Content is one recurring job. Your business has a dozen others — the monthly numbers, the customer replies, the launch checklist, the quarterly listing refresh. The ops layer turns those from "things you re-figure-out every time" into Standard Operating Prompts: saved, tested, run on a schedule. Same idea as an SOP in a real company — except the procedure is a prompt.
The SOP template
Every Standard Operating Prompt has the same five parts. Write it once, test it twice, save it where you'll find it. Then you never rebuild that task from scratch again.
SOP: [task name] GOAL: [what "done" looks like, concretely]. INPUTS: [the files/sources it may use — read-only unless noted]. STEPS: 1) … 2) … 3) … RULES: [never-do list; match my Context Pack; ask when unsure; draft only]. DELIVERABLE: [what to produce, named how, saved where]. REPORT BACK: [what to tell me + anything it wasn't sure about].
Four SOPs, written out
Tap any one to open the full prompt, copy it, and swap in your specifics. These four cover most of a solo business's recurring admin.
1 Monthly numbers reviewMonthly · read-only · ~1 hr saved
Turns a spreadsheet export into a plain-language read of the month — what moved, what to watch, one thing to try.
SOP: Monthly numbers review GOAL: A one-page read of last month's numbers + one thing to try next. INPUTS: my metrics export at [file] and my Context Pack goals doc. Read-only. STEPS: 1) Summarize the key numbers vs the prior month. 2) Name what moved and a likely why. 3) Suggest ONE focused change for next month tied to my #1 goal. RULES: Don't invent numbers — if a figure's missing, say so. No vanity metrics. Plain language. DELIVERABLE: A one-page summary in the chat (or a doc "Numbers — [month]"). REPORT BACK: The headline, the one number I should worry about, the one thing to try.
2 Customer-reply packOn demand · drafts only · ~40 min saved
Drafts on-brand replies to your common customer questions — never sends, always in your voice.
SOP: Customer-reply pack GOAL: Draft replies to the messages I paste, in my voice, ready for me to send. INPUTS: the messages below + my Context Pack (voice, products, standards). STEPS: 1) Group by type. 2) Draft a warm, correct reply each. 3) Flag any that need a real decision from me. RULES: DRAFT ONLY — never send. Match prices/product names to my Products doc exactly. If a question is sensitive or unclear, flag it, don't guess. DELIVERABLE: A labeled list of draft replies I can copy. REPORT BACK: Which ones you're confident in, which need my eyes. MESSAGES: [paste]
3 Product-launch checklist generatorPer launch · planning · ~2 hrs saved
Generates a dated, channel-by-channel launch plan from your product details — so nothing gets forgotten.
SOP: Product-launch checklist generator GOAL: A dated launch checklist for [product], from prep to post-launch. INPUTS: my Context Pack + these product details: [name, price, promise, launch date, channels]. STEPS: 1) Work backward from launch day. 2) List tasks by phase (build, warm-up, launch week, follow-up). 3) Note which channel each touches and a suggested date. RULES: Realistic for a one-woman team. No busywork. Flag anything I likely can't finish in time. DELIVERABLE: A dated checklist I can drop into my planner. REPORT BACK: The critical path, the riskiest date, what to cut if I'm short on time.
4 Quarterly listing refreshQuarterly · proposals only · ~2 hrs saved
Rereads your product or service pages and proposes refreshes — flagging stale copy and weak hooks without touching the live listings.
SOP: Quarterly listing refresh GOAL: A refresh plan for my listings — what's stale, what's weak, 2 title options each. INPUTS: my listings at [links/file] + my Context Pack (voice, products, prices). STEPS: 1) Read each listing. 2) Flag stale info, weak hooks, unclear benefits. 3) Draft 2 alt titles each. RULES: PROPOSALS ONLY — never edit or publish live listings. Keep prices/facts unless I confirm a change. DELIVERABLE: A doc "Listing refresh — [quarter]" with flags + title options. REPORT BACK: The weakest listing, the strongest new title, anything factually out of date.
Agents in production: the weekly roster
An SOP you run by hand is good. An SOP that runs on a schedule is an agent (you met these in the AI Agents guide). Graduate your best SOPs into a weekly roster — and keep a manager's review cadence so you stay the boss.
| When | Agent | Your review |
|---|---|---|
| Monday | Morning / weekly brief | 2-min read |
| Wednesday | Content repurpose (the splinter pass) | Voice QA before publish |
| Friday | Follow-ups & loose ends drafted | Read every draft, then send |
| Monthly | Numbers review + bookkeeping prep | Sign-off before anything's final |
An agent roster without review is how businesses drift into "metric theater" (Lesson 5). Keep it simple: a daily glance at scheduled outputs, a weekly voice check on anything published, a monthly sign-off on the numbers. Delegation, not abdication.
The automation triangle
Automations sound intimidating until you see the shape. Every one is the same triangle: a trigger → an AI step → a destination. Something happens, AI drafts a response, the draft lands somewhere for you to approve. Three worked flows — verify the exact tool steps in your automation layer, since connectors change:
| Trigger | AI step | Destination |
|---|---|---|
| New subscriber joins | Draft a personalized welcome in my voice | Saved as a draft for me to approve & send |
| New sale comes in | Draft a thank-you + a gentle review-ask | Draft queued; I click send |
| Weekly metrics land | Summarize into one dashboard row | Appended to my tracking sheet |
The AI step drafts; it doesn't send. Every flow above ends in a draft or a row you review — nothing goes to a customer or spends money without your click. Build the triangle so the last edge is always yours.
- Pick the one recurring admin task you most dread. Write it up with the SOP template.
- Test it twice on real inputs. Fix the prompt, not just the output, each time it's off.
- Decide: does it stay a run-by-hand SOP, or graduate to the weekly roster?