Running a shop, café or small team: 7 real cases

Design and build a restaurant terrace instead of paying €10,000+ with Codex

Quotes for a summer terrace at Be Fresh, the small Vienna restaurant run by Alex Gavrilescu's mother, started at €10,000 and ran as high as €20,000. Gavrilescu, a software engineer rather than a carpenter, gave Codex the permit's size and position rules, then asked it to search stores around Vienna for materials the family could actually buy. Codex redesigned the terrace around those parts, produced a shopping list with quantities and where to buy, and wrote 'IKEA style instructions' for cutting and assembly. It also suggested adjustable supports for the uneven street and rust-resistant screws. The build took five days, including shopping, and city inspectors approved it on the first visit. The terrace seats eight and roughly doubles the restaurant's space.

Scale catering recipes for 25–30 people up to hundreds with ChatGPT and Claude

Brian Reilly, co-owner of Arlington's Catering and Historic Venue in Springfield, took a six-week small-business AI workshop sponsored by Innovate Springfield. Reilly now uses ChatGPT and Claude to make routine checklists, such as cleaning procedures, and to scale recipes built for 25–30 people up to, say, 330. Reilly says accuracy depends on the input. When Reilly fed it information inaccurately, those mistakes were caught 'pretty quickly', and now Reilly screenshots the recipe into the tool instead to avoid user error. 'It might take a time or two, but it's been pretty accurate.'

Turn a retailer's carton labeling rules into a label file with Gemini

Jamie Kim, owner of the granola company Jamie's Farm in New York, uses Gemini to work through big retailers' dense delivery-guideline PDFs, where a mistake can mean major chargebacks. When one retailer required every shipping carton to be labeled with its number and destination warehouse, Kim asked Gemini to help generate a custom CSV file for the label template and point to online label-making software, turning a full day's work into one hour. Gemini in Sheets also helped write formulas for a spreadsheet tracking inventory, sales and cost of goods.

Have Claude draft replies to a small brand's customer messages, warranty claims included, with a human approving each one

idopmstuff runs several small consumer brands. For the most recent one they acquired, customers got in touch through Facebook and Instagram comments and messages as well as email, so they had Claude build an app that pulls those messages in as they arrive and drafts a reply, including any actions needed, for them to edit or send. Their edits feed back into its playbook, and it has got pretty good. It handles warranty claims (ask for a photo of the damage and the order number, create a replacement order, email a confirmation) and most common questions. Every reply is still approved by a human, though they plan to let it handle common cases on its own soon.

Run a restaurant's back office through an OpenClaw agent: supplier price-change alerts, invoice uploads and a fairer tip split

Christopher Thépot runs restaurant operations in Madrid and spent a month using OpenClaw. The agent answers point-of-sale questions over WhatsApp and combines supplier invoices, sales data, recipes and inventory counts to track waste and what's missing. It flags supplier price spikes; two pricing mistakes caught so far directly affected the P&L. It also pulls supplier invoices from their inbox into the accounting software's scanner, saving one to two hours a week, though they're still fine-tuning that. They used it to design and test a new tip policy in under an hour: 80% stays with the shift and 20% goes to a pool split by hours worked.

Turn a photo of a cafe's daily inventory sheet into shopping lists for three suppliers with a custom GPT

Cole Morgan runs a small cafe with no single major supplier, so they drive to three stores. Staff fill in two paper inventory forms each day and upload them to Slack, and turning those into a checklist for each store took 2–10 minutes. An off-the-shelf OCR tool that worked well cost over $250 a month, and a Python script they had ChatGPT write wasn't accurate enough. So they built a custom GPT, 'C&K Inventory Manager', trained on a spreadsheet of every item and its vendor. Now they upload the sheet photo and get a checkbox list grouped by vendor, ready to paste into Obsidian. It asks about items missing from the sheet, such as whole milk. Accuracy improved with GPT-4o with no change to the instructions, though it sometimes asks too many questions and they have to tell it to just output the list.

Auto-coach sales reps from call recordings with Zapier Agents

According to Zapier's customer story, Healthie's call-summary and coaching agents save each of about 20 reps and CSMs 2–3 hours a week, more than 60 hours a week across the team.

According to Zapier:60+ hours a week saved