How to Run a One-Person Business Without Drowning in Admin

Nobody starts a company because they enjoy admin. Run a business alone for a month, though, and the arithmetic is uncomfortable: the work you started the company to do is the minority of your week.

This is about shrinking the majority without hiring.

Four jobs, not one pile#

"Getting organised" hides four distinct problems that fail in different ways:

A tool that fixes triage does nothing for memory. This is why founders end up with five apps and the same problem.

Why it costs more than the time it takes#

Answering an email takes two minutes. Being interrupted to answer one costs considerably more.

Gloria Mark's group at UC Irvine followed office workers and timed what happened after interruptions. The widely cited figure from that work is roughly 23 minutes to return to the original task — and people typically handle about two intervening tasks before getting back to it (Mark et al., CHI 2008). Her later work found something equally relevant to solo founders: a large share of interruptions are self-inflicted — nobody pinged you, you went and looked.

So "be more disciplined about admin" misdiagnoses it. The problem is not that admin is slow. It is that admin is interleaved with work that needs a running start.

Why unfinished admin follows you around#

There is a second, quieter cost. Masicampo and Baumeister's experiments found that unfinished goals intrude on unrelated tasks — participants interrupted before finishing reported intrusive thoughts and did worse on work requiring concentration.

The useful half of that finding: the intrusions largely stopped when participants wrote a specific plan for finishing. Not finishing — planning (Masicampo & Baumeister, 2011, JPSP).

That is the mechanism behind every "get it out of your head" system. It also explains why a vague list makes things worse: "sort invoices" is not a plan, so it keeps intruding.

The rule for what to hand off#

The line is not boring versus interesting. It is whether there is a right answer.

Hand off when you would recognise the correct output instantly: the reply you have written forty times, pulling last quarter's numbers together, a reminder that a client never came back, filing a receipt where it belongs.

Keep pricing, hiring, whether to take the client, what to build next. Software offering to make those calls is selling you something.

Most people get this exactly backwards — automating the judgment, which produces confident wrong answers fast, while hand-doing the mechanics forever.

Four fixes, in order of payoff#

1. Triage the inbox in blocks, not continuously. Checking email in a limited number of daily blocks measurably lowers stress (Kushlev & Dunn, 2015), and it is the interrupt source for everything else. The inbox guide goes deeper.
2. Never walk into a call cold. What did we last agree, what did I promise, what changed. That is a lookup, not a decision — the highest-return five minutes in the week.
3. Close the follow-up loop. The quietest and most expensive failure. Why deals die in the inbox.
4. File as you go. Fifteen minutes weekly beats two days in January. No clever fix exists; attach it to something you already do.

Where a Mac assistant fits#

Everything above is triage and memory, which is what an on-device assistant is for. Jarvis is the one we make: it runs on your Mac, drafts inside your mail client rather than a chat window you copy out of, remembers your people and projects between sessions, and can run a routine overnight. Free and open source.

Honest limits: Mac-only, 2.0 is in beta, and it will not decide what to charge.

The rule worth keeping#

If you have done it three times the same way, it should not need your attention the fourth time.

It is also a decent filter for tools: if it cannot take over something you have already done three times, it is not solving your problem.