DEV Community: RoutineKit

Pretty Notion templates won't run your week for you

RoutineKit — Sat, 22 Aug 2026 12:10:50 +0000

If your Notion looks beautiful and your week still feels scattered, the problem usually isn't "more pages."

It's that tasks, content ideas, client follow-ups, and cash notes live in four different places — so every Monday you reopen the same question:

What am I actually looking at this week?

ChatGPT is open in another tab. That doesn't help if nothing tells you which prompt belongs to which row.
Content: Name · Channel · Status (Idea → Hook locked → Draft → Edit → Scheduled → Published) · Publish date · Hook · CTA · Proof · URL.

Proof is required on purpose: a real experience, number, or case for every piece. Cheapest anti-hallucination guard there is, because you fill it before asking an LLM to draft anything.

Clients: Name · Stage (Lead / Meeting / Proposal / Quote / Contract / Active / Collecting / Closed) · Next action · Next date · Amount hint · Temperature. Next action is a verb with an implied deadline — "Thu send quote," not "follow up."

Money: Name · Type (In / Out / Expected in / Expected out) · Amount · Date · Category · Proof. Weekly cash sense only — not tax filing, not a replacement for real books.

Rebuild in ~20 minutes (no live duplicate link)

A lot of "Notion templates" assume a one-click duplicate.

That fails when you want the structure inside your own workspace, under your own permissions, without inheriting someone else's clutter.

So the practical approach is:

Create a root page: Weekly Ops Board
Add the four databases with explicit properties
Add weekly views (This Week Board / Publish Pipeline / Touch this week / This Week Ca ## A simple weekly rhythm

Monday: 25m board cleanup + pick today's 3
Wednesday: 40m content slots
Friday: 20m light cash sketch
Sunday: 25m review, lock next week

110 minutes. No second brain museum. No 40 linked databases.

What a wired prompt looks like

Role: Weekly ops coach for a solo operator. Goal: from the Inbox/Doing list below, pick exactly 3 tasks to finish today.

Rules: select exactly 3, label everything else defer / delegate / delete, prefer P0 and near deadlines and revenue-critical work, and keep the sum of est. minutes at or under 80% of available time.

Output: today's 3 with est. min and a one-line why, a defer list with reasons, and which rows to flip to "Today" in Notion.

Two details make it usable: the 80% capacity rule (without it the model hands you a 9-hour day), and that last output line — every AI turn has to end as a row edit inside the board, not a paragraph in a chat window you'll never reopen.

Also: replace only the placeholders. Rewriting the Rules section is how you get mush. And check facts, money, dates, and names before anything goes out — AI output is a draft.

Soft CTA

If you want the full Markdown schema + prompt slots + a filled example week as a downloadable pack, I published it as Solo Weekly Notion Ops Board + AI Prompts ($9):

https://routinekit.gumroad.com/l/vnlmtc

Prefer to try before buying? Free sample (Tasks schema + 3 prompts):

https://routinekit.gumroad.com/l/alxurc

Unofficial template — not affiliated with or endorsed by Notion Labs, Inc. Not financial or tax advice.

— RoutineKit

I'm the maker. Questions about property choices or "how do you avoid Notion bloat?" welcome in the comments.
sh)

Paste a small prompt library (pick-today's-3, stuck-task breakup, content batch, Friday cash sketch, Sunday review)
Replace example rows with your real week

The dashboard page is just a callout (this week's 3 goals + a do-not list) plus those four linked views. One scroll shows tasks, people, content, money. That's the whole point.

A minimum weekly schema (not an aesthetic dashboard)

For solo founders and freelancers, I've found four databases are enough:

DB	Job
Tasks	What ships this week
Content	Publish pipeline
Clients	Next action, not a full CRM
Money	Cash in / cash out sketch

Two shared properties do most of the work:

Week — Select: Last week / This week / Next week
AI Prompt — Text: which prompt slot you used on this row (T-1, C-2, ...)

That Week select is the trick. Every dashboard view filters on Week = This week, so the board empties and refills itself instead of growing into a graveyard. And the prompt slot means the model isn't a random chat — it's wired to the board.

Skip relations on day one. You can add them in week three if you still want them. (You usually don't.)

The properties that actually earn their keep

Tasks: Name (start with a verb — "Send quote") · Status (Inbox / Today / Doing / Waiting / Done / Canceled) · Priority (P0-P2) · Due · Area (Sales / Delivery / Content / Ops / Learning) · Est. min · Actual min.

Waiting separates "I'm blocked on me" from "I'm blocked on someone else's reply" — completely different problems on a Friday.

