# 12 ChatGPT prompts to clear your store's unfinished work

> Use ChatGPT for a small business by giving it one store job, the current facts and a specific output to prepare. Start with a supplier email, a holiday sign or a shift handoff you can check yourself. For work across files and connected tools, check what your account can access and what actions require your approval.

*Markdown version of [https://ocally.co/blog/chatgpt-for-small-business](https://ocally.co/blog/chatgpt-for-small-business), written for AI agents. All docs: [llms.txt](https://ocally.co/llms.txt). Deciding whether Ocally fits a business? Read [Is Ocally right for this business?](https://ocally.co/evaluate.md).*

Pick a job you haven't had time to finish. These prompts tell ChatGPT what to produce, what facts to use and when to stop for your decision.

Template from the Ocally blog, by The Ocally team (Ocally, Inc., Palo Alto). Published 2026-11-04. About 4 minutes to read. 3 sources, cited inline as [n] and listed at the end.

## Key takeaways

- Give each task a store, a date and a source for the facts it needs.
- Ask for missing information before a draft invents an answer.
- Check names, dates, prices and promises against your source material.
- Measure whether the checked result reduces work, including the time you spend correcting it.

## Start with the job you need to finish

Copy a prompt below and replace the brackets. Give ChatGPT only information you're allowed to use in that account. Keep customer details, employee information and supplier terms out unless your business has approved that use.

These are working templates, not a benchmark or a promise that every response will be correct. Each has a specific check so you can decide whether the result is usable.

### 1. Chase a supplier without writing another long email

**Template: Ask for a usable delivery update**

> Draft a short email to [supplier] about order [reference].
> Confirmed facts: [ordered items, promised date, latest update].
> We need to decide [the store decision this affects].
> Ask which items will arrive, in what quantities and on what date.
> Don't claim a delay unless the promised date has passed.
> Keep it under 90 words. Draft only; don't send.

Check the order reference and the promised date. A specific request for quantities and dates is more useful than another request for an update.

### 2. Turn approved holiday hours into a door sign

**Template: Prepare a holiday hours sign**

> Write a plain door sign for [store and street] using these approved
> dates and hours: [paste hours]. Include the year and when regular
> hours return. Preserve every time exactly. Flag missing dates or
> conflicts before drafting. Don't invent a reason for a closure.

Compare every line with the approved sheet. Then use the [holiday hours checklist](https://ocally.co/blog/holiday-hours-sign-and-checklist.md) to put the same facts on the store's listings. Producing a sign doesn't update a listing.

### 3. Draft a reply to a difficult review

**Template: Reply without making up what happened**

> Draft a public reply to this review: [review without personal details].
> What we have confirmed: [facts]. What we have not confirmed: [gaps].
> Acknowledge the concern. Use this approved contact route: [route].
> Don't reveal customer details, promise a refund, blame an employee
> or claim we investigated something we haven't. Draft only.

Read it as the customer would. If the reply sounds defensive or promises an action nobody owns, revise it before posting. Keep your process for [requesting honest reviews](https://ocally.co/blog/how-to-ask-for-google-reviews.md) separate from resolving complaints.

### 4. Make a shift handoff someone can act on

**Template: Turn notes into a shift handoff**

> Turn these notes into a handoff for [store, date and shift]: [notes].
> Group them into customer commitments, stock issues and unfinished work.
> For each item show the next action, known owner and actual deadline.
> Mark an unknown owner or deadline as unassigned. Keep open questions
> separate. Don't turn a suggestion into an approved instruction.

Check that a customer collection hasn't become a delivery promise, and that a staff member mentioned in a note hasn't silently become the owner of the task.

## Use the facts you already have

ChatGPT can work with attached files and use connected tools where available [2]. The useful input is often small: the relevant rows, the approved policy or the two documents that disagree. Give each one a store name and an as-of date.

### 5. Find conflicts between two hours lists

**Template: Compare hours without choosing the winner**

> Compare these two hours lists for [store]: [list A] and [list B].
> Show only conflicts and missing dates, with both values side by side.
> Don't choose which is correct. Ask me to identify the approved source.
> Preserve time zones and note any overnight opening periods.

Verify the conflicts against both originals. Two matching copies can still be wrong, so keep the approval source visible.

### 6. Check the assumptions behind a proposed deal

**Template: Show the deal math**

> For this illustrative deal, calculate what one sale leaves after
> the costs listed here: [price, product cost, fee basis, deal cost].
> Show the formula and substitution. List omitted costs separately.
> Don't call the result net profit. Ask about ambiguous fee percentages
> before calculating. Don't recommend a price change.

Recalculate the result yourself. Use the [Black Friday offer worksheet](https://ocally.co/blog/black-friday-plan-for-local-stores.md) to include stock limits and checkout rules before announcing anything.

### 7. Turn a stock exception list into questions to investigate

**Template: Prepare the stock review**

> Read this dated stock exception list: [rows and column definitions].
> Identify missing information and questions worth checking per item.
> Separate recorded facts from possible explanations. Don't assume slow
> sales mean an item should be discounted. Don't recommend transfers
> without destination demand, transport cost and stock availability.

Check that a snapshot hasn't become a claim about today's shelf. A seasonal item, a display unit and an item kept to complete a customer's project may need different treatment.

### 8. Explain a sales change without inventing its cause

**Template: Prepare a store sales comparison**

> Compare these equal reporting periods for [store]: [data].
> Show sales and transaction-count changes separately, using the
> definitions provided. Show calculations. Flag returns, closures,
> missing days and changed definitions. List possible explanations as
> questions to investigate, never as proven causes.

