1What it does
A Make AI Agent checks supermarket stock rows against a purchasing policy, emails purchase requests to the owner and records its decision in the sheet. The owner approves every purchase.
2Who it is for
- Industry
- Retail
- Platform
- Make
- Category
- Operations
- Tools used
- Make AI AgentsGoogle SheetsGmailClaude (setup only)
3How it works
Every step, in order, is in Workflow at a glance, pinned on the right as you scroll.
Setup (one time)
- 01 Setup
Use Claude with the Google Drive connector only for setup: create the stock list, test data and purchasing_policy.md.
- 02 Setup
Upload purchasing_policy.md with approved suppliers, delivery days and payment terms. Set knowledge search query to let AI decide and limit to 50.
- 03 Setup
Give the agent Gmail Send an email, named send email to owner. Fix the recipient to the owner. Let AI decide the subject and raw HTML body.
- 04 Setup
Give it Google Sheets Update a Row, named update stock row. Map the row number directly and set last check date with formatDate(now; YYYY-MM-DD). Update exactly once per item after any email is sent.
Runtime
- 01 Trigger
Daily schedule at 7:00 am in the video. The description says 8:00 am.
- 02 Action
Search Rows in the stock sheet where needs_reorder = yes. The sheet uses a min-max method and includes stock, reorder level, packs to order, order value, priority, MD's approval and last check date.
- 03 AI / action
Pass row fields into a medium-size Make AI Agent with job-description instructions covering role, goals, steps and rules.
- 04 Action
Skip delisted items and flag missing approved suppliers. Never place orders or contact suppliers.
- Trigger
- Daily schedule at 7:00 am in the video. The description says 8:00 am.
- AI step
- The full instruction text is not read out in the video. The author puts the instructions, tool descriptions and policy file in the video description. Instructions cover the role, goals, steps and rules.
4How it went
The test sent per-item purchase requests, flagged sliced bread as urgent and sent an action-needed email for an item without an approved supplier. A rerun with no yes rows took no action. These are demo records, not business results.
Only what the source reports. The source did not report numbers.
5Worth knowing
- BrightMart in Wuse 2, Abuja is the supermarket example. Claude is used for setup only.
- The owner approves every purchase. Fixing the email recipient and row number keeps those decisions outside the AI.
- The demo rice row had 3 on hand, a reorder level of 5 and 13 packs to order, worth about 1.14 million naira. Kubwa Grains and Rice Depot delivers Mon/Wed/Sat and requires bank transfer before delivery.
- The email initially showed no currency. The author fixed the sheet with a Nigeria custom currency format.
- Resolve the 7:00 am video schedule versus the 8:00 am description when setting the daily run.
6Recipe
Adapted from source Reworded or reconstructed from what the source shows.
Adapted setup: Search Rows in stock where needs_reorder = yes. Map row fields into the agent and upload purchasing_policy.md. Fix send email to owner to the owner's address; let AI decide subject and raw HTML content. For update stock row, map the row number directly and set last check date to formatDate(now; YYYY-MM-DD). Tool rule: update the item's row to record the decision and supplier exactly once, after any email. Skip delisted items, flag missing suppliers and leave every purchase for owner approval.