8 NOVEMBER 2026 · 17:00 GST · 5 MIN READ · DUBAI CAREER NOTES

How to use ChatGPT for supply chain operations: the Paste, Prompt, Prove rule

By Prudhvi Polavarapu, supply chain and e-commerce operations leader in Dubai

If you want to know how to use ChatGPT for supply chain operations, start with the work that eats your morning, not with the technology. Most operators I know do not need a robot to run a warehouse; they need two hours back from emails, summaries and spreadsheets, and a chat assistant can give them that if they use it with discipline.

Where does ChatGPT actually help in logistics and warehouse work?

It helps with language and structure, and it is weak at anything that needs your live numbers. That one sentence saves most of the disappointment. A chat assistant is very good at turning messy text into clean text, and much less reliable at arithmetic on data it cannot see or at facts it was never given.

In my own operations the useful jobs are all of the first kind. Drafting a polite but firm message to a carrier about a late pickup. Turning a page of shift handover notes into five bullet points. Summarising a long email thread so a manager can decide in one minute. Translating a supplier message and checking the tone. Writing the first draft of an SOP from a voice note. None of these touch your stock records, and all of them return real time.

A morning at the dispatch desk

Picture 7:40 on a peak-season morning. Eleven emails sit unread, three of them about delayed shipments, one from a supplier who has changed a delivery date without saying why. A good coordinator used to spend the first hour writing replies from scratch. Now the same person pastes the thread, asks for a three-line summary and a reply that confirms the new date and asks for the reason in writing, reads it, fixes one sentence and sends it. Ten minutes, not sixty. The judgment was still theirs. The typing was not.

I ran peaks of 5,000+ orders a day, and what I learned is that the scarce resource on those days is never effort, it is attention. Anything that protects the attention of your supervisors for the floor is worth trying.

What is the Paste, Prompt, Prove rule?

It is the habit I teach every team member who starts using AI at work, in three short steps.

Paste only what is safe and what is needed. Give the assistant the email thread, the handover notes or the SOP draft, with customer names, phone numbers, addresses and anything commercially sensitive removed. If your company has a policy on AI tools, follow it; if it does not, ask for one before you paste anything about customers.

Prompt with a role, a task and a format. For example: "You are an operations coordinator. Summarise this thread in three bullets, then draft a reply of under 80 words that confirms the new delivery date and asks for the reason in writing." Specific prompts get usable answers; vague ones get essays.

Prove before you act. Check every date, quantity, name and promise against your own system. The assistant writes fluently even when it is wrong, so a confident tone proves nothing.

The assistant writes the first draft. You own the last check.

Can I use ChatGPT for inventory and demand planning?

Carefully, and only as a helper. It can explain a method, such as how a cycle count works, draft the checklist, or help you write the formula you want in a spreadsheet. It should not be your source of truth for what is on the shelf. Stock accuracy comes from counting, scanning and discipline, which is why I wrote about inventory accuracy and cycle counts before writing anything about AI. If you want a sensible order of adoption, the same logic is in start with the boring AI: begin with repetitive text work, then move toward forecasting once your data is clean.

How do I start without creating risk for my company?

Keep a short list of allowed tasks and a short list of forbidden ones. Allowed: drafting, summarising, translating, formatting, brainstorming checklists. Forbidden until your company says otherwise: pasting customer personal data, pricing agreements, contracts, or anything under a confidentiality clause. The UAE takes technology seriously and the government publishes national guidance on AI through its official portal, u.ae, so it is worth reading what applies to your sector and following your own employer's rules.

Then measure. Pick one task, time it for a week before and a week after, and keep the result. When a manager asks what AI is doing for the team, "it saved the dispatch desk about five hours a week on carrier emails" is a better answer than any slide.

The operators who will do well with AI in the Gulf are not the ones who trust it most. They are the ones who paste carefully, prompt clearly and prove everything before the truck leaves.