There are also agents that we can use for coding. There are many different agents that I can chose from, and I suspect there are other folks that have other/more options. And apparently, the company is testing/developing more.
We have about a dozen different versions to choose from, with about a 60/40 split between GPT and Claude. And those are just the copilot agents for coding. They don't include the couple of versions available for MS Office and/or desktop use.
One of the things that they ALWAYS stress, is that anything that you do with AI needs to be reviewed. I am in IT, but I'm not always an early adopter. As a matter of fact, this past week, I used AI for the first time for a few different things. The first thing that I did was take some feedback that I had for someone (positive, think "thanks, you're awesome" kind of thing), and run it through AI. I did it 2 or 3 times. The first couple of times, I didn't like what it spit out, but it eventually spit something out that I thought was an improvement, although I still made a couple of small changes.
I was then asked for a slightly different sort of feedback (to go with a year end review) and used it again for that. Again, it took me several shots and editing for me to be happy with the results.
Claude Sonnet 5.5
But what really blew me away was when I finally used it for something technical.
We have a script that we run on a system that is designed to push similar, but different information to 2 other systems. Once you push the info to the two systems, if I gave a person the script from the first system, and then the results of the script from the 2 target systems, the average thinking person would be able to see how the things fit together.
So we have a thing A-B that begets 2 different things that are 1-A and 2-B.
I have a need, to take ~1200 of the 1-A and 2-B things, and then reverse engineer them into one really long A-B.
I created 5 files.
file-1 with A-B
file-2 with 1-A
file-3 with 2-B
file-4 with 1-E
file-5 with 2-F
I explained that file 1 causes the creation of files 2 and 3 and then said, please take files 4 and 5 and create a file 6 similar to file 1.
Bam! It thought for a few secs and then spit out a correct file-6.
I then took the 1183 items similar to A from system 1, and the 1183 items similar to B from system 2, and pasted those into file 4 and 5. I said "I updated file 4 and file 5. Please repeat the previous process to create a new file 6 with the new data.
It took a minute and 20 seconds and spit out a file. With the file, it explained what it did including saying "I checked one block and confirmed that it is accurate."
I said "please repeat the previous action, but check every block".
It did it again and that time only took 23 seconds.
It was very impressive. I've got a colleague that is apparently number 4 in the company as far as token use. He's loving it. He said it's great for data analysis. For instance, writing formulas in Excel or performing complex data analysis in spreadsheets.
I can see more of it in my future. I hope to hang on to work as long as I can until I'm ready to retire which I assume is ~9yrs out. I think that learning to use and using AI is a requirement. Folks, especially in IT that aren't or don't use it, will be out of the loop and won't last indefinitely unless they have some task that just cannot be done by AI (not a lot of that in IT).
Quote:
Originally Posted by gacook
I'd caution overuse of AI in RFP responses. I recently let a contract and it was painfully obvious which proposals were written by AI and NOT properly edited/vetted after AI spit out whatever it spit out. Goes without saying, those props didn't do well.
|
Where I work they are pushing AI very hard. We have our own in house copies and a whole AI division (and have for at least 6-7 years, I think). There are several different AI options available at work, stuff that's specifically for working in Microsoft Office (emails, spreadsheets, word, etc...).