How an exporter scored three export markets
An exporter needed to rank three new export markets. No agency brief, no commissioned study. The criteria came from him, five of them, and the tool did the reading and the scoring. The work took a brief, a scoring table, and a source link on every claim.
What can I use AI for in my business?
Start with the decisions you keep postponing because the reading behind them is too big. For an exporter, that is market selection. You know your product, your yields, and what a buyer is worth. What you do not have is two weeks to read about duty lines and port costs in three countries you have never shipped to.
That reading is what the tool is good at. It sorts, lists, scores and drafts. It produces a table you can argue with. The same shape covers a supplier comparison, a tender you are deciding whether to bid on, or a first screen of distributors in a new state.
The trap is treating this as a search box. A search box gives you links. A scored comparison across criteria you chose gives you something you can take to a decision.
The criteria came before the prompt
The exporter picked five tests before anyone opened a chat window: demand, local capacity, landed cost, duty, and setup friction. Each one came from running the business, not from a research template.
Demand, because the grade a market buys decides whether you can serve it at all. Local capacity, because what the destination country already produces tells you whether you are supplying it or competing with it. Landed cost, because the number that matters is the one at the buyer's gate, not at your gate. Duty, because a tariff line can erase a margin you spent a year building. Setup friction, because the number of approvals before the first container moves decides whether the market is worth entering at all.
The five tests were the five things the owner already weighs when he says yes or no. Nobody had to teach him what mattered. He had to write it down.
What did the brief actually ask for?
The brief named the role, the context, the task, the output format, and the constraints. That is the structure worth learning once and reusing forever. Role: an export market analyst working for the exporter. Context: the plant, the grades it makes, the ports it can reach. Task: score the three candidate markets on the five criteria. Format: one row per country, one column per criterion, a score, a one-line reason, and a source link for every claim.
The format line does most of the work. Ask for prose and you get an essay you cannot check. Ask for a table with a source column and you get something an owner can audit in minutes. The constraint that matters most is this: no claim without a source you can open.
Where the owner overruled the output
An owner reads the table against what he already knows, and the first rows to go are the ones he can contradict from memory. That is the normal first move on a scored comparison.
You check four things. You open the source link behind every cost and duty figure and read the page yourself. You check the date on each source, because a tariff schedule from three years ago is a wrong answer dressed as a right one. You test the landed cost against a freight quote you can actually get. And you read the demand claim against what your own buyers and brokers tell you, because a trade statistic and a live order book are not the same thing.
Where the output is vague, you say so in the next turn and ask for the specific number or the specific regulation. Where a figure is wrong, you correct it and ask for the table again on the corrected input. Your judgment does not become less important. It becomes the only thing left to do, because the reading is off your desk.
This is the kind of work that happens in session four of Module 1 at A.D.A.P.T. AI Coaching, where an owner leaves with "a research brief on your own market".
What else do owners ask before trying this?
Do I need clean data before I start?
No. This work runs on public information and on what you already know about your own business. Your own numbers come in later, when you test the landed cost against a real freight quote. Start with the criteria, not with a data cleanup project you will never finish.
What if the output gets a duty rate wrong?
Assume it will get something wrong, and build the check in from the start. That is why the brief asks for a source link on every claim. You open the link, read the tariff line yourself, and correct the table. A wrong figure you caught costs nothing. A wrong figure you accepted costs a shipment.
Will the same brief work for a domestic market screen?
Yes, with the criteria changed. Swap duty and landed cost for credit terms, transport time and who is already selling there. The structure holds: your tests first, one row per market, a score, a reason, and a source link on every claim you will have to defend later.