Stop Hiring Humans for Work a Script Can Do (Hire Them for the Rest)

A coach we audited was paying her VA $20 an hour, ten hours a week. We looked at the task list. Four of those hours were: copy new client details from the intake form into the CRM, copy the same details into the invoicing tool, download podcast stats into a spreadsheet, and move signed contracts into the right Google Drive folder. That's $4,160 a year for data transfer between systems that have APIs. A script does all four jobs in seconds, for free, without typos, at 2 a.m. if needed.

To be clear about what this post is not: it is not anti-VA. A good assistant is one of the best hires a solo operator makes. This is anti-misallocation. When you pay a human to be middleware, you get the worst of both options — human costs and error rates, on work with software-grade repetitiveness. And you waste the one thing the human actually brought: judgment.

The cost math, honestly run

Say your VA spends five hours a week on mechanical transfer work at $18/hour. That's $90 a week, about $4,700 a year. The equivalent automation — a few Zapier workflows or a small custom script — costs somewhere between $0 and $50 a month to run, plus a one-time build. The automation pays for itself in the first quarter, then keeps paying forever.

But the hourly rate is the smaller half of the bill. The rest:

• Error rate. Humans doing repetitive transfer work make mistakes at a low but nonzero rate — a transposed digit in an invoice, a lead filed under the wrong name. Each error costs cleanup time and occasionally a client's confidence. Scripts do the same thing every time. For this task category, that's the whole point.
• Latency. Your VA batches the CRM updates on Thursdays. So from Friday to Wednesday, your pipeline data is wrong. An automation runs the moment the trigger fires.
• Management overhead. Copy-paste work still needs instructions written, questions answered, and spot-checks done. You're spending founder hours supervising work a cron job would do silently.
• Fragility. When the VA is sick or leaves, the invoicing stops. A process that breaks when one person takes a vacation isn't delegation. It's a single point of failure with a salary.

The split test: one question per task

Take your delegation list — everything you currently hand off or wish you could — and run each task through one question:

Does this task have fixed inputs, fixed outputs, and no judgment calls in the middle?

If yes, it's software work. Form data into CRM. Invoice generation from a signed proposal. Receipt filing. Weekly metrics pulled into a dashboard. Appointment reminders. Renaming and filing documents. If you could write the steps on one page and a stranger could execute them perfectly on day one, a script can too — and it should. Whether it needs a dumb rule or a smarter agent is a separate question (here's how to tell), but either way, it's not a human's job.

If no, it's human work. The tells: the word "depends" shows up when you explain it, taste matters, or a relationship is involved. Calming an unhappy client. Choosing which podcast guests are worth pitching. Noticing that your best client has gone quiet and figuring out why. Planning the retreat. Writing that isn't templated. This is work where a good person gets better over time — the opposite of script work, where the goal is to never vary at all.

Most delegation lists we see split roughly 40/60: forty percent software work being done by humans, sixty percent genuinely human. That forty percent is where the money leaks.

Re-scope the role upward

Here's the part people miss: automating the copy-paste work is not a reason to cut your VA's hours. It's a chance to upgrade the job description you probably under-wrote in the first place.

The coach from the opening kept her VA at ten hours. The four reclaimed hours went to work that actually compounds:

1. Exception handling. The automations run; the VA reviews the weekly log and handles the oddballs — the intake form with a typo'd email, the payment that didn't match an invoice. Ten minutes instead of four hours, and more valuable than the original job.
2. Client experience. Personal check-in messages, remembering the client whose launch was this week, sending the occasional handwritten card. Things a script should never do.
3. Judgment work she'd been hoarding. First-pass podcast guest research, drafting replies to non-standard inquiries, flagging which testimonials were strong enough to publish.

Same spend, radically better output. Her VA, for the record, preferred the new job — nobody's career goal is being a slow API.

How to actually make the switch

Don't announce a grand reorganization. Do this instead. Ask your VA to keep a simple task log for two weeks — same audit you'd run on yourself. Sort it with the split test. Pick the two most frequent software-work tasks and automate just those, keeping the VA as the fallback for the first month while trust builds. Then move down the list, and backfill every reclaimed hour with judgment work from the pile you've been meaning to delegate for a year.

The end state is worth naming plainly: software doing the work that has one right answer, your VA doing the work that benefits from a person, and you doing only the work that requires the founder. That's what delegation was supposed to mean before it turned into paying someone to move text between browser tabs.