The math nobody does before the next admin hire

Start with what the hire actually costs, because the salary line is where most of the math starts and stops.

Whatever number goes in the posting, a list of costs rides on top of it. Employer payroll taxes. Benefits, if you offer them, and pressure to offer them if you don't. A laptop, a desk or a stipend, a seat on every piece of software your team touches. The recruiting cost — either an agency fee or your own evenings reading resumes and your best people's afternoons running interviews. Then the ramp: months where you pay full price for partial output while someone senior spends real hours training, answering questions, and checking work.

It doesn't end when they're up to speed, either. Every employee carries a permanent management tax: one-on-ones, reviews, PTO coverage, the coordination overhead of one more person in every loop. None of these line items is exotic. You already know every one of them. The point is that almost nobody adds them up before the posting goes live.

Take an illustration, not a statistic. Suppose the posting says $50,000. Add your actual payroll tax rate, your actual benefits cost, your actual tool stack, and an honest estimate of the hours your team will sink into hiring and training. You will land meaningfully north of the sticker, and you're the only one who can say how far north. Now notice the most important property of that number: it recurs. Every year. With raises. A hire is a subscription, and the price only moves in one direction.

Now decompose the role, because "admin" describes a bundle of different jobs, not one.

Here's the exercise I ran on my own half-written posting. Write down what the person would do on a real Wednesday in month three. Not the aspirational bullet points from the job description — the actual day. Mine came out something like: move data from the inbox into the CRM, chase unpaid invoices on a schedule, assemble the same weekly report from the same four sources, book meetings, answer the same dozen customer questions, keep two systems in sync that refuse to talk to each other.

Read a list like that honestly and a pattern jumps out. Most of it is rule-following. Information moves from one place to another under conditions you could write down on a single page. In the roles I've decomposed inside my own companies, that routine core came out to well over half the job. The remaining slice was genuine judgment: the angry customer, the weird edge case, the call on what matters this week.

That split is where the real math lives, because the routine core is exactly what AI agents and workflow automation handle well. We run systems like this inside our own companies. The inbox gets triaged, the report assembles itself, the invoice chasing happens on schedule without anyone remembering to do it.

So run your own numbers. Take the fully loaded figure from the last section and multiply it by the routine share of the role you're hiring for. That product is the annual, recurring, raise-adjusted price of work a system could carry. The honest question was never whether you can afford the hire. It's whether the judgment slice, the part only a person can do, justifies a full seat once the routine slice is off the table.

Sometimes it is, plainly, and pretending otherwise would turn this piece into a sales letter.

Hire when the role is mostly the judgment slice. Some seats are almost entirely relationships, negotiation, taste, or physical presence — a site foreman, a senior bookkeeper who catches what software can't, your first salesperson. Automating the thin routine layer of a judgment-heavy role saves you little and delays a hire you actually need.

Hire when the process doesn't exist yet. Automation hardens whatever you feed it. If the workflow changes shape every week because the business is still finding its footing, a good generalist who can invent the process is worth more than any system, because there's nothing stable to build on. Automate after the process settles, not before.

Hire when you're buying a future, not a queue. If this person is meant to grow into running the team, the routine work is partly training. That's a legitimate trade, made with open eyes.

And often the honest answer is both, in sequence. Automate the routine core first, then look at what's left. Sometimes the judgment slices from three different half-roles combine into one excellent full-time job — a job that's far easier to hire for, because nobody's second month is thirty hours of copy-paste. I've hired more people since installing these systems in my own companies, not fewer. The difference is what the jobs look like.