Every freight manager wants the same thing on Monday morning: a short, honest picture of last week. How many jobs moved, which ones are late, where they are stuck, and what needs a decision.
In most companies that report is built by hand. Someone copies numbers out of a job register (the list of every shipment the team is handling), pastes them into slides, and hopes nothing was missed. It takes hours, it arrives late, and two departments' reports often disagree with each other.
An AI assistant can do most of that work in minutes. I have built these reports with AI and trained non-technical teams to build their own. This guide is the method, step by step, for someone who has never done it.
What AI can and cannot do here
An AI assistant, such as Claude or ChatGPT, is a program you talk to in plain language. You can give it a spreadsheet and ask it to count, sort, compare and explain, and it can turn the answer into tables, charts and even slides.
What it cannot do is know your business. It does not know that a job is "late" after ten days for air freight but thirty for sea, unless you tell it. It does not know which numbers your manager cares about. And it can make mistakes, especially with messy data. So the rule for the whole guide is simple: you set the rules, the AI does the counting, and you check the answer.
Step 1: Decide the five questions the report must answer
Before you touch any tool, write down what the report is for. A good weekly freight report answers five questions:
1. How many jobs did we open, move and close this week?
2. How many are late against the target time, and which ones?
3. How long do clearance and delivery really take, on average?
4. Where are jobs waiting longest, and why?
5. What does a manager need to decide this week?
If a table or chart does not help answer one of these, leave it out. A report nobody reads is not a report.
Step 2: Get the data out cleanly
Most teams already have the data, in an Excel job register or in their freight system. Export it to Excel and make sure:
- one row is one job;
- each column has one meaning, for example "Date arrived", "Date cleared", "Date delivered", "Status";
- dates are real dates, not text like "next Tuesday";
- the same thing is written the same way every time, for example always "Cleared", never "cleared", "CLRD" and "done".
This step matters more than any prompt. The AI can only be as accurate as the sheet you give it.
Step 3: Protect your clients' information
Before you upload anything, check two things:
- Use only an AI account your company has approved. A business account usually comes with stronger privacy terms than a free personal one.
- Remove what the report does not need. If the report only needs counts and dates, delete client names, contact details and bank details from the copy you upload. Never upload passwords or anything you would not want to see outside the company.
Step 4: Write your instructions once, and reuse them every week
This is the part that saves the most time. Instead of explaining the task from scratch every Monday, write one set of standing instructions, often called a master prompt, and save it. Most AI tools let you save instructions inside a project or a custom assistant.
Here is a simple example you can adapt:
You are preparing the weekly operations report for a freight forwarding team. I will upload this week's job register as an Excel file. Each row is one job.
Rules:
– A job is late if it has not been delivered within the target days in the "Target days" column.
– Clearance time is the days from "Date arrived" to "Date cleared". Delivery time is the days from "Date cleared" to "Date delivered".
– Ignore rows where "Status" is "Cancelled".
Produce:
1. A summary table: jobs opened, jobs delivered, jobs open, jobs late.
2. A table of every late job with its reference, mode, days over target and status.
3. Average clearance time and average delivery time, for air and sea separately.
4. Three short points a manager should act on this week.
Show the numbers you counted so I can check them. If any data looks wrong or missing, list it at the end instead of guessing.
The last line is the most important one. You want the AI to tell you when the data is bad, not quietly fill the gap.
Step 5: Check before you share
Never send an AI report without checking it. It takes five minutes:
- Spot-check three rows. Pick three jobs and confirm by hand that they are counted correctly.
- Check one total against the sheet. If the report says 42 jobs delivered, filter the sheet and see if it agrees.
- Read the "data problems" list. Fix the sheet, not the report, then run it again.
If the numbers are right three weeks in a row, you can trust the method. Keep spot-checking anyway.
Step 6: Turn it into the format your manager reads
Once the numbers are right, ask the AI to put them in the format your manager prefers: a one-page summary, a few slides, or an email. Keep the same layout every week, so the reader learns where to look and can compare one week with the last at a glance.
The five mistakes I see most often
| Mistake | What to do instead |
|---|---|
| Asking for "a report" with no rules | Write the five questions and the definitions first |
| Uploading a messy sheet and trusting the result | Clean the data in step 2, and ask the AI to list problems |
| Changing the instructions every week | Save one master prompt and only improve it on purpose |
| Sharing without checking | Spot-check three rows and one total, every time |
| Adding every chart possible | Keep only what answers one of the five questions |
Why this matters more than it looks
A weekly report is not paperwork. It is how a manager learns that a container has been waiting nine days, while there is still time to do something about it. When the report takes four hours, it arrives too late to help. When it takes twenty minutes, it arrives on Monday morning, and the conversation changes from "what happened?" to "what do we do?"
The bigger lesson is about who can use these tools. You do not need to be a programmer. You need to know your work well enough to write the rules down. That is a skill every good freight professional already has.
If you try this and get stuck, write to me. And if you want the sums done for you on duty, landed cost or demurrage, the free tools are here.