Note: The views expressed in the post reflect the AI capabilities at the mentioned time period. The capabilities have advanced since then.
The first time I used AI (ChatGPT) for problem solving was in Dec 2022.
We had an enterprise product with low customer adoption.
The root cause was deemed to be poor onboarding experience. There were too many manual steps needed to onboard the product. Invariably something would go wrong during onboarding and customer would either file support tickets or abandon the thought of using the product.
When I looked at the steps (written in English) it looked lengthy, cumbersome and occasionally ambiguous.
I wondered why we were asking the customer to do this when we could automate it all?
I knew it could be automated. But the catch was that it needed to be in Windows Powershell as the customer admin base was entirely Windows.
All my scripting experience was in Linux / Unix shell.
I didn’t have the time and enthusiasm to search through Stack Overflow to get Powershell snippets to compose the utility I needed.
That’s where ChatGPT came to my help.
I knew exactly what to ask for each line of the script because I had the equivalent bash script in my head.
I could test and verify the responses and tweak them to my needs.
In a short time I had the utility ready for demo.
Since it was a working demo, it was scoped in for the subsequent development cycle. (To complement the code and make a case, I wrote a Wiki page with convincing arguments and data.)
A developer still had to work on it to iron out wrinkles, handle corner cases I didn’t think of, add to the security model (signing) and getting it production ready.
We shipped it and customers found it useful. It helped us during customer calls because we had built in a diagnostic mode in it.
It was a win for us and the customer.
Would this have happened if AI was not there?
May be. May be not.
Could it have been done only by AI without all the thinking, validation, and refinement we did?
May be not.
Did AI help us?
Absolutely!