- Automate first whatever is repeated, predictable, and costing you the most: for most small businesses that's lead follow-up, appointment reminders, invoicing, review requests, or reporting.
- The filter is a 2x2: how often a task runs, and how much judgment it needs. Repeated-and-predictable automates first; rare-and-judgment-heavy stays human.
- One workflow is the right starting size. A cattle farm turned a 2.5-hour weekly inventory chore into one 31-minute phone call, roughly 104 hours a year, from a single automation.
- Don't automate a broken process: automation makes a workflow faster, not smarter. Fix the steps on paper first, then wire them up.
The best first automation is almost never the flashiest one. It's the chore you'd be embarrassed to admit how much time you spend on: the invoice chasing, the same reminder text, the lead that sat until Thursday. McKinsey's automation research found that in about 60% of jobs, at least a third of the tasks could already be automated with existing technology. This post is the filter for finding which third of yours goes first. If you're still deciding whether automation pays at all, start with the ROI math and come back.
What makes a workflow worth automating first?
Two questions decide it: how often does the task run, and how much judgment does it need? Plot every chore in your week on those two axes and the answer stops being a debate:
The top-right quadrant, repeated and predictable, is the only place a first automation belongs. Those tasks have written-downable steps and run weekly or daily, so the build pays for itself on volume. The judgment-heavy quadrants aren't off-limits forever; they just come later, with a human approval step wired in, once the easy wins have paid for the plumbing.
Which five workflows pay first for most businesses?
Across the small businesses I work with (home services, farms, professional practices), the same five candidates keep winning. The tell for each one is a symptom you already recognize:
| Workflow | The symptom you have it | What the automation does |
|---|---|---|
| Lead follow-up | Inquiries answered "when I get a minute," which is tomorrow | Instant reply + booking link the moment a lead lands |
| Appointment reminders | No-shows, and mornings spent confirming by hand | Automatic text/email sequence before every appointment |
| Invoicing + payment chasing | You are the accounts-receivable department, at 9pm | Invoices generated and politely chased on a schedule |
| Review requests | Happy customers, thin Google profile | A well-timed ask after every completed job |
| Recurring reports | Copy-paste from three systems into one spreadsheet | The numbers assemble themselves and arrive by email |
If several apply, pick the one closest to revenue. A no-show costs you a slot; a lead that goes cold costs you the whole job.
How do you actually find yours?
Don't guess from a list, audit a week. The procedure is short:
Track one honest week
Every time a task repeats, write it down with a minutes count. No fixing yet, just counting.
Filter with the 2x2
Keep only the repeated, predictable ones. If you couldn't hand it to a new hire with written steps, it's not first.
Rank by bleed
Multiply minutes by frequency, and weight anything that touches leads or cash flow. The top of that list is your first automation.
Fix the process on paper
Clean up the steps BEFORE wiring anything. Automating a broken process just makes mistakes faster.
Build one, measure, expand
Prove it with a before-and-after number, then let workflow two reuse the same plumbing.
That fourth step is the one everyone skips. Automation is an amplifier: point it at a clean process and it multiplies the output, point it at a mess and it multiplies the mess.
What does a good first automation look like in practice?
A cattle farm I built for had a classic top-right-quadrant chore: weekly inventory meant 90 minutes in a freezer reading box weights to a helper, then another hour of someone retyping it all into spreadsheets. Repeated weekly, completely predictable, three people involved. We replaced it with one phone call to a voice agent that writes everything to Google Sheets and Square directly: 2.5 hours became 31 minutes, on the first run, and the owner got roughly 104 hours a year back from that single workflow.
Notice what made it the right first pick: it ran every week, the steps never changed, and nothing about it needed judgment. Weighing a box is not a decision. That's the pattern to hunt for in your own week.
Starting with one workflow isn't just less risky, it's cheaper: a single-purpose automation is a small fraction of what a full multi-agent system costs, and the accounts and integrations it sets up get reused by everything you add later.
If you want the audit done for you, book the 15-minute call and walk me through your week. I'll tell you which workflow I'd automate first and what it would cost, and if the honest answer is "clean up the process and spend nothing yet," that's the answer you'll get.
