I used IA for a project, which has nothing to do with me. For this problem belongs to the Support team. So there is this customer who emailed us hisPO. I entered the PO into our system, and yet this customer kept emailing me that the SOconfirmation email was wrong. We went back and forth, no, I went back and forth with him with elaboration, which I pointed out which item he had a wrong pricing. He basically just repeated his statement again and again, without further reasoning. I got frustrated, and had Emily checked the SO for me. She did not want to check it, because one glance, she knew I was correct. Then I decided to go to David. After David heard me while searching around. He confirmed my statement that the customer quoted a product incorrectly. He then wrote an email to the customer.
The next morning, while I was picking, he came and asked me if I saw the customer's reply. I said yea, the customer agreed with our conclusion. But he said that another person from the customer chimed in, saying that it was our fault through a pricing spreadsheet, which we shared with the customer. The customer asked us to honor the wrong data this time. I was kind flabbergasted, and David said that he now agreed with me that the customer knew the problem and set the trap for us.
Internally, the support team was trying to correct the mistake, but they found it's really hard. I kept hearing David and Ashley saying it's really hard. I thought it was a great opportunity to try out the iA. I started on Sunday for 1.5 hours, downloading the data, rearrange it, engaging the iA a little bit. Just to understand what we were dealing with. Then I spend 3+ hours last night trying to get it going using ChatchatGpt. From the start, iA gave me three ways to go, I forgot the first one, second one was powerQuery, since it was an excel sheet. The last option was iA could write a VBA for me to automate.
At first, I used powerQuery to do text to column. But for some reason, it failed, because some pricing appears like this: 231,356,087. power Query reconize some of this as 1 number, rather than 3 different numbers, although the delimiter is comma. I eventually solved this issue with manual operation using excel's build-in function text to columns. Then I started to use powerQuery to split the weight, this one was very easy. But for some reason, iA told me to unpivot while highlihgting the SKU column. Then I got a data range called long format, which is a name I learned today from dipsik.
" In the context of data analysis and statistics, a "long-range dataset" or "long format" refers to a dataset where each row represents a single observation or measurement, with multiple observations or measurements for the same entity (like a person, place, or thing) stacked in separate rows. "
But I feel it might be helpful down the road, for some reason unknown to me. Anyway, the rest of the 3+ hours, I got lazy, so I just tell ChachatGpt to give me a VBA code to automate everything, like give me a converted weight column, since spreadsheet does not recognize mg, g, ug, we have to use one unit to really compare the weight. It was not smooth sailing, iA gave a few VBA codes, but some failed, some work partially, until I thought it worked. Then I asked if we could make the attributes, like price, price2, and size.1 and size.2 to correspond with heaviest and most pricy and second heaviest, and second pricy. That was the one task I could not finished last night. iA kept gave me VBA codes, I kept trying to give it feed back on the codes. Until 1230 am, I found a huge error. Previously I thought the converted weight was correct, I am not sure if it's because all those VBA codes for correct re-labeling attributes, or it was just wrong beforehand. The ug had failed to convert to mg. That was the last straw of the night. I gave up and turned in.
I tried it again this afternoon at work. I used dipsik at first and was really impressed by the interaction with it, especially when it guided me to debug the problem of converting ug to mg. Though I tried and tried, and dipsik was extremely patient, I myself lost patience and found a cheating method to bypass the error. Even now I am still wondering just why can't I convert ug to mg...
But after the bypass, I could not engage with dipsik again. I felt the instruction was kind too simple and bland. I think it must have something to do with my prompt. Anyway, I jumped back to chachatGpt. it has the same problem like dipsik. The instruction seems very unclear and vague, I had a hard time to follow it. I found last night's chat, it has a lot VBA codes. I decided to follow that chat. Also I forego the manual process, I asked iA for VBA code like last night. This time, he gave me about 3 wrong codes, I told him each time, what's wrong, what the error message was saying, which code did the debug highlight. Eventually I got a code that worked!! That's another 3 hours passed. The finality was to pivot the long form range. After a tweak of the datatype, change the Value column of mg, ug and pricing to text. I got the table I wanted. I was so happy I went and told Emily. Emily even came over to take a look. Though I know it's a polite thing to do, I appreciated her effort.
Anyway, I still need to check the sheet tomorrow to make sure it is correct. lol
8 hours working on a sheet is really kind dumb. But considering that I know very very little about power query, VBA and other stuff. The amount of things iA did was amazing. It was not in anyway perfect, in fact, it did a lot of trial and error. Of course, there's problem with the prompts that I gave. But if it just me, I know it will not work with 100 hours. iA really is quite amazing under the right hand. I think.