Buying ChatGPT Is Not an AI Strategy
A customer pings WhatsApp: “Can I have the quote today?” Sales is waiting for the boss to approve a discount. The boss is in a meeting; the phone will not stop. Half an hour later the customer drops a voice note: the spec has changed a little. Sales opens last year’s Excel and a pile of old PDFs, hunting for a similar sheet. After the discount is approved, customer service copies SKU, price and delivery address into MYOB or the old system, cell by cell. One extra zero, and tomorrow’s reconciliation turns into an argument.
Typing can get faster. The customer is waiting for a quotation with a discount, a lead time and terms. That sheet still goes through old files, the owner’s calendar, then into the system. Faster typing does not skip those three waits.
A Hong Kong trading or wholesale firm can run that process a dozen times in a day. When owners buy ChatGPT Team or Copilot, they usually mean: let people work faster. Three months later, emails and meeting notes really are quicker. The quote is not. Some then assume the team cannot write prompts, and buy a half-day of training. Monday morning, the customer still chases, the boss is still in a meeting, someone is still copying cells.
These days we rarely walk into a company where nobody uses AI. Marketing polishes English. Admin drafts notices. Warehouse asks how to roster. Someone on a free account at night pastes a customer thread and asks how to reply. Each person has their own login and their own habit. Quoting, posting invoices and chasing orders still run on Excel, email and WhatsApp taped together. The tool reached the desktop. It has not reached how the company ships or posts the books.
Typing got faster. The company did not.
Fixing a sentence or drafting a WhatsApp reply helps the person at the keyboard. It does not decide whether a quote leaves the building today, or whether an invoice hits the ledger. Ask three colleagues the same price and you can get three answers: different prompts, different accounts, someone still on last year’s list. However fluent ChatGPT sounds, it cannot see the customer in your WhatsApp, the fields MYOB expects, or the discount only the boss can sign. People are still waiting. Hands are still copying.
That is why many firms take a wrong next step: they train harder on prompts and expect the quote to move by itself. A prompt can make a follow-up email nicer. The customer is still waiting for the quotation, not a fluent email.
Another seat and an AI strategy are two different purchases. A seat is a login. A strategy is being able to say which hours and which deals you want to move this year. If you cannot say that sentence, staff will use AI to speed up their own paperwork. The company stays as slow as before.
The quote waits in a few short places
Count it yourself. The customer asks on WhatsApp, sometimes with a voice note, a photo, and an old order mixed together. Sales hunts through old Excel and old mail for “the last similar sheet,” then guesses today’s cost and lead time. A discount waits until the boss leaves a meeting. A PDF goes out. The customer confirms. Then customer service moves the same words into the order and the books. One miss, and warehouse and accounts shout at each other the next day.
Which stretch takes longest differs by company. Sometimes it is approval: five meetings in a day. Sometimes it is finding the old price, scattered across three folders. Sometimes it is posting: one person, a PDF and a system, copying all afternoon. ChatGPT at the next desk can draft “we are still working on it, please wait.” The customer still has no quote.
The same pattern shows up in matching documents. A supplier invoice, your PO, a delivery note: three papers, three layouts. Eyes miss a line and the count goes short. Asking AI to summarise an email does not reconcile that pile. Slow here, as with quoting, is not typing. It is scattered files, waiting, and moving numbers by hand.
A strategy is being able to name the first process
You do not need the whole company to change at once. An owner can ask three plain questions. Which line of the business should move this year? Where does that work wait in a day? Is it waiting for someone, hunting a file, or copying cells? The answer is often short: enquiry → quotation → discount approval → order → books. Pick the one you actually want faster this year. Move that one first.
How do you know it moved? How long a quote takes now, how many leave in a day, how many posting errors in a week. With those numbers you can see whether anything changed. Without them, the project becomes “everyone should use AI,” and afterwards you still cannot say where the time went.
Name the process first, then talk tools. Some stretches an off-the-shelf product can hold. Some need your WhatsApp, Excel and accounts to connect. That choice can wait. Rush a catalogue and add seats, and each person only gets faster at their own desk. Shipping and posting stay as they were.
Frasertec Limited has spent twenty-odd years with Hong Kong SMEs on systems and automation. We start with where people wait, copy cells, and mismatch numbers. Which licence you bought can wait. If you already have ChatGPT and cannot say how it serves this year’s figures, bring the slow quote or the matching error. That is enough. We have written separately about staff using public AI. This piece is the step before: which process you want faster.
