It had been almost I year since I last saw L and this meeting was on the brink of not happening. But we did get coffee at the place we both love and as always there was much that I learned in the couple of hours. She mentioned working with an engineering team that has veered between being hostile and rejective about using AI to ramp their productivity to being in public denial but using it surreptitiously. The later almost did not make sense because in her company there is a push for AI adoption all the way from the top. So it would only help rank and file SDEs to toe the line.
Recently after a minor release was pushed out, over half the users were locked out of the system causing a terrific amount of chaos, fire-fighting and blamestorming. L was in the thick of it all, trying to assess root cause, implementing checks to prevent it and non-stop damage control. That was the week leading up to our coffee chat. She look completely exhausted. It turns out that the culprit was some AI generated code, being reviewed automatically (another AI I presume) and the human being the final backstop.
The human in question was one of those tenured engineers who hated AI with his guts and did not believe it was able to provide any measurable benefit to the likes of him. He had apparently reviewed this bit of code pre-release but not caught something that seemed to be pretty obvious. L chalked it up to him being disenchanted and disengaged ever since the brass started to make it near mandatory for developers to use the set of AI tools the company had provided to do more with less. Then the question was if B had infact been delinquent in doing his job as the code reviewer or it he had relied too much on AI having seen prior releases had gone without a hitch.
The bigger question for L as she moves into a new role with much larger remit is how she should message the use of AI for her team which will include technical and non-technical folks. Her own preference is to leave it up to the person and only measure outcomes and value delivered, not interfere with the process but she does have to drive the AI adoption mandate. Clearly, she did not come upon any clarity that afternoon as we chatted. I shared this story with her in hopes that other voices might help her with it.
Comments