The One-Person Back Office: Running Support, HR, Training, and Knowledge Without a Department of Any of Them

Somewhere between employee five and employee fifty, your company quietly acquires four departments — and hires none of them.

Customer support, HR, training, and knowledge management all show up, and they live in the margins of somebody's actual job: the founder answers tickets, the office manager became "HR" the day someone needed a leave policy, training is "shadow Sarah for a week," and the knowledge base is Sarah.

I've spent twenty years inside these functions at enterprise scale — contact centers, QA organizations, training departments with real budgets — and the most useful thing those years taught me is what actually requires the department and what only requires the system. This playbook is the cross-functional answer: the operating pattern, the stack principles, and where the deep-dive playbooks live for each function.

The pattern: all four jobs are the same job

Strip the vocabulary and every back-office function runs the identical loop: intake → triage → respond → document → improve. A support ticket, an employee question about parental leave, a training request, a "how do we do X" — each one arrives, gets prioritized, gets answered, should get written down, and should make the next occurrence cheaper.

This matters because it means you don't need to learn four professions. You need to install one loop four times, with different nouns. And it means the failure mode is also shared: the loop breaks at document every time, in every function, because answering feels urgent and writing down feels optional. The whole craft of the one-person back office is making the document step cheap enough to survive a busy Tuesday — and that's exactly the step AI just changed (more below).

Function by function: the 20% that does the 80%

Support. One queue (not inbox-plus-DMs-plus-text), a written triage rule, five great response templates with a human seam, and a deflection loop where every recurring question becomes a doc that kills the next ten tickets. The full operating model is in the CS team-of-one playbook; the template system is here.

HR. You are not building enterprise HR; you are building defensible, humane basics: clean hiring (a job description that attracts — playbook), pay transparency where the law already requires it (state-by-state), onboarding that's written down, and documentation habits that protect both sides. The survival map is the HR manager-of-one playbook.

Training. Skip the course catalog. Define the behavior you need changed, build the minimum that changes it, and measure at thirty days whether it stuck (the no-consultants measurement stack). Most "training problems" at small scale are actually documentation problems wearing a costume — which is convenient, because you're already fixing those in the next function.

Knowledge. Two rules carry the whole function: the second time anyone asks a question, the answer gets documented where the third asker will find it; and every doc gets an owner and a review date, or it joins the graveyard. When someone irreplaceable resigns, there's an emergency protocol for that — but the second-ask rule is what makes it rarely necessary.

The stack principles (tools change, these don't)

One source of truth per noun. Customers live in one place, employees in one place, docs in one place. The moment a noun lives in two systems, you've hired yourself as the sync engine, and you don't show up reliably. Before adding any tool, the question isn't "is it good" — it's "which existing source of truth does it replace or read from?"

AI drafts, humans send. The single biggest unlock for a team of one, applied identically in all four functions: AI writes the first 70% of the ticket reply, the job description, the training scenario, the SOP-from-transcript — and a human adds the judgment, the specifics, and the send. Volume was always the reason these functions needed departments. Drafting was most of the volume. (How to tell tools that actually work this way from tools that sprinkled AI on the brochure.)

Document on the second occurrence. Same rule, every function — second ticket on a topic, second policy question, second "how do I." It builds exactly the library people need, in the order they need it, with zero documentation projects.

Weekly digest, not dashboards. A team of one doesn't have time to visit four dashboards, and won't. Invert it: each function reports to you, once a week, in one email — what happened, what needs a decision, what can wait. If a tool can't summarize itself proactively, it's an unstaffed dashboard, which is a decoration.

Buy the loop, build the glue — almost never the reverse. The intake-triage-respond-document loop is commodity; your judgment layer is not. The build-vs-buy playbook covers the decision in full, but the short version: every hour spent maintaining homegrown ticket plumbing is an hour not spent on the 20% of work that actually required you.

The honest part

This pattern — same loop, four nouns, AI on the draft step, digest instead of dashboards — is the thesis TranscendByDesign is built on: CSByDesign, HRByDesign, LearningByDesign, and KnowledgeByDesign are the four functions as software, designed for the person running all four at once. That's the pitch, and you should discount it accordingly.

But notice what this playbook actually asked you to install: a triage rule, five templates, a behavior sentence, a second-ask rule, an owner on every doc, a weekly summary. None of that requires our software or anyone's. The companies that drown don't drown for lack of tools — they drown because answering felt urgent and documenting felt optional, four times a day, for three years.

Closing

You can run a credible back office alone: one loop, four nouns. Support gets a queue, a triage rule, and templates with a human seam. HR gets defensible basics and honest job postings. Training gets a behavior sentence and a 30-day pulse. Knowledge gets the second-ask rule and owned docs. Everything gets AI on the draft step and a human on the send button, reporting to you weekly instead of waiting on a dashboard you'll never open.