gates, that is a custom engagement. Three to five connected Routines, full guardrails, and your team trained to own them.
Symphony AI
Strategic AI Consulting
From Prompt to Power.
You shipped your first Routine. Imagine five of them running your operations.
Symphony AI wires Routines into your full stack with auth, audit, and guardrails. No subscriptions, no lock-in, just systems that ship.
Book a Strategy Call