Stop Pressing Go

57%.

That's how much of the average workday McKinsey* says can be automated with technology that already exists. Not in five years. Today.

Most people are still pressing go every time. The gap between 57% and what most businesses are actually automating isn't the technology. It's the schedule.

*McKinsey: a major business consulting firm that studies workplace trends.

The florist whose orders fill her inbox while she sleeps
A florist in Perth hands every after-hours inquiry to an AI receptionist. It handles questions, confirms bookings, and sends details while she's home. By morning, her inbox has confirmed orders she never had to process. No staff staying late. No missed leads at 11 PM. How she put it: the difference isn't the AI. It's that she stopped needing to be there when it ran.

The invoice watcher that checks email every four hours
A small business owner built an agent that reads his email every four hours. Not when he remembers. Not when his accountant asks. Every four hours. When an invoice arrives, it pulls the amount and due date and logs it to a spreadsheet. He's touched the system once since setup. His accountant called it the tidiest books she's seen from a one-person operation. Around 5 hours saved a week. One timer.

90 minutes of tab-switching, replaced by a 7 AM digest
A content creator describes what her morning used to look like: open Reddit, scan for relevant threads, switch to X, search her topics, check Hacker News, take notes, close 14 tabs. Ninety minutes, most days. She set up a scheduled agent to do the same sweep every morning at 7 AM and deliver a ranked digest to her inbox before she wakes up. She hasn't opened Hacker News on purpose in six weeks.

The consultant who got her Mondays back

A marketing consultant with 8 clients used to spend every Monday the same way. Log into each client's analytics dashboard. Pull the week's numbers. Compare to last week. Write a plain-English summary. Format it. Send it. Ninety minutes per client, times eight. Twelve hours. Every Monday, for three years.

She built a scheduled agent. Every Friday at 5 PM, it pulls numbers from each client's analytics and ad accounts, runs the week-over-week comparisons, and writes a draft summary in plain language. By Monday morning, eight reports are sitting in a folder. She reads each one, adjusts anything that needs context, and sends.

Review time: about 20 minutes per client instead of 90. Eight clients: under three hours instead of twelve. Same deliverable. Same quality. Nine-plus hours back every week.

She used the time to take on two more clients. Same schedule. New revenue.

The agent runs every Friday at 5 PM whether she's thinking about it or not. That's what scheduling does that manual work can't: it doesn't depend on you remembering.

This pattern applies to any professional who delivers regular reports on the same data: bookkeepers, financial advisors, property managers, PR firms. If you send the same type of summary to the same clients every week, a scheduled agent can draft it. You review. You send.

Week 6 of 6: Stop pressing go

Last week you ran a research workflow manually. You typed a prompt, Claude browsed and wrote, you got a file. This week you make it run automatically, on a schedule, without you ever touching it again.

Open Claude Code in your project folder. Give it this prompt, filling in your day and time:

At your chosen time, the workflow runs on its own. The output file will be waiting.

You don't need to understand how scheduling works. That's why you have Claude.

Stuck? Reply to this email. I'll help.

Six weeks. You went from opening a terminal for the first time to running a workflow that does real work while you sleep. That's the arc. Now the training wheels come off.

Next issue: the agent that finds and pitches new client work on my behalf while I'm not watching. Not a demo. The real one. What it decides, what it gets wrong, and what it sent while I was asleep.

Manu