Morning Prep & Inventory Agent
A daily prep list built from your actual sales patterns and tomorrow's weather.
Every evening after close, this agent pulls the day's sales data and the next day's weather forecast, compares it against the last several weeks of the same weekday, and produces a specific prep list for tomorrow morning — how many pastries to thaw, how much milk to pull, whether to bump cold brew production for a 78-degree Thursday. It also watches your rough inventory counts and flags anything trending toward a stockout before it becomes a 7am scramble. Instead of your opener eyeballing yesterday's numbers and guessing, they walk in to a printed or texted list with actual numbers behind it — built from your shop's real patterns, not generic advice.
What this agent takes off your plate
- A specific tomorrow-morning prep list instead of an opener's best guess
- Prep quantities that adjust automatically for weather and day-of-week patterns
- Early warning on ingredients trending toward a stockout, days before it's a crisis
- Less over-prep waste from "just make extra to be safe" habits
- A written record of prep decisions you can review and adjust monthly
The stack
Real costs & ROI
- Claude Pro or ChatGPT Plus$20/moneeded for the daily automated run — the free tier's message limits get tight with a nightly job plus normal use
- Make.com Core plan (or Zapier Starter)$9–20/mothe free tier's ~100 tasks/month isn't enough for a daily multi-step automation; Make's Core plan starts around $9/mo
- Google Sheets$0/mofree with any Google account; stores your sales history and prep log
- Weather forecast$0/moWeather by Zapier is a built-in trigger with no API key; Make users can call Open-Meteo (free, no key) or type tomorrow's forecast into the close-out form
The math
Opening prep guesswork costs money two ways: over-prep that gets thrown out at close, and under-prep that means a 9am sellout and a customer who doesn't come back. Even a modest reduction in waste plus a couple of avoided sellouts a month is worth $150–300/month for a typical single-location shop — often more than the $20–40/mo this costs to run. On top of the dollars, your opener gets a specific list instead of pattern-matching from memory, which saves 20–30 minutes of morning decision-making, worth another $10–15/day in labor efficiency. This is the one agent in the pack where the tool cost is real, but the payback shows up directly in reduced waste, not just saved time.
Build steps
Build your day-of-week sales baseline
The agent needs to know what "normal" looks like for each weekday before it can tell you how tomorrow should be different.
Data you need
Export 60–90 days of hourly/daily sales by item from your POS if it supports it — otherwise even a rough manual log of "how many of X did we sell" for a few weeks of each weekday works.
I want to build a day-of-week sales baseline for prep planning. My cafe: [CITY, format — full cafe/grab-and-go/hybrid] My pastry/food items: [LIST] My sales data for the last 60–90 days (paste export or describe pattern by weekday): [PASTE OR DESCRIBE] Please analyze: 1. **Average units sold per item, broken out by day of week** (Monday sells differently than Saturday — I want both) 2. **A "safe range" per item per day** — a low estimate (light day) and high estimate (busy day), not just an average 3. **Which items are most weather-sensitive** (cold brew on hot days, hot drinks on cold days) vs. which are stable regardless of weather 4. **Which items have the tightest sell-through** (sell out fast, worth slightly over-prepping) vs. **which trend toward leftover** (worth slightly under-prepping) 5. **A starting prep-quantity table** by item and day of week that I can hand to whoever's setting the automation up next Keep the output as a table I can drop into a spreadsheet — item, Monday through Sunday baseline quantities.
Set up your daily data feed
The agent needs two fresh inputs every evening: today's sales by item and tomorrow's forecast. Pick the sales path that matches your comfort level — all three land in the same Google Sheet, so you can start simple and upgrade later. Path A — closer's form (simplest, works with any POS): a Google Form called "Close-out" with a number field per tracked item (croissants sold, gallons whole milk used, cold brew batches, etc.) plus a "sold out early?" and "thrown out at close" field. The closer fills it in 2 minutes. Form responses land in a sheet tab automatically. Path B — scheduled POS report email: Square (Dashboard → Reports → Item Sales → schedule/export) or Toast (Reports → Scheduled Reports) can email a daily item-sales report. Forward that email to a Zapier "Email Parser by Zapier" mailbox, teach the parser which numbers to grab, then Trigger: Email Parser — "New Email" → Action: Google Sheets — "Create Spreadsheet Row". Set this up once; it runs nightly. Path C — POS API via Make (advanced): Make.com HTTP module → Square Orders API (or Toast API) → Iterator → Google Sheets "Add a Row". Only worth it if you're comfortable reading API docs or have someone who is. Weather: no API key needed. In Zapier, "Weather by Zapier — Tomorrow's Forecast" is a built-in trigger — give it your zip code and a run time (8pm). Make users can call Open-Meteo (free, no key) with an HTTP module, or just add a "tomorrow's weather" field to the close-out form. Sheet layout: one "Baseline" tab (the table from Step 1), one "Daily" tab (date, day of week, per-item sales, waste notes), one "Prep log" tab (what the agent recommended, filled by Step 4).
Data you need
You'll need admin access to your POS reports, a Google account for the form and sheet, and your zip code for the forecast. No API keys for Paths A or B.
Agent Pack
Unlock the Coffee Shop Agent Pack.
One payment unlocks all 3 blueprints in this pack — build steps, the complete system prompt, guardrails, a test checklist, and the maintenance routine — for $29.
one-time, unlocks all 3 blueprints in this pack on this browser. Secure checkout via Stripe.