Stop Quoting "Around $500": A 5-Part Freelance Ops Loop You Can Run With AI

RoutineKit — Sat, 22 Aug 2026 11:47:24 +0000

Most freelance money leaks don't happen in the work. They happen in the paperwork loop around it: the quote you typed into a chat window, the "tiny tweak" you never priced, the receipts living in your camera roll until tax week.

Here's the loop I run, plus the prompts I paste into ChatGPT/Claude at each step. Not legal or tax advice: these produce drafts, and a human still checks before you sign or file.

1. Quote: itemize or lose the argument

"Around $500" is not a quote. It's a future dispute. A usable quote has line items, exclusions, a revision count, and a validity window.

Write a client quote.
Project: {{one-line scope}} | Deliverables: {{list}}
Rate basis: {{hourly/day/fixed + number}}
Output: (1) itemized table, hours and price per line,
(2) "Included / Not included" two-column list,
(3) revision policy: {{N}} rounds, extras priced,
(4) validity: {{14}} days.
No filler adjectives. Client must approve or reject each line.

Before: "Website redesign - $2,000." After: 6 line items, 3 exclusions, 2 revisions, expires in 14 days. Discount requests now hit a line item instead of your self-esteem.

2. Light agreement

A $2k gig doesn't need 12 pages. It needs scope, payment schedule, IP transfer on final payment, kill fee, late-payment terms. Ask the model to draft it from the quote in under 700 words, and to flag in brackets anything you must confirm with a lawyer. That bracket rule is what stops it from inventing enforceable-sounding clauses.

4. Books: 10 minutes a week

One sheet: date, client, category, amount, tax-relevant y/n, note. Paste the week's receipts, let AI categorize, you correct. Ten minutes weekly beats a lost weekend in April.

5. Advisor prep

Before a CPA call, generate a one-pager: entity type, revenue range, top expense categories, three specific questions, three things you're unsure about. You'll get more out of 30 minutes than most people get from three meetings.

The pattern

Every step is the same shape: SOP, then a prompt with real variables, then a filled example, then a pre-send checklist. The variables do the work; empty ones give you generic mush you'd never send.

I packaged this loop out as templates:

Free sample (the full quote module): https://routinekit.gumroad.com/l/alxurc
Freelancer Quote, Contract & Tax Prep Kit, $15: https://routinekit.gumroad.com/l/opcrcu
Solo AI workflow kit, $12: https://routinekit.gumroad.com/l/lsbdil

Start with the free one. If the quote module doesn't change how your next proposal looks, the rest won't either.

Which step in your ops loop still eats a whole evening?

3. Change requests: a script, not a mood

"Can you just..." is where margin dies. One template: acknowledge, cite the scope line, then two options - swap for something in scope, or add it at {{price}} and +{{N}} days - then ask for a yes/no. Two options reframes it from "are you charging me?" to "which one?"

You Use ChatGPT. Your Workload Didn't Shrink. Here's a Weekly OS.

RoutineKit — Sat, 22 Aug 2026 10:34:48 +0000

You already use ChatGPT daily. Yet the week still fills up: proposals eat the evening, content stays at "I'll post this week," replies get uneven, quotes start from a blank page. More prompts don't fix that — cadence does. Here's the weekly OS I run solo.

Slot A — Client proposal (~45 min)

Give the model a spine: restate the client's problem in their words; the outcome in one measurable sentence; in scope / out of scope / assumptions; timeline by phase with a deliverable each; price + payment split. Prompt shape: role = senior consultant writing to a busy non-technical client; constraints = under 400 words, no jargon, one price, explicit out-of-scope list. The win isn't typing speed, it's never re-deciding the structure.

Slot B — Content batching (60-75 min weekly)

Dump 10 raw ideas from real work. Pick 3. Write each as claim + proof in one line. Let the model expand the middle; rewrite the first and last paragraph by hand. Then schedule. If it isn't claim + proof, it isn't a post yet.

Slot C — Inquiry replies + scope fences (30-40 min)

Good fit: confirm the problem, propose a 20-minute call, name your price range up front.
Unclear: ask exactly two questions, no more.
Not a fit: decline in three lines, refer someone.

Fence I reuse: "That's outside this phase — happy to quote it as a separate add-on."

Slot D — Quote + simple SOW (40-50 min)

Deliverables, revision count, change-order trigger, payment schedule, validity date. Template it once; only numbers change.

Slot E — Weekly review (25-35 min)

What shipped? What slipped, and why? What are next week's five slots? Lock the calendar before closing the laptop.

An LLM removes typing friction, not decision friction. Fixed slots remove the deciding.

The packaged version

Every slot as SOP + prompts + filled example + checklist: Solo Operator AI Weekly System — $17
Narrower pieces: AI Workflow Kit — $12 and Quote, Contract & Tax Prep Kit — $15 (not legal advice). I'm the maker; questions about running slot B or C on a real week are welcome in the comments.