For 3-4 years now we have been seeing the rise of AI, AI chats, coding using AI agents like copilot, Anthropic Claude. There is also has been a lot of buzz in the market regarding job loss, AI take over. This all in some sense is true also and false too. It all really depends on where you stand.

Let’s first try to see what capabilities of AI are available, how organizations/people are using them. Having made a case for this, then we shall try to gauge how AI affects us, mostly from an Engineering point of view.

Our organization provides github copilot licenses to each of us developers/leads. So, while reviewing code 1-2 days back I saw one of the developers had implemented a Stored Procedure. Just to explain, there was a staging table, from which his Stored proc was reading the columns and there was a standard schema as per which he had to extract data from each column and load it into the final table. The code looked very structured as any AI coding agent would do.

I connected with him over teams and asked did you write this code. He said Yes!! I asked him did you use copilot, to which he said yes. So, I went on and asked, what did you provide as input to copilot. To which he said I gave copilot the schema document and prompted to provide SQL. I would say till this point the things are as expected, we want and mentor developers to make use of copilot to generate code.

Next, I asked him to explain the logic of SQL. to which he kept quite and then started stammering. I pushed him to explain the code further, but he just couldn’t. You see the problem here. Its very frightening to hear this as a lead/Engr Mangers/Tech Mangers, simply because developer is introducing a layer of uncertainty into the overall picture. The solution might work very well, no doubt about it but you are pushing to trunk something that you are not sure of.

So you see here is where I bring in the human accountability into the overall picture. While developing the code with copilot its very important to understand the code, understand the motive and different paths in the code. So humans are never out of picture in this case.