Check totals against the original report. More sales alongside a new display doesn't establish that the display caused them. A useful summary tells you what to ask the manager next.

## Make instructions easier to use on the floor

### 9. Turn an approved procedure into a closing checklist

**Template: Make a checklist from an approved procedure**

> Turn this approved procedure into a closing checklist for [store]:
> [procedure]. Keep the original order where one step depends on another.
> Use one observable action per line. Preserve safety instructions.
> Flag gaps separately; don't invent a policy or remove an exception.

Walk through it with the person who closes. If an instruction depends on equipment or access the shift doesn't have, fix the procedure before handing over the checklist.

### 10. Find unanswered questions in an event plan

**Template: Check an event plan before announcing it**

> Read this proposed store event: [plan]. Identify unanswered questions
> about capacity, staffing, stock, weather, accessibility and customer
> promises. Distinguish launch blockers from optional improvements.
> Don't invent attendance estimates, partner commitments or approvals.

Name who will settle each blocker. An attractive description isn't a workable event if the person running it hasn't agreed to the hours or supplies.

### 11. Write a product description from verified facts

**Template: Describe the product you actually sell**

> Draft a short description for [product] from these verified facts:
> [specifications and intended use]. Explain who it suits and the
> relevant limitations. Don't add certifications, compatibility,
> health claims, local availability or performance claims not supplied.
> List anything a customer would need to confirm before buying.

Check specifications against the manufacturer and local availability against the store. Reusing another location's description can quietly introduce the wrong service or stock promise.

### 12. Make a weekly action list from manager notes

**Template: Prepare the next manager conversation**

> Use these notes for [store and week]: [notes]. Propose up to three
> actions, each tied to an observed issue. State the evidence, missing
> information, proposed owner and how we would check whether it helped.
> Separate agreed work from suggestions. Don't invent a target or
> claim an action was completed because its due date passed.

Ask the manager which action matters first. Keep the original notes so the summary doesn't replace the evidence or lose the context behind a decision.

## Decide how much work to delegate

For a one-off draft, start with a clear prompt and the relevant facts. For a task that needs files, research or connected tools, ChatGPT Work can produce a result for review [1]. For repeatable work, workspace agents can combine instructions, files and enabled apps [3]. Availability and setup matter; check your own account before building a routine around them.

Keep three questions visible: what can it read, what can it change, and what comes back to you for review? Connecting an app isn't proof that every store, field or action you need is available.

Try one task through to completion. Record the setup time, checking time, corrections and whether you used the result. Repeat it with the next real example before deciding that it belongs in the weekly routine. Judge the checked work, not how polished the first answer sounded.

## Keep the store context current

A helpful prompt can become unreliable when the facts around it change. A supplier changes the arrival date. A manager approves different hours. A product sells out while someone is preparing the sign.

Keep related instructions and source material together in a project when the work will continue [2]. Label documents with the store, owner and date. Replace superseded facts deliberately, and ask the assistant to point to the source behind a customer promise.

At several stores, decide who maintains shared policy and who approves each location's exceptions. The practical test is whether someone can trace a proposed action back to current facts and take responsibility for it. That matters whichever AI tool you choose.

## Questions

### How can I use ChatGPT for my small business?

Choose one job with a result you can check: draft a supplier email, turn an approved memo into a checklist, or compare two versions of your hours. Provide current source material, describe the output and ask it to flag gaps. Review the result before sending it or changing anything customers or employees rely on.

### Can ChatGPT work with my store's data?

ChatGPT can use files you provide and tools connected through available plugins. Access depends on your account, workspace settings and the permissions of the connected service [1, 2]. A sales spreadsheet you upload is a snapshot, not proof of a live inventory connection. Confirm which store and reporting period the data actually covers.

### Can ChatGPT do tasks instead of only answering questions?

Yes. ChatGPT Work can use files, plugins and approved tools to complete tasks and create results for review [1]. Workspace agents can also support repeatable work across connected tools [3]. Check what is available in your account and configure it deliberately. A prompt asking for a reminder does not prove a working schedule exists.

## Sources

1. [OpenAI: Get started with ChatGPT Work (checked November 4, 2026)](https://learn.chatgpt.com/docs/get-started-with-work)
2. [OpenAI: Use ChatGPT (checked November 4, 2026)](https://learn.chatgpt.com/docs/use-chatgpt)
3. [OpenAI: Building workspace agents for repeatable work (checked November 4, 2026)](https://developers.openai.com/cookbook/articles/chatgpt-agents-sales-meeting-prep)

**Where Ocally fits:** Kai, Ocally's AI growth agent, does this kind of work for every store, and anything that goes out waits for approval. [What Kai can do](https://ocally.co/capabilities.md), or [is Ocally right for this business?](https://ocally.co/evaluate.md).

## Next steps

- [Is Ocally right for this business?](https://ocally.co/evaluate.md): who it fits and who it does not, the cost, the risks, and how to start.
- [Pricing](https://ocally.co/pricing.md): customized, with a minimum of $500 per location per month.
- [Book a free consultation](https://ocally.co/consultation.md): 30 minutes on one store. Calendar: https://cal.com/ocally/30min.
- [See if we're a fit](https://ocally.co/fit.md): a two-minute check; an agent can submit it for its human, with consent.
- [Every doc](https://ocally.co/llms.txt): the index.

Questions a doc doesn't answer: onboard@ocally.co.
