Ravindra Dastikop with ChatGPT
The situation
Every Friday, you combine sales figures from several spreadsheets. You ask AI to write a script that does it for you. You run it on last week’s files, and the total looks right. Would you trust it next Friday?
The old habit
When we learned to code, getting a program to run was a milestone. It’s easy to carry that habit into everyday work and treat the first successful run as proof that the job is done.
Unlearn this
Stop writing every line of routine code yourself by default. Describe the output you need, give AI a small sample, and let it draft the script.
Start learning this
Practise defining what correct means before you run the code. If the same sale ID appears in two files, should the script count it once, count it twice, or flag it for review? Find out what a repeated ID means in your workflow. Then turn that decision into a test.
Try it once
- In your first sample spreadsheet, enter sale A101 for ₹1,000 and sale A102 for ₹500.
- In the second spreadsheet, enter the same A101 for ₹1,000.
- For this exercise, assume a repeated ID represents the same sale. Before running the script, write down the expected total of ₹1,500.
- Run the script on both files. If it reports ₹2,500, it counted A101 twice. The script ran successfully, and your test caught the wrong result.
How will I know I am learning this?
I can design a small test, predict the correct result, and explain why a plausible-looking output is wrong. On the next script, try writing the test case and expected result before asking AI to draft any code.
AIYING · Accomplishing with AI


No comments:
Post a Comment